GMAT Focus

The 12-week GMAT study plan

From arithmetic to test-day ready. Start at Week 0 if GMAT Focus itself is new to you — fifty minutes on format, scoring and how to record practice, with no questions. Concepts build through weeks 1–8 with four full-length mocks calibrating you along the way, and weeks 9–12 convert accuracy into speed.

Every week below expands into its topics, and every topic carries a brief written for a candidate who is already competent: what the exam actually does with the material, where the disguises are, and what "done" looks like. Read the whole map or work one week at a time.

• 0 Get to know the exam sections, scoring, how it adapts, how to track practice START HERE 50 min read

  Start here if GMAT Focus itself is new to you. Fifty minutes on format, syllabus, scoring and how to record practice so it can be reviewed later — no practice questions. Skipping this week is defensible if you have already sat the exam; skipping it and then discovering in week 6 that you have been recording nothing is not.

  Exam skills

  What GMAT Focus actually contains

  Three sections and only three: Quantitative Reasoning, Verbal Reasoning, Data Insights.

  Quant is problem solving only — no Data Sufficiency paper sits there any more. Verbal is Reading Comprehension and Critical Reasoning; Sentence Correction is gone, and no amount of grammar drilling will earn a mark. Data Insights absorbed the old Integrated Reasoning section and Data Sufficiency with it, so it is now the only place a DS item can appear — still a fixed set of five answer choices, but sitting in a section whose other questions look nothing like it.

  The other four DI formats, which you should be able to name on sight:

  • Multi-Source Reasoning: tabs of text or tables, several linked questions
  • Table Analysis: a sortable spreadsheet, one yes/no judgement per row
  • Graphics Interpretation: a chart with drop-down statements to complete
  • Two-Part Analysis: two columns, one pick from each

  DI is also the only section with an on-screen calculator; Quant gives you none.

  The disguise: identical underlying content reappears across sections in different clothing. A rate or a ratio is a problem-solving item in Quant, a DS stem, or a two-part row in DI. So classify an item by its interface before its content — the format decides how much you must commit to before you see the answers.

  Strong candidates lose here mechanically: they build a quant-and-verbal plan, leave DI for later, and meet five unfamiliar interfaces cold.

  You are ready to move on when

  • You can name all three sections and, for each, list its item types without looking them up
  • Shown any Data Insights question, you can identify its format before reading the question text
  • Your study schedule lists Data Insights as a section alongside Quant and Verbal, not as later revision

  Exam skills

  Scoring, and what "adaptive" changes about strategy

  GMAT Focus does not score you on how many you got right. It scores you on the difficulty level at which you were still answering correctly, estimated one question at a time — and each section runs its own estimate, resetting at the section boundary.

  Three consequences do the work.

  Early questions move the estimate most. Before the algorithm has evidence, your answer is the evidence; after twenty responses, one more barely shifts anything. The first few questions in a section therefore carry more weight than the last few — the reverse of the instinct to warm up.

  Difficulty is feedback about the algorithm, not about you. A stretch of hard questions means it is probing near your ceiling; a stretch of easy ones means the estimate slipped. Neither tells you how you are doing, and neither is a reason to change pace. Strong candidates lose points precisely here: they read hard questions as failure, slow down to be careful, and reach the final third of the section short of time.

  No partial credit, no penalty for wrong answers. A blank and a wrong answer score identically, so an unanswered question is a pure loss. And the cost of a stuck question is not that question — it is the two later ones you rush. Set a per-question budget from the section's length and the clock, and abandon at it.

  Worth memorising: you score the level you can handle, not the count you got right. Chasing "how many correct" misjudges every timing and abandonment decision.

  You are ready to move on when

  • You can state your per-question time budget for a section before starting it, and name the point at which you will guess and move on
  • In a timed practice set you left no question blank, and can identify at least one question you deliberately abandoned
  • After reviewing a section you can describe its difficulty trajectory without treating it as a verdict on your performance

  Exam skills

  Section order, and the bookmark-and-review tool

  You choose the order of the three sections before the clock starts, and you place your single optional break in the same step. This is the only strategic decision you make in advance and cannot revise, so settle it now and rehearse it in every practice test rather than deliberating on test day.

  Two variables matter. First: lead with your strongest section for rhythm, or with your most fragile one to get your freshest hour? Second, and easier to overlook: the break can only fall between sections, so the order you choose also decides which two sections run back-to-back. Put the break in front of the section you would most resent starting tired.

  Within a section, bookmarking is free and carries no scoring weight. The review screen lists every question as answered or not, flagged or not — and the moment you leave a section it seals. That screen is your last chance to change anything.

  Three consequences:

  • Clear blanks first. Unanswered questions score as wrong, so a guess made in the final minute is strictly better than nothing.
  • Treat flags as a worklist, not a parking lot. Flag on a specific doubt ("I assumed $x>0$"), not on unease, or the review pass becomes a re-read of questions you already handled well.
  • Do not reopen questions you finished confidently. The second-guess tax is where strong candidates quietly lose points.

  Pacing anchor: each section is $45$ minutes against roughly twenty questions — about two minutes each. A four-minute bank buys two hard questions.

  You are ready to move on when

  • You fixed a section order and break placement before a practice test and did not reconsider it mid-exam
  • You reached the end of every section with zero unanswered questions, having used the review screen to find them
  • You left each review pass having touched only questions flagged for a stated reason, and left every other answer as first written

  Exam skills

  How to record practice so it can be reviewed

  The official score report gives you section accuracy and timing. It will never tell you that you dropped a rates question because you read "per hour" as "in total". Your own log is the only instrument that can, and it works only if you write in it at the moment you answer, not after reading the explanation. Reasons written afterwards are reconstructions, and they are reliably kinder than what actually happened.

  Four fields carry nearly all the value:

  • Source ID, precise enough to reopen the question without the answer showing
  • The option you chose and the specific thing that made it attractive
  • The split: the two options you genuinely hesitated between, and the word, number or assumption that broke the tie
  • Error class: did not know the rule; knew the rule and chose wrong anyway; knew it and ran out of time

  That last distinction is the one most people collapse, and collapsing it is how strong candidates lose points here. They tag everything "weak in ratios", re-read solutions until the questions look familiar, then meet the same item in a mixed set and repeat the error. Familiarity is not retrieval; only a blind re-attempt is. Log a re-attempt date and honour it: reopen the source, re-answer without notes, then compare reasoning rather than answers, so you can read $Accurac{y}_{retry}$ beside $Accurac{y}_{first}$.

  One habit to refuse: recording a topic name as your diagnosis. It names the subject, not the failure, and nothing can be done with it.

  You are ready to move on when

  • For any recent miss you can state the exact pair of options you hesitated between and the detail that decided it
  • Every log entry carries a source ID precise enough to reopen the question without the answer showing
  • You can sort your log by error class and separate content gaps from execution errors from timing
  • You have re-attempted logged questions blind after a delay and recorded the second result beside the first

  Open week 0 on its own page →

• 1 Foundations in arithmetic, CR & DI ACCURACY FIRST 10–15 hrs

  Accuracy before speed, without exception. Everything in weeks 9–12 is built on the assumption that when you are not rushed, you are right. If that is not true yet, speed work is premature.

  Quantitative

  Arithmetic fundamentals and estimation

  GMAT Focus uses this topic as scaffolding, not as a destination. A bare computation is not the deliverable. What is tested is controlled approximation: deciding whether exactness is required at all, and knowing the error you are carrying when it is not. In Data Sufficiency, the skill is stopping the moment an expression's size, sign, or parity settles the question.

  Disguises: answer choices spaced so widely that one rounded division eliminates three of them; "must be a multiple of" questions where only divisibility structure matters; digit and units-digit questions; and rate, percent or weighted-mix word problems whose numbers are ugly on purpose.

  Three decisions separate the correct answer from the tempting one. Measure the answer choices before computing: if they differ in the leading digit, estimate; if they differ in the second, compute exactly. Track the direction of your approximation — whether it over- or under-states the true value is what picks the choice, not its size. In DS, stop as soon as the statement can no longer change the answer.

  The mechanical way strong candidates lose points: chaining rounded intermediates so error compounds across steps, then finding two choices sitting inside their own uncertainty band. Round once, at the start, or not at all; carry exact values forward.

  Worth having in memory: for consecutive integers beginning at 1, the sum of the first $n$ terms is $\frac{n(n+1)}{2}$.

  You are ready to move on when

  • Given the answer choices, you state estimate-or-exact before writing any arithmetic
  • You can say of any approximation you make whether it over- or under-states the true value
  • On integer and divisibility questions you reach the answer using units digits or parity alone, with no full computation

  Quantitative

  Fractions, decimals and percents as one system

  GMAT Focus does not ask you to convert $\frac{3}{8}$ into a decimal. It asks you to notice that $\frac{3}{8}=37.5\mathrm{\%}$, that the answer choices are written as fractions, and that multiplying by $\frac{3}{8}$ is therefore cheaper than multiplying by $0.375$. The topic is one quantity wearing three costumes, and the exam tests whether you pick the cheapest costume for the operation in front of you.

  The disguises: data-sufficiency statements give one value as a fraction and another as a percent; a word problem says "reduced to $\frac{1}{4}$" when you expect "reduced by $\frac{1}{4}$"; a chart labels in percent while the question asks for a count.

  Two decisions decide most questions. First, choose the representation that makes the next operation one step: percent for a quick multiplier, fraction for exact cancellation, decimal only for ordering or adding. Second, successive percent changes multiply, never add — a rise of $20\mathrm{\%}$ then a fall of $20\mathrm{\%}$ is $1.2\times 0.8=0.96$, a $4\mathrm{\%}$ net fall.

  The mechanical loss is precision drift: rounding $0.375$ to $0.38$ mid-chain, then choosing the answer choice built from that error. Keep one exact form until the final line.

  Have in memory: $\text{new}=\text{old}\times (1\pm r)$ with $r$ as a decimal, so $0.4\mathrm{\%}$ means $r=0.004$.

  You are ready to move on when

  • Convert any fraction with denominator 2 to 12 to its percent form and back without long division
  • Rewrite every percent change in a problem as a single multiplier before doing any arithmetic
  • Compare a mixed set of fractions, decimals and percents by converting to one form only, and refuse any answer that depends on a mid-calculation rounding

  Quantitative

  Ratios and proportions

  GMAT Focus rarely asks you to define a ratio. It asks you to translate a verbal relationship into a scale-free comparison, then decide what absolute anchors the scale. The exam's favourite move is to give you a ratio and one real quantity, or to hide a ratio inside a rate, a mixture, or a percentage.

  Disguises: 'for every 3 ... there are 5 ...', 'per', 'out of', 'the same proportion', and any rate (miles per gallon, students per teacher). In Data Sufficiency, the question may ask for a fraction of a total, a ratio, or an absolute; the statements will mix ratio and absolute information, and you must know which is which.

  Three decisions separate correct from tempting. First: is the ratio part-to-part or part-to-whole? Boys:girls is part-to-part; boys:students is part-to-whole. Second: do not treat ratio numbers as counts. Write $A=mk$, $B=nk$. Third: check what is asked — a ratio statement alone cannot yield an absolute; an absolute alone cannot yield a ratio unless the other part or the total is known.

  The mechanical loss: reversing the order ($B\mathrm{/}A$ instead of $A\mathrm{/}B$) or adding parts to form a total when the ratio is part-to-part but the question asks for a part-to-whole fraction. At this stage, write the multiplier explicitly rather than doing ratio arithmetic in your head.

  Memory: if $A:B=m:n$, then $A=mk$, $B=nk$, and $A\mathrm{/}(A+B)=m\mathrm{/}(m+n)$. Use it whenever you need to convert between a ratio and actual counts or a fraction of the total.

  Illustration: flour:sugar $=5:2$ and 15 cups flour gives sugar $=2\times (15\mathrm{/}5)=6$ cups.

  You are ready to move on when

  • Given a word problem with a ratio, you can label it part-to-part or part-to-whole before writing any equation
  • You can convert any two-term ratio $m:n$ into $mk$ and $nk$ and use one absolute to solve for $k$
  • In Data Sufficiency, you can decide whether a statement supplies an absolute or only a ratio, and whether that is enough for the question asked

  Verbal

  Critical Reasoning: finding the conclusion

  On GMAT Focus, the conclusion is the claim the rest of the passage exists to support. The exam rarely labels it. It appears as a recommendation, a prediction, a causal claim, or a rebuttal to an opposing view, and it can sit first, last, or mid-paragraph.

  The two decisions that separate right from tempting:

  Is the statement doing the supporting or receiving it? Insert "therefore" between candidate pairs. If statement A therefore statement B reads as a coherent argument, B is the conclusion. This test works only when both statements belong to the same argument.

  Whose claim is it? A vivid opposing view is not the author's conclusion. The author's conclusion is what the author wants you to accept; opposing views are premises in a rebuttal. Also, distinguish the main conclusion from an intermediate conclusion, which is supported by other premises and itself supports the final point.

  You will lose points mechanically by grabbing the last sentence, or the most extreme sentence, without checking what supports what. In an argument ending "Clearly, the policy will fail," that sentence is often the conclusion—but verify by asking what the author would say if challenged with "why?" The answer reveals the premises.

  Compact illustration: "The new toll will reduce congestion; after all, similar tolls cut traffic elsewhere." The conclusion is the first clause.

  You are ready to move on when

  • You label the conclusion in the stimulus before reading the question stem, and your label matches the credited answer's reference
  • You can state which sentence would survive if you deleted the others as support, and you name that sentence as the conclusion
  • You spot and discard an intermediate conclusion without being distracted by its confident tone or position

  Data Insights

  Reading tables and charts without assuming

  Data Insights rarely asks you to construct anything. It hands you a table or a chart and asks for one value, one comparison, or one yes/no condition — which means almost every point on this topic is lost the moment you start reading the picture instead of the labels.

  What is actually tested is the chart's apparatus: axis ranges and tick spacing, units, the footnote that redefines a denominator, and whether a series is a count, a rate, a running total, or an average. Graphics Interpretation gives you a handful of drop-downs, each usually settled by one number read off correctly. Table Analysis gives you a sortable table and a condition — sorting is a tool you apply, not a property of the data, and a table that arrives unsorted punishes anyone who assumes otherwise.

  The disguises: a non-zero baseline; a second series on a right-hand axis with its own scale; two legend entries a shade apart; a 'total' row drawn from a different population than the rows above it; a chart in thousands with options in millions; adjacent columns expressed as percentages of different bases; an x-axis with a missing year.

  Three decisions separate right from tempting. Read axes, units and footnotes before any comparison. Establish whether the question wants a level, a change, or a ratio — and against which base. If the chart shows shares, recover counts before comparing groups.

  Strong candidates lose here by answering from visual impression: the taller bar, the steeper slope, the obviously bigger group. It bites hardest on "largest increase" questions where the x-axis is unevenly spaced.

  Keep one relation: $\text{segment value}=\text{segment share}\times \text{category total}$, valid only when the share is taken of a total the chart actually states.

  You are ready to move on when

  • Before answering the first sub-question of any chart or table, you can state its units, axis range and any footnote that changes a denominator
  • For each wrong answer you chose in a practice set, you can name the specific unstated assumption it required you to make
  • You sort tables and convert shares to counts deliberately rather than by default, and can point to the row or tick mark you read each answer from

  Open week 1 on its own page →

• 2 Number properties, CR assumption & DS PREDICTABLE LOGIC 10–15 hrs

  Two of the exam's most rule-driven areas in the same week, deliberately. Number properties and Data Sufficiency both reward a candidate who can enumerate cases rather than one who can calculate quickly.

  Quantitative

  Number properties: parity, primes, factors

  Parity on GMAT Focus is rarely about a single integer. It is about expressions: whether $n^2+n$ is even, whether $p+q$ can be odd, whether a product of consecutive integers is even. The structural fact to run first: adding or subtracting any even number preserves parity, so strip the even terms and look at what remains. $n(n+1)$ is always even, and so is $n^2-n$.

  Primes have one fact doing most of the work: 2 is the only even prime. When a stem pairs "prime" with a parity condition, it is usually forcing one variable to 2. Treat "odd prime" as a deliberate restriction — the test writes it because 2 would otherwise be live.

  For factors, hold this: if $n=p_1^{a_1}{p}_2^{a_2}\cdots$, the number of positive divisors of $n$ is $(a_1+1)(a_2+1)\cdots$. The count is odd exactly when $n$ is a perfect square. The same fact settles "how many pairs $(a,b)$ with $ab=n$": half the count, rounded up — and when $n$ is a square the middle pair contributes one.

  The standard way to lose points is listing divisors by hand. You forget $n$ itself, or you stop at $\sqrt{n}$ and double without noticing that a square has already been counted once.

  The decisions that separate the answer from the trap:

  • Whether $n$'s value matters or only its shape. Many stems ask about form; computing wastes time and invites slips.
  • Whether the variables may be 0, 1, or negative. Testing only 2 and 3 is the common failure; 1 breaks most prime assumptions, 0 breaks most positive ones.
  • Whether "prime" was intended to exclude 2.

  You are ready to move on when

  • You factor any integer under 200 into primes without hesitation and read off its divisor count in a single step
  • For any stem pairing "prime" with a parity condition, you identify within seconds whether 2 is live or excluded before you start solving
  • On parity and factor questions that permit it, you test 0, 1, and a negative value before committing to an answer

  Quantitative

  Divisibility and remainders

  The exam rarely asks you to divide. It asks you to bound an unknown integer or to decide whether a divisibility condition is pinned down at all.

  Nearly every remainder problem reduces to one identity: $N=dq+r$, where $q$ is an integer and $0\le r<d$. The inequality does the work, not the equation. "$N$ leaves remainder 3 when divided by 7" plus "$N<40$" leaves a handful of candidates; a quotient named in the stem often exists only to be bounded.

  Expect disguises. Trailing zeros of $n!$ (count factors of 5 — 2s are never scarce). "How many multiples of 3 lie between $a$ and $b$" (a difference of floors). Parity questions, which are remainders mod 2. DS stems asking "is $N$ divisible by 12", where you need 4 and 3 independently — divisibility by 6 and by 2 gets you nothing.

  Two decisions separate the right answer from the tempting one. First, whether a statement constrains the modulus you need: "$N\equiv 2(\mathrm{m}\mathrm{o}\mathrm{d}4)$" says nothing about $N<mrow><mi\; mathvariant="normal">m</mi><mi\; mathvariant="normal">o</mi><mi\; mathvariant="normal">d</mi></mrow>8$. Second, whether arithmetic on remainders is legitimate — you may add and multiply them under the same modulus, not under a different one.

  The mechanical way strong candidates drop the point: finding $N\equiv 2(\mathrm{m}\mathrm{o}\mathrm{d}3)$ and then treating 2 as the value of $N$.

  Worth holding: for coprime $a$ and $b$, the residues of $N$ mod $a$ and mod $b$ determine $N$ mod $ab$.

  You are ready to move on when

  • You rewrite every 'leaves remainder' clause as $N = dq + r$ and extract a bound on $q$ before reading the answer choices
  • On a divisibility DS item you can name which modulus each statement constrains and which it leaves free, without plugging numbers
  • You produce the trailing-zero count of $n!$ and the number of multiples of $d$ in a range by method, and can defend both in one sentence

  Verbal

  Critical Reasoning: assumptions and the negation test

  The word "assumption" may never appear. GMAT Focus asks what the argument depends on, requires, or presupposes, and hides the same demand inside boldface and evaluate stems. The correct answer is necessary, not merely helpful — and the tempting wrong answer is precisely the statement that would strengthen the argument if true but that the argument can survive without.

  Two decisions separate right from tempting. Necessary versus sufficient: reject anything the conclusion stands without. Linkage: the assumption must connect these premises to this conclusion, not to a broader claim the author never made.

  The negation test settles the first decision mechanically. Negate the logical force, not the sentence:

  • "All A are B" → "at least one A is not B"
  • "Some A are B" → "no A is B"
  • "No A are B" → "at least one A is B"

  If the negated statement leaves the conclusion standing, discard the choice; if it destroys the argument, keep it. Suppose a sales jump followed an ad campaign. The needed assumption is not "many people saw the ads" but "nothing else caused the jump" — negate the latter and the conclusion collapses.

  Where strong candidates bleed points: they negate, then argue with the negation instead of asking whether the conclusion still follows; and under time pressure they pick the strongest-sounding statement. Negate every finalist, including the one you like.

  You are ready to move on when

  • You negate any answer choice by reversing its logical force rather than rephrasing its wording, out loud and without hesitation
  • Across a set of ten assumption questions you reject the strengthening-but-unnecessary answer every time and can say why the conclusion survives its negation
  • You identify the required assumption inside boldface or evaluate stems before reading the answer choices

  Data Insights

  Data Sufficiency: the sufficiency judgement

  Data Sufficiency does not test whether you can find an answer; it tests whether the information fixes one. Every question reduces to: across all cases the statement permits, is the answer to the question asked always the same? Sufficiency is not truth — "Is $x$ negative?" answered with a definitive no is sufficient.

  The disguises are where statements look decisive but are not. A statement that fixes $x+y$ leaves $x$ alone. A word problem about people or objects silently requires an integer. A geometry statement fixing two sides of a triangle does not fix the third. A statement giving $x^2=16$ looks like it yields a value; for "What is $x$?" it does not.

  Three decisions separate the right answer from the tempting one:

  • Value questions: one working case proves nothing. Two valid cases with different answers prove insufficiency — this is your entire tool for eliminating.
  • Erase statement 1 before judging statement 2. Carrying a value across is the most common mechanical way strong candidates lose points here.
  • Two solutions for the variable can still be sufficient if both produce the same answer to the question actually asked.

  Have this in memory: a statement is sufficient exactly when the set of permitted cases maps to a single answer to the question posed — no more, no less.

  You are ready to move on when

  • You can produce two concrete permitted cases with different answers for every statement you label insufficient
  • You evaluate each statement from scratch and can name any place where statement 1's value leaked into your statement 2 reasoning
  • You have caught a question where a statement was sufficient despite not fixing the variable — a definite no, or two roots giving one answer

  Open week 2 on its own page →

• 3 Algebra, statistics & Mock 1 BASELINE SCORE ACCURACY EARNS SPEED 10–15 hrs

  The first full-length mock goes here, not in week 1. A baseline taken before you have seen half the syllabus measures how much you have not studied yet, which you already know.

  Quantitative

  Linear equations and algebraic manipulation

  GMAT Focus tests linear equations less as "solve for $x$" than as translation and sufficiency. The topic rarely announces itself. It arrives as a word problem — rates, mixtures, ages, digits, profit — or as a Data Sufficiency stem where the variable you want will not isolate.

  Two decisions recur. First, count independent equations against unknowns, then check for hidden constraints. "Positive integers", "distinct digits" or a bounded range can make one equation in two variables determinate; miss it and you call a statement insufficient when it is sufficient. Second, in DS, ask whether a unique numeric value is forced. $x=5-y$ is insufficient for $x$ — but may be sufficient for $x+y$. When the target is an expression such as $2x+3y$, do not solve for $x$ and $y$ separately; manipulate the given equations toward the expression.

  The mechanical loss is arithmetic drift: solving for variables when the target is an expression, and dropping a sign when distributing a negative. Add or subtract equations first to see what cancels. If you clear a variable from a denominator, record the value it cannot take; that value will appear among the traps.

  For $a_{1}x+b_{1}y=c_1$ and $a_{2}x+b_{2}y=c_2$, a unique solution exists exactly when $a_1{b}_2-a_2{b}_1\mathrm{\ne }0$. When that determinant is zero, the system has either no solution or infinitely many, according to consistency.

  You are ready to move on when

  • You translate a word problem into equations before computing, and on a mixed set no error traces to a misread relationship
  • You decide sufficiency in a two-variable DS item without solving for either variable, by testing integer or positivity constraints
  • You manipulate two equations to isolate a requested expression, and you verify the sign when distributing a negative

  Quantitative

  Mean, median, mode and range

  GMAT Focus rarely asks you to compute a mean or median from a clean list. It asks what happens when you add, remove, or replace a value, or whether a statement pins down the median. The arithmetic is trivial; the trap is distribution.

  Two decisions separate answers. First, when a value changes, compare it to the current mean to predict the mean's direction, but locate it relative to the sorted set for the median. Adding a value equal to the mean leaves the mean unchanged; the median may still move if the new value's position shifts the middle. Second, in Data Sufficiency, mean and range are weak constraints: many sets share the same mean and range but have different medians. Ask whether the statements force one sorted arrangement or merely allow one.

  Disguises: 'average' usually means arithmetic mean, but a question about 'typical' salary or home price is testing median's resistance to outliers. A set with duplicates makes mode and median position-counting traps. Removing a value can shrink the range only if it was the min or max; otherwise range is unchanged.

  Worth memory: for $n$ sorted values, median position is $(n+1)\mathrm{/}2$ when $n$ is odd; for even $n$, average the two middle values. New mean after adding $x$ to $n$ values with old mean $M$: $(nM+x)\mathrm{/}(n+1)$.

  Strong candidates lose points by computing the correct new mean and then assuming the median moved with it, or by reporting a mode for a set where every value appears once. The exam expects 'no mode', not a fabricated one.

  You are ready to move on when

  • You predict the direction of a mean change from a single added or removed value without recomputing the full sum
  • You decide whether the median moves after an insertion or removal by tracking sorted position and duplicates
  • In Data Sufficiency, you reject a mean-and-range statement as insufficient when it allows multiple medians

  Quantitative

  Weighted averages

  The exam almost never says "weighted average." It says overall average, combined mean, average price per unit, per-capita, or hands you a two-row table and asks for the mean of the whole column. Mixture problems — a 40% solution blended with a 70% one — are the same question in different clothing. The trigger to slow down is any setup that collapses two or more groups into a single figure without stating that their sizes are equal.

  The mechanical failure that costs strong candidates points here is averaging the averages. If 30 values average 12 and 20 values average 17, the combined mean is $\frac{30(12)+20(17)}{50}=14$, not $14.5$. When the counts are equal both routes agree, so the habit goes unpunished until it does not.

  Two decisions do most of the work:

  • What is the weight — count, dollars, hours, distance? For average speed over equal distances, weight by time; that is why the answer is $\frac{2ab}{a+b}$, not $\frac{a+b}{2}$.
  • Where does the answer sit relative to the midpoint? Always between the two group means, nearer the larger group.

  Carry one formula: $\frac{w_1}{w_2}=\frac{\mathrm{\mid }{x}_2-\bar{x}\mathrm{\mid }}{\mathrm{\mid }\bar{x}-x_1\mathrm{\mid }}$ — the weights (counts) are inversely proportional to the distances from the combined mean. It answers ratio-of-group-size questions in one step, without simultaneous equations.

  You are ready to move on when

  • On a fresh set, you never compute a combined mean by averaging two given averages, even when the group sizes differ by awkward amounts
  • You name the weight (counts, dollars, hours, distance-as-time) before writing any expression, and you correctly default to time in equal-distance speed problems
  • Given a combined mean and both group means, you recover the ratio of group sizes from $\frac{|x_2-\bar x|}{|\bar x-x_1|}$ in one line

  Verbal

  Critical Reasoning: strengthen and weaken

  Strengthen and weaken questions are assumption questions wearing a different verb. What is tested is not whether you can weigh evidence, but whether you can find the unstated link between the evidence given and the conclusion drawn, then judge which option moves that link.

  Disguises: Which of the following, if true, would most seriously undermine the argument, provides the strongest support for, the argument is vulnerable to criticism on the grounds that, would most weaken the conclusion. The phrase if true is the tell — you are judging the effect of the option, not its plausibility.

  Three decisions do the work:

  1. Fix the conclusion. It is usually the prediction, recommendation, or causal claim, often carried by an implied therefore. Options that attack a premise, or a claim the author never made, are wrong.
  2. State the gap before reading the options: what must hold for this evidence to support this conclusion?
  3. Check direction and scope — the tempting option usually concerns a related quantity the argument never mentions.

  The mechanical loss is treating strengthen as prove. The right answer only shifts likelihood; a modest effect on the assumption beats a striking fact about something else, and under time pressure strong candidates take the first true-and-relevant option and stop reading.

  Use the negation test: if denying a candidate assumption would collapse the argument, that assumption carries the weight, and the correct answer is the option affirming or denying it — provided it stays inside the argument's terms.

  You are ready to move on when

  • Before reading the options, you can write the conclusion and the gap as two single lines
  • You can sort all five options into 'moves the gap' and 'moves something else' without using outside knowledge
  • For every rejected option you can name the word or scope shift that disqualifies it

  Exam skills

  Mock 1: taking a baseline that means something

  Mock 1 is a measuring instrument, not a rehearsal for a good day. Its job is to record what you do by default when a section turns hostile and the clock is real. That is the only thing it can tell you that practice sets cannot, and the data is easy to corrupt.

  Corruption has three usual sources: pausing the clock once ("just to finish this one"); taking it in slices across a day, which hides the fatigue that lands hardest in Data Insights, sat last; and best-effort mode, where you abandon your process to prove you can crack a hard question, then review a run that no longer resembles test day.

  Three decisions carry the value:

  • Finish every section, guessing the last questions rather than leaving blanks. A blank is a decision you would not make under the real clock.
  • Sit it once, at the hour of your actual appointment, with the scratch materials you will be given.
  • Log before you read explanations. Your memory of your reasoning overwrites the evidence of your behaviour.

  The log is four cells: fast/slow x right/wrong. Fast-and-wrong is a content gap or a trap. Fast-and-right is ready. Slow-and-wrong is an abandonment rule you have not written yet. Slow-and-right is your speed backlog, and it is where most recoverable points sit — which is why reviewing only your errors is the standard way strong candidates waste a baseline.

  Average pace $=$ section minutes $\mathrm{/}$ questions in section. It is an average, not a per-question cap: overspending on one question is fine only if you have banked the time elsewhere.

  You are ready to move on when

  • A completed four-cell tally (fast/slow x right/wrong) for all three sections, filled in before reading any explanation
  • A quotable pace figure per section: average seconds per question, and the count of questions that ran past twice it
  • One written abandonment rule and one written protocol (start time, single sitting, no clock pauses) you will repeat identically on the next mock

  Open week 3 on its own page →

• 4 Exponents, inequalities & CR reasoning ABSTRACT THINKING 20–25 hrs

  The heaviest conceptual week. Exponents and inequalities are where the candidate set quietly doubles — a square root, an even power, a multiplication by an unknown sign — and where most avoidable Quant losses live.

  Quantitative

  Exponents and roots

  GMAT Focus seldom asks you to simplify a power. It hides exponent structure inside comparisons, number-property constraints and Data Sufficiency traps.

  The common disguise is a question that never says "exponent": comparing $2^{30}$ with $3^{20}$, counting trailing zeros of a factorial, or finding a units digit. The move is mechanical — force a common base or a common exponent. For the comparison, rewrite as $(2^3{)}^{10}=8^{10}$ against $(3^2{)}^{10}=9^{10}$ and read it off.

  Three decisions separate the credited answer from the tempting one:

  • An even root is not the inverse of squaring. $x^2=81$ has two real solutions; $\sqrt{81}$ has one. The real test is whether the stem's constraint (integer, positive, non-zero) kills the negative branch.
  • $\sqrt{x^2}=\mathrm{\mid }x\mathrm{\mid }$, not $x$. Substituting a negative number to check is faster than reasoning about it.
  • Constraints travel with the variable. "$n$ is a positive integer" and "$n$ is a positive number" produce different answer counts on the same-looking DS stem.

  The reliable loss here is mechanical: distributing a power over a sum, $(a+b{)}^2\mathrm{\ne }{a}^2+b^2$, and dropping the negative root whenever the stem hands you an even power.

  Worth having: $a^m{a}^n=a^{m+n}$ and $(a^m{)}^n=a^{mn}$, valid for $a>0$; for $a<0$ they hold only when the exponents are integers.

  You are ready to move on when

  • You rewrite a comparison of two large powers onto a common base or common exponent without hesitation
  • Given an even-power equation you state the full solution set, then state which solutions each constraint type eliminates
  • You spot an exponent problem embedded in a units-digit, trailing-zeros or factor-counting question before doing any arithmetic

  Quantitative

  Inequalities and the sign of the unknown

  On this test, inequalities are not a manipulation exercise; they are a sign-management exercise. Nearly every hard question turns on the fact that the algebraic move you want to make is legal only when the unknown has a known sign — and the question is usually built so that it doesn't.

  The load-bearing rule: for $a>b$, multiplying by $c$ preserves the inequality if $c>0$ and reverses it if $c<0$. Nothing else in the topic does as much damage.

  Disguises: cross-multiplication in a ratio ($\frac{1}{x}>\frac{1}{y}$); comparing squares, where $x>y$ does not give $x^2>y^2$; "which of the following must be true" items where every option is an ordinary rearrangement; Data Sufficiency statements that pin down $\mathrm{\mid }x\mathrm{\mid }$ without pinning the sign of $x$.

  Three decisions separate right from tempting:

  • Never multiply or divide by anything containing the unknown unless the stem licenses its sign.
  • Test $0$, a small positive, and a negative before committing — signs and boundaries, not magnitudes.
  • In Data Sufficiency, a statement settles the sign only if it forces one: $x^2>4$ doesn't; $x>2$ does.

  The mechanical loss: a candidate cross-multiplies in a DS item, reaches a clean expression, marks it sufficient, and the answer hinges on the negative case they never tested. The arithmetic was flawless; the sign was never checked.

  Worth memorising: $\frac{1}{x}>1$ is equivalent to $0<x<1$, not to $x<1$.

  You are ready to move on when

  • On any inequality containing the unknown in a denominator, you state the permitted sign before evaluating the statements
  • For every boundary or range question you test zero plus one number on each side of it, and can say why zero was included
  • When a statement gives a squared or absolute-value condition, you explicitly record whether it fixes the sign and treat it as insufficient for a directional comparison until it does

  Quantitative

  Absolute value

  GMAT Focus rarely asks you to compute an absolute value. It uses one as a sign-suppression device, and the actual question is whether you can tell when a quantity's sign is knowable and when it is not.

  The workhorse reading is distance: $\mathrm{\mid }x-a\mathrm{\mid }$ is the distance from $x$ to $a$, so $\mathrm{\mid }x-3\mathrm{\mid }<2$ means "within 2 of 3". Reach for that before case-splitting, because it collapses a two-case problem into one interval. Worth having: $\mathrm{\mid }x\mathrm{\mid }<a⟺-a<x<a$, and $\mathrm{\mid }x\mathrm{\mid }>a⟺x<-a$ or $x>a$ — both only for $a>0$. Applying the first form with a negative right side is a silent error.

  Disguises: $\sqrt{x^2}=\mathrm{\mid }x\mathrm{\mid }$; even powers such as $x^4=16$ or $x^2=(x-2{)}^2$; "how far apart are $p$ and $q$"; and any Data Sufficiency stem demanding the sign of a variable, where $\mathrm{\mid }x\mathrm{\mid }=x$ if and only if $x\ge 0$.

  The decisions: is this a distance (solve geometrically, no cases) or a definition (two cases, take both)? Must the non-absolute side be non-negative — $\mathrm{\mid }2x-5\mathrm{\mid }=3x$ admits no negative root, so $x\ge 0$ frames everything? And does each candidate satisfy the case assumption it was solved under, not merely the equation you produced?

  The mechanical point-loser is squaring both sides to kill the bars, then not substituting back. Squaring invents roots that satisfy the squared equation and not the original; test makers place those invented roots in the answer choices, usually as the tempting middle option.

  You are ready to move on when

  • You state the sign restriction on the non-absolute side before solving, and reject roots that violate it
  • You rewrite square roots of squares, even powers, and distance phrasing as absolute values without being prompted
  • You convert $|x - a| < b$ and $> b$ into intervals correctly and never apply the $-a < x < a$ form to a negative right-hand side

  Verbal

  Critical Reasoning: evaluate, flaw and paradox

  Evaluate, flaw and paradox questions rarely name themselves. They appear as ordinary arguments: a company's sales rose after a rebrand; a city's crime fell after a curfew; a study links a gene to a disease. What GMAT Focus tests is not your ability to spot the label but to locate the gap between evidence and conclusion—the unstated assumption, the causal leap, the scope shift.

  The disguises are thin. A flaw question may ask why the argument is vulnerable to criticism; an evaluate question may ask what would most usefully be known; a paradox may present two facts as a surprising contrast. In each, you are hunting the same thing: what must be true for the conclusion to follow?

  Three decisions separate correct from tempting. For evaluate, the answer must be a question whose answer could cut either way—strengthen or weaken—and would change the argument's force. For flaw, you must name the precise error: correlation treated as causation, a necessary condition treated as sufficient, a sample too small or unrepresentative to support the generalisation. For paradox, the resolution must reconcile both facts, usually by introducing a third factor that makes both true; explaining one away is wrong.

  Strong candidates lose points by answering the question they expected. On evaluate they pick a strengthener or weakener; on paradox they discredit one fact; on flaw they accept a vague objection. The mechanical fix: before reading options, articulate the assumption and negate it. The argument's conclusion collapses if that negation is plausible, and that is your answer.

  You are ready to move on when

  • You can state the unstated assumption in an evaluate question before looking at the options
  • You can distinguish a flaw answer that names the precise logical error from one that merely weakens the argument
  • You can propose a resolution to a paradox that leaves both facts intact

  Data Insights

  Two-Part Analysis

  Two-Part Analysis is not two questions glued together. The exam gives a short scenario and two columns of options; you must choose one option per column so that the pair satisfies a joint condition. The condition is often an equation, but it can be logical: one column might be a cause, the other an effect; one a premise, the other a conclusion.

  Disguises: the coupling hides in a clause you read past—'...the same amount...', '...twice as many...', '...if and only if...'. Or the two columns look like independent variables, and the options are arranged so that the best option in each column alone is a tempting pair that violates the link.

  Three decisions separate the correct pair. First, before evaluating options, write the relation that binds the columns. Second, treat the columns as a system: if one choice is fixed, the other is usually forced, so use the more constrained column as your filter. Third, verify the pair against every constraint, not just the one you used to find it.

  The mechanical loss: strong candidates solve one column correctly, then pick the other column's option that 'fits the story' rather than the one that satisfies the joint condition exactly. Or they assume independence and select the best option in each column separately.

  Useful memory: if two quantities have sum $S$ and difference $D$, then $x=(S+D)\mathrm{/}2$ and $y=(S-D)\mathrm{/}2$ (with $x\ge y$). Use it when the question gives a total and a 'how many more' in any form.

  You are ready to move on when

  • You write the coupling condition in one sentence before looking at the options
  • You solve the more constrained column first and use it to cut the other column to at most two candidates
  • You can name the exact constraint that each tempting wrong pair violates

  Open week 4 on its own page →

• 5 Work & rates, RC foundations & tables STAY ORGANIZED 10–15 hrs

  Word problems and long passages fail for the same reason: the information arrives faster than it is organised. This week is about the notation and the map, not the arithmetic.

  Quantitative

  Work and rate

  Work-rate questions on GMAT Focus are rarely about multiplying rates. They test one habit: rewrite every participant as output per unit time, then add. The rest is bookkeeping.

  The stem seldom says "rate". It says a pump fills a tank, a machine stamps parts, a drain leaks. Negative rates are ordinary — a leak is just a rate with a minus sign, and the answer turns on whether you kept that sign.

  Two decisions separate correct from almost-correct. First, what counts as one job. If the stem gives counts (widgets, pages, litres), write $W=rt$ and keep the count as $W$. If it gives only times, choose $W$ deliberately — the LCM of those times makes every rate a whole number. Second, the inversion: "A is twice as fast as B" is a claim about rates, so $t_A=t_B\mathrm{/}2$. Rate and time pull in opposite directions, and carrying the 2 straight across into the times is the most common error on this topic.

  For simultaneous work, $\frac{1}{t_{\text{together}}}=\frac{1}{t_A}+\frac{1}{t_B}$, valid only while both actually work. Combined time is never the mean of the individual times: for three hours and five hours it is $15\mathrm{/}8$.

  The mechanical loss is dropping the reciprocal — computing a combined rate correctly, then offering that rate as the answer to a question that asked for time. Under pressure it goes unnoticed.

  When someone joins or leaves mid-job, split the timeline at that point. Work done in each phase adds; average rates do not.

  You are ready to move on when

  • Across a mixed set of ten problems, you converted each participant to output per unit time before writing any equation
  • You solved at least one answer-in-terms-of-$x$ work problem and verified the result has the dimensions of time, not rate
  • You handled a negative-rate stem and a mid-job join-or-leave stem by splitting phases, without needing the wording to flag it

  Quantitative

  Speed, distance and time

  GMAT Focus rarely asks for a bare $d=rt$. It asks you to hold two or three rates in one consistent frame, and the whole question turns on which quantity is held fixed. Distance fixed across legs gives a harmonic average; time fixed gives an arithmetic one. The exam sets up the first and punishes the second.

  The disguise is always a story — two cyclists, a train clearing a platform, a tailwind, a return journey. Nothing says "average speed". Your job is to build the table (distance, rate, time; one row per leg; every unit identical) before writing any equation. That organisation is the topic.

  Three decisions separate the answer from the trap. First, average speed is $\frac{\text{total distance}}{\text{total time}}$, never the mean of the speeds; for equal distances at $a$ and $b$ it is $\frac{2ab}{a+b}$, and that shortcut only holds when the distances are equal. Second, relative speed is $a+b$ for opposing motion and $a-b$ for chasing. Third, convert before combining: $1\text{ km/h}=\frac{5}{18}\text{ m/s}$.

  The mechanical loss is splitting time evenly when the problem split distance evenly, or the reverse. Assign the faster leg the smaller time explicitly, then check the answer sits between the two speeds.

  You are ready to move on when

  • You build a distance-rate-time table with one row per leg, in unified units, before writing any equation
  • You identify correctly whether the question holds distance or time constant, and apply the matching average
  • You compute a round-trip or two-leg average speed in under a minute and verify it falls between the two leg speeds

  Quantitative

  Mixtures and concentration

  On GMAT Focus, mixtures are weighted-average questions in costume. The exam will not ask for a chemistry idea; it gives you two stocks at different concentrations and asks for the blend, the amount to add, or the fraction to replace. This is bookkeeping, and the week's theme applies: keep a ledger.

  Track solute in absolute units. Percentages don't add; amounts do. Write solute-in + solute-added = solute-final, with volumes alongside, before computing. Final volume is not automatically initial plus added — removals and evaporation shrink it.

  Three decisions decide most of these:

  • Never average the two concentrations unless the volumes are equal. The unweighted mean is the standard trap.
  • In replacement, the portion removed carries the current concentration, and each cycle multiplies the remaining solute by the same factor.
  • "Add water" and "evaporate" both leave solute fixed — resist recomputing it.

  Formulas: blend concentration is $\frac{V_1{c}_1+V_2{c}_2}{V_1+V_2}$. For a target $c$ between $c_1$ and $c_2$, alligation gives amounts in ratio $V_1:V_2=(c_2-c):(c-c_1)$. After $n$ replacements of volume $x$ from total $V$: $c_n=c_0(1-x\mathrm{/}V{)}^n$ — needs full mixing and solvent-only replacement.

  Disguises: alloys, fuel blends, diluting juice, draining a tank of solution, and the "average percentage across two groups" question.

  The mechanical way strong candidates lose points: they track volume through a removal but not solute, so they divide the wrong quantity; or they apply the replacement factor once rather than compounding; or they reverse the alligation ratio.

  You are ready to move on when

  • You set up any two-stock blend as a solute ledger and can state the final concentration before looking at the answer choices
  • On a drain-and-replace question you write $(1-x/V)^n$ immediately, with $V$ taken as the total volume before removal
  • You can restate an alloy, fuel-blend, or two-group average-percentage question as a weighted average and name both weights on the first read

  Verbal

  Reading Comprehension: mapping a passage

  GMAT Focus RC is a location test, not a memory test. Your notes exist to tell you where to look and who is speaking — not what the passage says.

  Map for function, not content. For each paragraph, write a few words naming its job: sets up a puzzle, states the orthodox view, offers evidence, concedes, rebuts. Add one marker for the author's stance verb (argues, concedes, merely notes): a passage that reports a theory and one that endorses it look identical until a question asks what the author would accept.

  Disguises: "The author implies" and "It can be inferred" name no line, so your map is the only pointer. "According to the passage" questions do name a line, and the trap is answering from your gist instead of rereading that sentence. Watch choices that are true in the world but absent from the text, and choices that quote a view the author cites only to reject.

  Two decisions separate answers: whether the claim is the author's or a reported party's, and whether the question's scope is the whole passage or one sentence. A primary-purpose answer must account for every paragraph; a detail answer needs one sentence.

  Keep this rule in memory: hinge words (however, therefore, moreover) mark a paragraph's function, and every paragraph gets exactly one label.

  The mechanical loss is over-mapping — paraphrasing whole sentences — which burns the time you needed to reread the single sentence that settles the question.

  You are ready to move on when

  • You can label every paragraph's function in a few words and mark the author's stance before you reach the first question
  • You can name the one sentence that supports your answer to each question, inference questions included
  • For every choice you reject, you can say whether it failed on author-versus-reported-view, scope, or absence from the text

  Data Insights

  Table Analysis and sorting

  GMAT Focus does not test whether you can read a table. It tests whether you can convert a verbal claim into a row filter and a column comparison, then use the sort control to make the relevant rows adjacent. The table is a search space, not a dataset to be averaged.

  The statements are disguised queries. 'Every site with capacity above 500 has a vacancy rate below 10%' is a universal claim: sort by capacity, find the first row above 500, then scan the vacancy column for all rows in that block. 'At least one project has both a budget over 2 million and a duration under 6 months' is existential: sort by either condition and look for the first row satisfying both. 'The median of column A exceeds the median of column B' asks you to sort each column separately, not to compute means.

  Three decisions separate correct from tempting:

  • Universal vs existential: for 'all', one counterexample defeats it; for 'some', one example proves it. Sorting to the extreme is usually the fastest test.
  • Within-row vs across-row: the statement may compare two columns in the same row, or aggregate one column across many rows. Do not mix them.
  • Strict vs inclusive: 'more than' and 'at least' change which rows qualify after sorting.

  The mechanical loss: you answer a nearby question. You verify that the highest value is above a threshold when the statement asks whether every value is. Or you read the top row after sorting but forget the table scrolls, missing rows below.

  Formula worth having in memory: for $n$ sorted values, the median is the $((n+1)\mathrm{/}2)$th if $n$ is odd, and the average of the $n\mathrm{/}2$th and $(n\mathrm{/}2+1)$th if $n$ is even.

  You are ready to move on when

  • You can take any Yes/No statement and label it universal or existential before touching the sort control
  • You can sort by a condition column, identify the contiguous block of qualifying rows, and check the target column across every row in that block
  • You can answer a median question by sorting and locating the $((n+1)/2)$th value for odd $n$, rather than computing a mean

  Open week 5 on its own page →

• 6 Sets & sequences, RC purpose & Mock 2 HALFWAY CHECK STAY STRUCTURED 15–20 hrs

  Mock 2 is the halfway check, and its job is comparison rather than a number. What moved since week 3, and did it move in the sections you actually worked on?

  Quantitative

  Overlapping sets and the double matrix

  GMAT Focus sets questions almost always reduce to two binary attributes plus a total — which is why the 2×2 grid beats a Venn diagram. Set up rows for one attribute, columns for the other, add a total row and a total column, and every cell becomes arithmetic. Venn diagrams earn their place only when three attributes are in play.

  The question rarely says "sets". It says employees who passed certification and speak Spanish, patients given the drug and reporting side effects, respondents who read the report or attended the briefing. Translate "or" to the union, "and" to the intersection, "only" to a single cell, "neither" to the bottom-right corner. Given percentages with no counts, set the total to 100.

  Three decisions separate the credited answer from the tempting one:

  • Whether the stated total includes the "neither" group. "Of 200 surveyed" usually does; "of the 180 who answered" does not.
  • Whether "or" is inclusive. In counts it is, unless the question says "but not both" — then you want the union minus the intersection.
  • With three attributes, "exactly two" is not the sum of pairwise overlaps. It is that sum minus $3\mathrm{\mid }A\cap B\cap C\mathrm{\mid }$, since each triple-overlap member sits in all three pairwise cells.

  The mechanical loss: filling the "only A" cell with A's total instead of subtracting the overlap, then carrying the error down the totals column. The grid stays self-consistent, so it survives a sanity check.

  Keep in memory: $\mathrm{\mid }A\cup B\mathrm{\mid }=\mathrm{\mid }A\mathrm{\mid }+\mathrm{\mid }B\mathrm{\mid }-\mathrm{\mid }A\cap B\mathrm{\mid }$, and when a neither group exists, $\mathrm{\mid }A\cup B\mathrm{\mid }=\text{Total}-\mathrm{\mid }\text{Neither}\mathrm{\mid }$.

  You are ready to move on when

  • You convert any two-attribute word problem into a filled 2×2 grid without drawing a Venn diagram
  • You state before solving whether the given total includes the 'neither' group, and adjust when it does not
  • You automatically subtract 3× the triple intersection when a three-set question asks for 'exactly two'

  Quantitative

  Sequences and series

  On GMAT Focus, sequences are rarely named. The section tests two moves: turning a definition into a specific term, and summing evenly spaced numbers. Most items arrive as word problems — weekly savings that rise by a fixed amount, a population that doubles, a repeating alphabet or digit pattern, "how many multiples of 6 lie between 200 and 500".

  Three decisions separate the credited answer from the tempting one:

  • Term, sum, or count? The distractors are built from the other two. Computing $a_n$ when the question asked for $S_n$ is the most common trap on offer.
  • Where does the index start? A definition like $a_n=a_{n-1}+3$ with $a_1$ given is 1-based; a "0th" term shifts every answer.
  • Brute force or cycle? For recursive and repeating sequences, compute $a_1$ through $a_5$ or $a_6$, identify the period, then use $n<mrow><mi\; mathvariant="normal">m</mi><mi\; mathvariant="normal">o</mi><mi\; mathvariant="normal">d</mi></mrow>\text{period}$. Candidates who grind out terms run out of clock and drop a sign.

  The mechanical loss is the count. Terms from $p$ to $q$ in steps of $d$: $(q-p)\mathrm{/}d+1$. The $+1$ is where points go. With the count, an evenly spaced sum is count × average, and the average is $(\text{first}+\text{last})\mathrm{/}2$ — valid only when consecutive terms differ by a constant.

  For a quantity growing by a fixed percentage, $a_n=a_1{r}^{n-1}$. Not $r^n$. Both exponents will appear among the choices.

  You are ready to move on when

  • You write the count as (last − first)/step + 1 before attempting any evenly spaced sum
  • You find the period of a recursively defined sequence within six computed terms and answer a large-n question by remainder
  • You state the exponent as n−1, not n, for any geometric growth question, and can say why

  Verbal

  Reading Comprehension: primary purpose and structure

  The GMAT Focus does not ask what a passage says. It asks what the author is doing and how the parts are arranged. Two question families dominate: primary purpose ("The primary purpose of the passage is to...") and structure/function ("The author mentions X in order to...", "Which of the following best describes the organization..."). The subject—astronomy, art, economics—is bait.

  Disguises: purpose questions offered as topic summaries. A choice that accurately states the subject but omits the author's stance—evaluate, reconcile, challenge, explain—is wrong. Structure answers often describe one paragraph's move as if it were the whole passage. The reverse also appears: a sweeping abstraction that ignores a crucial qualification.

  Three decisions separate correct from tempting. (1) Scope: whole passage, not the most memorable paragraph. (2) Verb: what is the author doing to the material? (3) Stance: endorse, critique, or neutral? If a passage presents a debate and then resolves it, the purpose is usually to adjudicate, not to survey.

  Strong candidates lose points mechanically by reading for content, then answering from memory. The fix: after each paragraph, note its function in the margin—"received view," "counterexample," "qualified conclusion." Answer from that skeleton, returning only to verify a contested choice.

  Worth having in memory: $\text{primary purpose}=\text{subject}+\text{author’s action}+\text{scope}$. Condition: for argumentative passages, the action must be a stance verb, not "describe."

  You are ready to move on when

  • After reading a passage, can state each paragraph's function in five words or fewer without looking back
  • On a mixed set, can eliminate primary-purpose choices that are topic-accurate but lack an author-action verb or exceed the passage's scope
  • Can correctly separate "main point" from "primary purpose" when both appear as options

  Data Insights

  Multi-Source Reasoning

  Multi-Source Reasoning hands you two or three tabs — usually a memo, a table or graph, and a third piece that looks like scenery. It tests triage, not reading speed. The real question is whether you can hold a claim in your head and know which tab can adjudicate it.

  Map the tabs before you read the first question: one line each on what it uniquely contains. The third tab is rarely filler. It often sets scope (a policy that applies only above a stated order value) or supplies the join key — a product code, a region name — linking the other two. A definition buried in the prose tab quietly overrides a column heading in the table.

  The disguises: a question using a term defined in a footnote; a comparison whose two numbers sit in different tabs; a qualifier column (such as pending) that invalidates a row you would otherwise count.

  Two decisions separate right from wrong. First, name the source before you calculate — if lookup answers it, do not compute; most prompts need at most one operation. Second, verify each row of a select-one-per-row item independently; one wrong judgement loses the whole question, and the standard failure is stopping once two rows feel right.

  Your mechanical leak: anchoring on Tab 1, then answering a Tab 3 question with Tab 1's numbers. Re-anchor per question, not per stimulus.

  No formula worth memorising here, just claim + source = verdict. Of each statement ask: which tab could falsify this?

  You are ready to move on when

  • On a fresh set, you write a one-line map of each tab before opening the first question
  • You name the tab or join key behind every sub-question before doing any arithmetic
  • You catch the footnote, qualifier or scope line that flips an answer, and you check all rows of a select-one-per-row item

  Exam skills

  Mock 2: reading a mid-preparation mock

  A mid-preparation mock is a measuring instrument with a sample size of one. The composite is a noisy reading; the answer log is the result. Read the log first, and deliberately delay reading the score until you have written down what the log says.

  The thing worth extracting is where accuracy breaks inside each section. Accumulated time debt usually converts into three or four rushed questions in a row; find the question number where your first rushed answer occurs and write it down. That number is more actionable than any section score, and it moves in ways the score hides.

  Two decisions carry most of the value. First, classify every miss as content (you did not know), process (you knew, chose badly) or clock (you knew, ran out of time). Content misses send you back to a topic; process and clock misses send you back to a rule, and mid-preparation mocks produce far more of the latter than people want to admit. Second, change strategy only on a pattern seen twice. One mock is an anecdote; the same collapse in two consecutive mocks is a fact.

  The mechanical loss: reviewing only the questions you got wrong. Add the ones you got right slowly, and the ones you flagged as a guess. Those are latent misses, and they distort the accuracy picture you are about to base decisions on.

  Review at least as long as you tested. For a $2.5$-hour mock, budget $2.5$ hours of review — otherwise you have bought a score, not a diagnosis.

  You are ready to move on when

  • I classified every miss in my last mock as content, process or clock, and the three counts are written down
  • I can state the question number in each section where my first rushed answer occurred
  • I logged the questions I answered correctly but slowly, or correctly by guess, alongside the ones I missed
  • I can name the single rule I am changing before the next mock, and it rests on a pattern visible in two mocks rather than one

  Open week 6 on its own page →

• 7 Probability, RC details & MSR DECIDE YOUR APPROACH 10–15 hrs

  Route selection week. Most questions here have two valid approaches with very different costs, and choosing between them in ten seconds is the skill being trained.

  Quantitative

  Combinatorics: counting without listing

  GMAT Focus rarely asks 'how many ways'. It hides counting inside a scenario — seating, codes, committees, grid paths, schedules — and expects you to name the structure and compute. The topic contains factorials and binomial coefficients; the exam tests whether you can decide which structure applies before calculating.

  Three decisions separate the correct count from the tempting one:

  • Order: would swapping two chosen items create a new outcome? If yes, it is a permutation; if no, a combination. A committee with named roles is ordered; a committee without roles is not.
  • Distinctness: are the objects distinct? Repeated letters or identical items mean you must divide by the factorial of each repeat. This is where careful candidates double-count.
  • Restriction: 'at least one' almost always means total minus none. 'Not adjacent' means arrange the free items, then place the restricted ones in the gaps.

  The formula worth carrying: $\left(\genfrac{}{}{0px}{}{n}{k}\right)=\frac{n!}{k!(n-k)!}$, valid when order is irrelevant, the $n$ objects are distinct, and no object is reused. For ordered selections, multiply by $k!$.

  The mechanical way strong candidates lose points is listing: they build cases by hand under time pressure, or they set up overlapping cases and count an outcome twice. Decide the structure first; list only to verify a small residual.

  You are ready to move on when

  • You can label a counting prompt as permutation, combination, or complement before doing any arithmetic
  • You can identify the repeated-element divisor in an anagram-style count without listing arrangements
  • You can convert an 'at least one' restriction into total minus the zero case and compute both correctly

  Quantitative

  Probability

  GMAT Focus probability is rarely a formula test. It tests whether you pick the smallest sample space before computing — and the exam rewards deciding the approach first: complement, symmetry, a two-way table, sequential multiplication, in that order of preference.

  The giveaways for complement are at least one, not all, some. $P(\text{at least one})=1-P(\text{none})$, and the 'none' product needs independence — verify it rather than assume it. The giveaways for a table are of those who, given that, any two-way breakdown. Conditionals are almost always cheaper as a restricted denominator than as $P(A\cap B)\mathrm{/}P(B)$.

  Two decisions separate the answer from the trap. First: are the outcomes you are counting equally likely? $\{HH,HT,TT\}$ is not a valid sample space; $\{HH,HT,TH,TT\}$ is. Second: did removal change the denominator? Without replacement, re-check every later fraction — but first look for symmetry, because the probability that the third card drawn is a heart is $13\mathrm{/}52$: positions are exchangeable. Strong candidates lose points here by grinding out a sequential product that symmetry settles in one line.

  Keep $P(A\cup B)=P(A)+P(B)-P(A\cap B)$; subtract the overlap unless the events are genuinely disjoint.

  You are ready to move on when

  • Before computing, you classify each question as complement, symmetry, two-way table or sequential product, and your classification is right without backtracking
  • You write out the sample space on any question with two or three stages and confirm each listed outcome is equally likely
  • On without-replacement questions you either adjust every subsequent denominator or name the symmetry that makes adjustment unnecessary

  Verbal

  Reading Comprehension: detail and inference

  On GMAT Focus, detail questions are location tasks with a paraphrase attached, not comprehension tasks. The stem names something in the passage; one line settles it. Read that line completely — the answer usually turns on the qualifier you skimmed (usually, until the 1890s, except in France) rather than on the main clause.

  Inference stems ("suggests", "would most likely agree with") are not licences to reason freely. The credited answer is forced by the text, almost always by a single sentence and occasionally by two adjacent ones. Anything you know about the subject from outside the passage exists in the options as a trap.

  Three decisions separate correct from tempting:

  • Scope. If the option introduces a cause, mechanism, or comparison the passage never draws, it fails however sensible it sounds.
  • Degree. Only, all, never, primary and the first need explicit warrant; hedged options usually match hedged prose.
  • Subject. The classic trap is accurate about the wrong entity, period, or group.

  Strong candidates lose points mechanically in two ways: answering from the first read and then defending the pick from memory, and re-reading the whole passage when the stem's keyword was a single searchable phrase.

  Worth memorising: for a "must be true" inference, negate the candidate answer — if the passage survives the negation, the answer is unsupported. That test does not apply to "suggests" stems, where the bar is merely consistency with the author's stated attitude.

  You are ready to move on when

  • Before looking at the options, you can point to the one line that settles each detail question
  • For every option you rejected, you can name which of scope, degree, or subject it violated
  • You apply the negation test to 'must be true' inference options and discard those the passage survives

  Data Insights

  Multi-Source Reasoning: synthesising across tabs

  MSR hands you two to four tabs — typically a prose scenario, a table, and a chart or second table — and either three statements to classify or one question whose options each need a different cross-tab join. The tabs are deliberately insufficient individually. Every answer lives at an intersection.

  The disguise is that nothing says "combine". A statement reads like a plain fact check — "Lyon exceeded its 2021 quota" — but quota is defined in the prose tab, output sits in the table, and the year must be filtered in both. Work from a single tab and you will still find a number that appears to settle the statement. That is the trap.

  Three decisions separate the correct answer from the tempting one:

  1. Which tabs does the stem name or imply? Read the stem first, then read only those tabs.
  2. What is the join key — the entity, period, or category that must match across tabs?
  3. Which tab has authority? A prose definition ("net revenue is stated after returns") overrides a column header you would otherwise read at face value.

  The mechanical loss for strong candidates: they read all tabs thoroughly in order, then answer from whichever tab is freshest in memory. Thorough reading is not synthesis. Before computing, you should be able to say in one clause, "this statement needs tabs 1 and 3".

  Worth keeping: where tabs conflict, the prose tab governs unless the question says otherwise.

  You are ready to move on when

  • For every statement in a set, you name the specific tabs required before doing any arithmetic
  • You can state the join key (entity, period, or category) for each statement in one clause
  • When a prose definition contradicts a table column, you resolve in favour of the prose without hesitating
  • You finish a set having used every tab at least once, or can explain why one was genuinely irrelevant

  Open week 7 on its own page →

• 8 Last new topics: coordinate geometry, boldface & TPA the coordinate plane, still tested under Functions EVERYTHING COVERED 15–20 hrs

  The last new material. After this week nothing on the exam is unfamiliar, and every remaining hour goes into execution rather than coverage.

  Quantitative

  Coordinate geometry

  On GMAT Focus, coordinate geometry is less a topic than a translation service. The exam rarely hands you a labelled graph. It gives you two vertices of a square in the $xy$-plane, a linear cost formula, or the statement $\mathrm{\mid }x-4\mathrm{\mid }<1$, and expects you to convert that into a line, a distance, or a region.

  The disguises are consistent. Linear functions are lines: in $C=40+3n$, the $3$ is a slope and the $40$ is an intercept with meaning attached. Absolute value is distance from a coordinate, so $\mathrm{\mid }x-4\mathrm{\mid }\le 1$ describes a vertical band, not an equation to solve. Inequalities describe regions, and the question then asks whether a named point sits inside, above, or on the boundary. Data Sufficiency leans hard on loci: one statement may define an entire line or circle of possible points, which is rarely enough to pin down a single point.

  Three decisions separate right from tempting. First, decide whether the question wants a signed value or a distance — midpoints and coordinate differences keep their signs, distances never do. Second, before using the perpendicular-slope rule, confirm neither line is vertical; the horizontal/vertical pair is the trap and the rule simply fails there. Third, subtract coordinates in the same order in numerator and denominator; flipping only one reverses the sign of the slope.

  For memory: $m=\frac{y_2-y_1}{x_2-x_1}$ and $d=\sqrt{(x_2-x_1{)}^2+(y_2-y_1{)}^2}$, with perpendicular slopes satisfying $m_1{m}_2=-1$ whenever both slopes are defined.

  You are ready to move on when

  • You can read a word problem with a constant rate and immediately name the slope, both intercepts, and what each means in context
  • You decide within a few seconds whether a Data Sufficiency statement fixes a point or merely defines a locus
  • You check for vertical and horizontal lines before ever applying the negative-reciprocal slope rule

  Quantitative

  Functions and invented notation

  GMAT Focus does not ask you to analyse functions. It asks whether you can execute a rule you have never seen, exactly, under time pressure. The rule arrives as an invented symbol ($a\mathrm{\square }b$), a named function with a piecewise definition, or a table. The mathematics is trivial; the discipline is not.

  Disguises: a function defined only for positive integers, so $f(-1)$ is undefined; a rule that looks symmetric but is not ($a\star b=a-b$); composition written as $f(f(x))$ or $f(g(2))$ with $g$ defined separately; a piecewise rule where the boundary value belongs to one branch only. Read the domain sentence before you substitute anything.

  The decisive moves are: substitute the entire input expression, not just the variable; preserve order when the operation is not commutative; and check whether the output feeds back as an input (composition) or whether the question asks for $x$ given $f(x)$ — a different task.

  Strong candidates lose points mechanically by distributing: assuming $f(a+b)=f(a)+f(b)$ for an arbitrary invented rule. Nothing licenses that. Compute $f(a+b)$ directly.

  For any function, $f(\text{input})$ means replace every instance of the variable in the definition with that entire input. That is the whole method. The condition: the input must lie in the stated domain.

  You are ready to move on when

  • Given an invented symbol or piecewise function, you substitute a compound expression like $a+b$ into every variable without simplifying prematurely and get the correct unsimplified form
  • You can identify the domain restriction in the rule before evaluating and reject an input that violates it
  • You can evaluate nested expressions such as $f(g(3))$ by working inside out and keeping track of which rule applies at each step

  Verbal

  Critical Reasoning: boldface and argument roles

  Boldface questions do not ask you to evaluate an argument; they ask you to map its architecture. The exam presents an argument with one or two boldfaced statements and asks what role each plays. The content is irrelevant except as a cue: a boldfaced sentence may be the main conclusion, an intermediate conclusion, a premise, a concession, an opposing view, a prediction, or a background fact. The same sentence type can play different roles depending on what supports what.

  Disguises: a boldfaced claim attributed to "critics" or "some scientists" is not the author's conclusion unless the author adopts it. A concession introduced by "although" or "granted" is a premise the author accepts but does not rely on as the main support. A prediction or recommendation can be the main conclusion. An intermediate conclusion is supported by other statements and in turn supports the main point; it is neither the main conclusion nor a bare premise.

  Two decisions separate the correct answer from the tempting one. First, who asserts the boldfaced statement, and does the author agree? Second, does the statement support another statement (premise) or is it supported by others (conclusion)? If it is supported by some and supports the main point, it is an intermediate conclusion. The tempting wrong answer often mislabels an intermediate conclusion as the main conclusion, or an opposing claim as the author's premise.

  Strong candidates lose points by labelling from content—"this sounds like a conclusion"—without tracing support arrows. Read the argument twice: first for the author's main point, then for the role of each boldface.

  Rule: a statement is a conclusion if other statements are offered as reasons for it, and a premise if it is offered as a reason for another statement. In "Critics say the policy will fail. But their evidence is flawed, so the policy will succeed," the first boldface is an opposing claim; the second is the main conclusion.

  You are ready to move on when

  • Given a boldface question, you can state the author's main conclusion before reading the answer choices
  • You can correctly label an intermediate conclusion when it is supported by evidence and itself supports the main point
  • You can distinguish an opposing claim from the author's conclusion in a debate-style argument without relying on content alone

  Data Insights

  Graphics Interpretation

  Graphics Interpretation is not a chart-reading test. Every question is a short sentence with two or three blanks, each behind a dropdown, and exactly one combination makes the sentence true. The graph is a constraint, not a story.

  What it tests: whether you can pull a specific value, compare two series, or reason about a plotted relationship — then express it in the units the sentence demands.

  The disguises. Dual-axis line charts where the left and right scales differ, so a series that appears to overtake another never does. Axes that do not start at zero, making a 3% change look like a doubling. Units swapped between graph and sentence (per-capita against total, thousands against millions). A line of best fit drawn through the points, inviting a precise claim the data cannot support.

  Three decisions. Read axis labels, units and the plotted range of $x$ before reading shape. Choose the dropdown options jointly: the first blank often has two defensible answers until the second eliminates one. Separate interpolation from extrapolation — a fitted line licenses claims only within the plotted domain, so "the model predicts" for an unplotted $x$ is the wrong option, not a calculation to attempt.

  The mechanical loss: candidates pick the first dropdown by eyeballing the trend, then hunt for a second option that fits, instead of testing candidate pairs against a specific plotted point.

  Worth memorising: for a fitted line, $m={\displaystyle \frac{y_2-y_1}{x_2-x_1}}$, using two clear points on the line itself, not data points.

  You are ready to move on when

  • On a chart with two vertical axes, you state each series' units and scale aloud before touching a dropdown
  • You select dropdown options as pairs and can justify the choice by naming the plotted point that falsifies the rejected combination
  • You can classify each blank as interpolation or extrapolation before calculating, and refuse out-of-range claims without working them

  Open week 8 on its own page →

• 9 Mixed practice & Mock 3 ALL THREE SECTIONS INTEGRATION PHASE 25–35 hrs

  The first week in which you never know what is coming next. Topic-blocked practice overstates readiness, because recognising the topic is half of most questions and the block already told you.

  Quantitative

  Quant: mixed review

  In a mixed quant set nothing labels itself. The classification step — deciding which tool the problem wants — is what the section tests, and it happens before you write anything down. Burning that step is the loss.

  The disguises are cross-topic. A weighted average arrives as a mixture, a rate, or a class-average table. A ratio arrives as two counts with a known total. A constraint — "how many", "at least", "distinct", "integer" — sits in the final sentence and silently removes candidates from the answer set. Data Sufficiency statements that look sufficient are usually one case from failing: test a non-integer, a negative, or zero before you commit.

  Three decisions separate the correct answer from the tempting one:

  1. Before computing, decide whether the question asks for a value or a comparison. If it asks which is greater, stop solving.
  2. In DS, ask whether testing cases settles sufficiency; it is often faster and safer than algebra.
  3. Fix the constraint set first. Half the traps are a solution those constraints exclude.

  The mechanical loss for strong candidates: finishing algebra they did not need. Mixed sets frequently stop at "sufficient" or "greater than", and continuing costs the minute you needed two problems later.

  Worth holding: for two groups, $\frac{n_1}{n_2}=\frac{\mathrm{\mid }{x}_2-\bar{x}\mathrm{\mid }}{\mathrm{\mid }{x}_1-\bar{x}\mathrm{\mid }}$ — group sizes sit in inverse ratio to their distances from the combined mean. It requires exactly two groups and a computable overall mean.

  You are ready to move on when

  • In a timed mixed set you named the topic and the first move for each problem before writing, and still finished the section
  • For every DS item you missed, you can state the single test case that would have exposed it
  • You solved two-group weighted-average problems via the distance ratio instead of setting up equations

  Verbal

  Verbal: mixed review

  Verbal on Focus is Critical Reasoning and Reading Comprehension only. Mixed review tests the switch between them, and the fact that neither arrives labelled: a stimulus never announces its conclusion, a passage question never announces that it is really a strengthen.

  The disguises: an RC question that is functionally strengthen/weaken applied to one paragraph; a CR stimulus long enough to read like a passage, usually with two voices; a method-of-reasoning item decided by one structural word; a boldface question whose two phrases play different roles.

  Three decisions carry most of the marks. Name the conclusion and its evidence before reading the choices — CR distractors are typically true, on-topic and silent about the gap. Ask what the question demands: "according to the passage" and "the passage suggests" license different degrees of confidence, and the trap for the first is a valid inference the passage never states. Match the strength of the language: a new term or a superlative needs explicit warrant.

  The mechanical way strong candidates bleed points is carrying the previous question's mode forward. Ninety seconds inside a dense passage makes the next CR stimulus get read as prose to be understood rather than an argument to be attacked, so the answer is chosen because it sounds plausible rather than because it closes the gap.

  Keep one tool loaded: the negation test. Negate a candidate necessary assumption; if the argument survives, discard it. It applies only where the answer must be required — not to sufficient-assumption or strengthen questions.

  You are ready to move on when

  • For every CR question missed in a mixed set, you can say in one sentence whether the failure was naming the conclusion, naming the gap, or matching an answer choice
  • You classify each RC question as stated or inferred before reading the options, and can point to the sentence that licenses your answer
  • You notice and restart when you have read a CR stimulus as prose, and you use the negation test only on necessary-assumption questions

  Data Insights

  Data Insights: mixed review

  Mixed review is where Data Insights stops being five topics and becomes one job: work out what a screen is asking before you touch the data.

  The question types are not the skill. The skill is target-fixing — deciding in the first fifteen seconds whether you owe a single value, a comparison, or a yes/no verdict. A Two-Part Analysis can be a weighted average in disguise; a Graphics Interpretation can be a percent change; a Table Analysis can be pure logic where you sort and compare and never divide. Multi-Source Reasoning usually hides its decisive constraint in the tab you opened second.

  Three decisions carry the marks. First, name the target before reading the evidence — candidates who read the data first drift toward whatever number is easiest to compute. Second, in Data Sufficiency take the stem's constraint literally; the trap is a statement that feels sufficient because it is consistent, rather than because it determines. Third, choose compute-or-bound deliberately. Many Graphics and Table items are settled faster by bracketing (above or below 50%?) than by arithmetic.

  The mechanical loss is partial answering: every multi-part question is all-or-nothing, and a blank column scores as a wrong one. Never submit with a column untouched.

  Worth having in memory: percent change $={\displaystyle \frac{\text{new}-\text{old}}{\text{old}}}$, on the condition that 'old' is the base you are changing from, not the later figure.

  You are ready to move on when

  • You can name the question type and the target quantity for each DI item before reading its data, on nearly every item in a mixed set
  • You finished a mixed set with no blank columns and no half-answered multi-part questions
  • You can point to specific items you solved by bounding rather than calculating, and say why that was the faster route
  • On Data Sufficiency items you stated aloud whether the stem demanded a value or a yes/no before you judged either statement

  Exam skills

  Why mixed practice scores lower, and why that is fine

  Mixed practice scores lower than topic practice. That is not a knowledge problem. It is the price of removing the cue. In topic practice, every question announces its type through the set you are in. In mixed practice, and on the real exam, the question does not. Your accuracy on any question is $P(\text{correct})=P(\text{classify})\times P(\text{correct}\mid \text{classified})$, assuming classification is all-or-nothing. The second term is what you have been training. The first is what mixed practice exposes.

  The disguises are ordinary. A Data Sufficiency stem with a single value looks like Problem Solving. A short Critical Reasoning argument looks like Reading Comprehension. A Data Insights question about percentages looks like Quant. Strong candidates lose points mechanically: they carry the previous question's schema into the next one. After three DS items, they evaluate the fourth (PS) for sufficiency. After a rate problem, they see a ratio and set up the wrong equation. This is set-shifting cost, and it is invisible in topic practice.

  Two decisions separate a correct answer from a tempting one. First, before solving, spend ten seconds naming the type and the task. Second, if the question resists your first method, abandon it; do not force the previous schema. Pacing also resets per question, not per block.

  Why the lower score is fine: the real exam is mixed. Your mixed score is the honest predictor. The gap between topic and mixed accuracy is your integration deficit, and it closes with mixed sets, not with more topic drills.

  You are ready to move on when

  • You classify every question in a mixed set before solving, and your classifications match the official types without correction
  • You trace each wrong answer in a mixed set to either a knowledge gap or a switching error, and you can name which
  • You finish a mixed set with no instance of applying the previous question's method to the current one

  Exam skills

  Mock 3: testing a section order

  All three Focus sections run $45$ minutes and carry equal weight, so your order changes nothing about the clock — only the state you are in when each section begins. That is the entire variable, and it is what Mock 3 exists to measure. Treat the resulting score as evidence about state, not about ability.

  The disguises are quiet. A dip in the third section reads to you as a content weakness. A sluggish first section reads as unusually hard questions. A rushed final five minutes reads as bad luck. All three are order artefacts, and all three will follow you into the real exam if you diagnose content instead.

  Two decisions matter. First, where the single $10$-minute break falls: after section 1, both remaining sections start rested; after section 2, only the last one does. Second, identify the section most damaged by a cold start — by careless errors, not by difficulty — and place that one after the break rather than first.

  The standard mechanical failure is moving several axes at once: new order, new break position, new pacing rule. Mock 3 then tells you nothing. There are $6$ orders $\times$ $2$ break positions $=12$ configurations; spend one per mock. And since every Focus section is question-level adaptive, its opening questions cost the most — you cannot warm up on question 1. Run a $30$-second checklist in the gap before each section: section name, pacing checkpoints, the one trap you keep re-entering. You can still change order at the test centre, so this is rehearsal, not lock-in.

  You are ready to move on when

  • Before starting Mock 3 you wrote down the order, the break position, and the single axis you changed from Mock 2
  • You ran the pre-section checklist in the gap before all three sections, and answered the first question of each at your intended pace
  • In review you separated order effects from content effects by checking whether the dip tracked the break position rather than the topic

  Open week 9 on its own page →

• 10 Timed sprints & when to guess TIME PRESSURE 20–30 hrs

  Speed work belongs here, after accuracy is established, because speed applied to a shaky method produces wrong answers faster. The most valuable skill this week is leaving a question.

  Quantitative

  Estimation as a deliberate method

  On GMAT Focus Quant, estimation is a deliberate bound, not a lazy shortcut. The exam rarely asks you to estimate explicitly. Instead it spaces answer choices widely enough that exact computation is optional — and includes questions where that spacing is a trap, because two choices differ by less than the error your rounding introduces.

  The disguise is a word problem with messy numbers: profit margins, mixture ratios, rates with decimals. You can eliminate three choices by order of magnitude in seconds. But the same technique fails when the question asks for an exact property — a remainder, a unit digit, whether a value is an integer — because approximation can suggest exactness but cannot confirm it.

  Three decisions separate correct from tempting. First, compare the gap between adjacent choices to your expected rounding error; estimate only if that error is less than half the smallest gap. Second, round factors in opposite directions so errors cancel. For $a\times b$, if you raise $a$ by $p\mathrm{\%}$ and lower $b$ by $q\mathrm{\%}$, the product changes by roughly $(p-q)\mathrm{\%}$; $19.8\times 5.1$ is better as $20\times 5=100$ than as $20\times 5.2=104$. Third, in Data Sufficiency, estimation can show a statement is insufficient by yielding two different plausible values, but it cannot prove sufficiency.

  The mechanical loss: strong candidates round every number the same way, compound the error, and land on a distractor that differs from the true value by exactly that compounded amount.

  You are ready to move on when

  • You can look at a Quant question's answer choices and decide within ten seconds whether estimation is safe, by comparing the smallest gap to your expected rounding error
  • You can solve a messy arithmetic problem by rounding one factor up and another down, landing in the correct choice without exact computation
  • You can identify a question where estimation is tempting but invalid because it asks for an exact property, and switch to exact reasoning
  • In Data Sufficiency, you can use estimation to demonstrate insufficiency but refuse to use it to claim sufficiency

  Exam skills

  Pacing benchmarks per section

  GMAT Focus does not test whether you can solve a question quickly. It tests whether you can allocate a fixed clock across a section whose questions vary wildly in cost. The benchmark is not a per-question cap. It is a checkpoint: a time-remaining figure at a known question number that tells you whether the current question is affordable.

  The published section clocks give you the only averages you need. Quant: $45\mathrm{/}21\approx 2:09$ per question. Verbal: $45\mathrm{/}23\approx 1:57$. Data Insights: $45\mathrm{/}20=2:15$. Treat these as the slope of a line, not a rule. By the halfway question you want roughly 22 minutes left; by the three-quarter question, roughly 11.

  The disguise is a question that feels fast — a short Quant prompt or a single-paragraph Verbal — but requires a full algebraic solve or a subtle scope shift, eating three minutes against a two-minute benchmark. The reverse disguise: a dense Data Insights set early that makes you think you are behind, when the checkpoint at question 5 shows you are ahead.

  The decisions that separate scores: check the clock only at fixed question intervals, never continuously; abandon a question when its elapsed time exceeds its budget and your checkpoint says you cannot recover; spend surplus time only on questions you are certain you can solve. The mechanical loss is enforcing the average as a cap: rushing a three-minute question at two minutes, getting it wrong, then burning the saved minute on a question you cannot solve anyway.

  You are ready to move on when

  • After a timed section, you can state your time remaining at each quarter-section checkpoint without looking at your notes
  • You can name the moment in a practice set when you abandoned a question solely because its elapsed time exceeded its budget, and you did not return to it
  • You can identify a question that cost you more than its share of the benchmark and explain what you would have done differently at the decision point

  Exam skills

  When to guess and move on

  Guessing on GMAT Focus is not what you do to questions you cannot solve. It is a decision you make about questions you could eventually solve, before they take a second question down with them. There is no wrong-answer penalty, so every question must end with a selection; the only variable is what that selection costs you.

  The real test is triage under sunk cost. A guess at ninety seconds and a guess at four minutes score identically. The second one also compresses everything after it. That escalation — not ignorance — is where strong candidates bleed: they refuse to abandon a question they are nearly finished with, guess anyway, and start the next question rattled and behind.

  Triggers are mechanical, not emotional: you have re-read the same sentence twice with no new information; you can name two answers but no reason to prefer either; you are doing work the question did not ask for; the setup looks familiar but your first step is not converging.

  Two decisions separate the useful guess from the expensive one. Guess on your own terms — before the clock decides, while you can still read the next question properly. And guess with an elimination, not blind: removing one option turns $1\mathrm{/}5$ into $1\mathrm{/}4$, which is often a full minute of work avoided. In Data Insights, partial selections earn nothing, so a half-built multi-part answer is worth exactly a blank — commit to a full guess.

  Flag and return only if you finish the section with time and a specific new idea. A second look rarely changes the answer and always spends time you will want on the questions still ahead.

  You are ready to move on when

  • You can state your guess trigger conditions out loud before starting a practice section, and afterwards identify each question where you applied them
  • In a finished practice set, every question has a selected answer and you can list which ones were guesses rather than solves
  • Your guesses are never blind — you can name the option you eliminated and why, on every question you abandoned

  Exam skills

  Timed sprints and how to run them

  A sprint is four to six questions, one skill, a hard buzzer, and a post-mortem longer than the sprint itself. Skip the post-mortem and you have practised answering quickly, which is not the skill.

  Sprint on one question type, or one decision — "does the second statement add anything?" — never a mixed set. Mixed sets produce a number, not a diagnosis. Use material you can already solve untimed at roughly 75–85%; harder questions make you slow for good reasons, and you will train yourself to abandon reasoning that was about to pay.

  Set the clock from your own data: sprint minutes $=n\times m$, where $m$ is your recent untimed median for that question type. Hold the stop, then finish the question anyway and log the overrun. Whether the overrun changed the answer is the only evidence that speed, rather than knowledge, is your binding constraint.

  The recurring error is scoring the sprint on time alone. Six correct in eight minutes is worse training than five correct with one candid loss. Score correct-at-target, and keep time-to-first-move separate from total time: most lost points in this range come from entering late, not from working slowly. A question you committed to at $t=40$s and closed at $t=140$s costs what one opened at $t=0$ and closed at $t=100$s costs — but only the first shows you where the time actually goes.

  You are ready to move on when

  • You can produce, for your last sprint, both a median time-to-first-move and a median total time, and say which of the two moved when you went faster
  • Your sprint log holds at least three sprints of the same single skill at similar difficulty, so accuracy-at-target is comparable across them rather than a fresh impression each time
  • You can name every question in the last sprint that ran past the buzzer and state whether the overrun changed your answer

  Open week 10 on its own page →

• 11 Harder questions & recovery MOCK 4 HARDER QUESTIONS 25–35 hrs

  Deliberately hard practice, plus the thing nobody rehearses: what you do in the eight minutes after a question goes badly. Recovery is a trainable skill and it is worth points.

  Quantitative

  Quant at the hard end

  At the hard end of GMAT Focus quant, the computation is rarely the obstacle. What separates the top band is whether you extract the constraint before you start manipulating. You will be handed a system with more unknowns than equations, plus one sentence that closes it — a domain restriction in a parenthetical ($n$ is a positive integer), a units conversion in a subordinate clause, or a geometric fact (the line is tangent) placed in the prose rather than the diagram.

  Its disguises: answer choices that are expressions in a variable rather than numbers; "which of the following could be" instead of "must be"; percent changes applied in sequence and described in words; and constraints stated as divisibility rather than equality.

  Three decisions carry the marks:

  • Solve for what the question asks. Hard items routinely produce $x$ and then ask for $x+y$, a ratio, or a change from the original value.
  • Test the choices before solving. When the choices contain a variable, pick a number consistent with every stated constraint — integrality, sign, and range included.
  • On Data Sufficiency, a statement that links two unknowns is almost-sufficient. Ask whether the other statement supplies the missing absolute value.

  The mechanical loss: carrying a correct algebraic result straight to the answer without re-reading the final clause — "how much greater than", "what fraction of the original" — or silently assuming positivity, or integrality, that the question never granted.

  Worth memorising: for an evenly spaced set, $\text{sum}=(\text{count})\times (\text{middle term})$. The condition is constant spacing.

  You are ready to move on when

  • You can name the single stated constraint that makes a hard quant problem solvable before doing any algebra
  • On problems with variables in the answer choices, you eliminate at least one choice by testing values rather than solving
  • In review, you can point to the exact clause you misread or the assumption you supplied that the question never granted

  Verbal

  Verbal at the hard end

  Hard Verbal is not harder to read. At this level stimuli and passages are often plainer than mid-level ones; the difficulty is loaded into the choices, where five options all sound like things the author might say. What separates them is scope, and scope is checkable.

  In Critical Reasoning, the conclusion is often the second clause — a “however”, “therefore” or “clearly” buried mid-paragraph — while the opening sentence is a premise that reads like one. Locate it before you look at the choices, then fix the argument's exact commitments: same group, same comparison, same time horizon. Hard correct answers tend to add a qualifier (“among those surveyed”, “in the first year”) that makes them narrow and dull; hard wrong answers are broader, smoother, and sound more like the passage.

  Assumption questions reward the same instinct. At the hard end the right answer is usually a defender — ruling out a rival cause or an unmentioned group — not a bridge linking two stated ideas. On Reading Comprehension, watch the stem's verb: “the passage suggests” licenses one step beyond the text, “according to the passage” licenses none.

  The formula worth keeping: to test a necessary assumption, negate the choice — normally just the quantifier or the verb, not the whole clause — and ask whether the argument survives. Quantifier ladder: $some\ge 1$, $most>50\mathrm{\%}$, $all=100\mathrm{\%}$; the negation of “most” is “half or fewer”.

  The mechanical way strong readers lose points: they eliminate an answer for using vocabulary absent from the stimulus, then pick one that reuses the stimulus's vocabulary but answers a different question. Match words to the reasoning, not to the passage.

  You are ready to move on when

  • On a set of hard CR questions, you state the conclusion in your own words before reading any choice, and can name the argument's scope commitments
  • For every hard RC inference answer you select, you can point to the single sentence in the passage that licenses it
  • Reviewing your hard Verbal misses, you can name the one quantifier, qualifier or scope term that makes each wrong choice wrong, without reading the explanation

  Exam skills

  What makes a hard question hard

  Hard GMAT Focus questions are rarely hard because they test an idea you have not seen. They are hard because they stack decisions: three steps compound error (if each step has success probability $p$ and steps are independent, your chance is roughly $p^3$), the core operation is not named, and wrong answers are built to satisfy most of the conditions.

  The disguise is usually thematic, not mathematical. A weighted-average question arrives as a mixture problem; a divisibility question hides behind 'possible values'; a necessary-assumption question reads like a main-point question. What makes it hard is that you must recognise the core before you can use it.

  The decisions that separate correct from tempting:

  1. Do you solve, or do you test the hinge condition first? Hard questions often have one non-obvious constraint — an integer requirement, a 'not', a unit change. Find it before doing algebra.
  2. Do you switch representation when stuck? The mechanical way strong candidates lose points is to re-read and re-solve the same way for four minutes. If 90 seconds pass with no new information, change form or leave.
  3. Do you answer the exact question? The most common hard-question error is choosing an answer that is true but answers something else.

  Difficulty is not a signal to try harder on the same path. It is a signal to check whether you are on the right one.

  You are ready to move on when

  • I can name the hinge condition in a hard question before I begin solving it
  • I can classify every hard-question error I make as a setup error, a trap-answer error, or an execution error
  • I can switch representation within 30 seconds of recognising I am stuck, rather than re-reading the same text

  Exam skills

  Recovering mid-section

  Recovery is forward-only. You cannot return to a question, so everything here concerns the next sixty seconds, not the last four minutes.

  What breaks a section is rarely the hard question itself. It is the decision that follows it. Two shapes account for most of the damage: the sunk-cost hold, where you stay on one question until the clock rather than your judgement ends it, and the over-correction, where you try to reclaim the lost minutes inside the very next question, read at speed, and miss something easy. Both convert one lost question into three.

  The in-the-moment signal is not "this is hard". It is physical: you have re-read the stem without gaining information, you are re-deriving something you already established, or you hold two answers you cannot separate on any stated criterion. Any of those is a cut signal, not an effort signal. Guess, mark it, move.

  Keep one number. For a section of $Q$ questions in $T$ minutes, your position at question $k$ should be near $kT\mathrm{/}Q$. Behind that line, recovery is a budget change for the next question only — not a permanent speed increase. Ahead of it, do not automatically spend the surplus.

  The counterintuitive move: after a blow-up, take the next question at normal pace. A second consecutive miss is what compounds, and restoring a little accuracy is worth more than the minutes it costs.

  Do not narrate. "The section is gone" is a story, not information. The engine is scoring answers, not your commentary.

  You are ready to move on when

  • You can state, before starting a section, the question number and elapsed time of your checkpoint for that section, from memory
  • In a timed set you cut at least one question at your own threshold rather than when the clock forced it, and can name the trigger you used
  • You can review a set where you missed two or more consecutive questions and label each loss as sunk-cost, over-correction, or genuine difficulty
  • After deliberately abandoning a question in practice, your time on the next question falls within your normal range

  Exam skills

  Mock 4: the last full rehearsal

  The last full mock is neither a score forecast nor a content diagnostic — you already know which content is weak. It is the only rehearsal of the procedure you will actually run on test day, and it is the last one. So freeze everything before you sit it:

  • Section order
  • Whether you take the optional break after section one
  • Your abandon trigger — the condition that makes you cut a question and move on

  If you are still choosing a section order or trialling a new pacing rule, you are measuring the wrong thing. Run Mock 4 in exactly the configuration you will use for real, start to finish, no pauses, nothing open in another tab. A mock taken in fragments teaches you nothing about stamina or clock behaviour.

  The output to harvest is timing, not marks. Per section your budget is $\frac{45}{N}$ minutes per question — roughly $2.1$ for Quant's 21, $2.0$ for Verbal's 23, $2.25$ for Data Insights' 20 — but treat that as an average, never a per-question rule. What matters is whether you were behind at two consecutive quarter-section checkpoints. One checkpoint behind is noise; two is a genuine deficit. Reaching Verbal question 12 at minute 30 leaves you about six minutes — three questions — in the red, and that is when the trigger fires, not when you feel stuck.

  The mechanical failure: strong candidates rework every wrong question for content, learn a lot, and finish with no written pacing plan. Or they change strategy because the total moved, which is chasing noise.

  Carry forward one page: order, break, four checkpoint times per section, trigger. Change nothing after this mock except volume.

  You are ready to move on when

  • You can sit Mock 4 start-to-finish in one sitting in your fixed section order, with the break decision made before question one
  • For each section you can state where you stood at every quarter checkpoint and name the checkpoint you were behind at
  • You can recite your complete test-day procedure — order, break, checkpoints, abandon trigger — from memory, and Mock 4 was the last change you made to it

  Open week 11 on its own page →

• 12 Final week: test readiness TEST READINESS Flexible

  Taper, do not cram. The score is already largely set; this week is about arriving able to use it.

  Exam skills

  What to do — and stop doing — in the last seven days

  The last seven days cannot add much content. They can add reliability. Judge everything you do now by whether it makes your existing ability show up under the clock.

  Do one final full mock no later than five days out, then stop. Its purpose is not a score; it is a timing map. For each section, note where you crossed your checkpoint — roughly $2$ minutes per question, so question $7$ by minute $15$, question $14$ by minute $30$. The pattern matters more than the result.

  After that, switch to timed sets of $5$–$10$ questions in your two slowest areas. Untimed practice now is worse than nothing: it rehearses a test that does not exist. Stop reading new strategy. Stop starting new question sources. Stop re-doing questions you already get right, unless you got them right slowly. And stop taking full mocks in the last three days — a late bad score changes nothing except your confidence.

  The decision that separates a good final week from a wasted one: when a question runs past $2$ minutes and you cannot see the end, guess and move. You will not out-think the clock; you will only out-decide it. Apply this from the first question, not after you are already behind.

  You are ready to move on when

  • You have taken your final full mock and written a section-by-section timing map from it
  • You can state your per-question abandonment threshold and have applied it in a timed set without reviewing the answer first
  • You have a written list of the two or three question types where you overspend, and you have completed a timed set on each

  Exam skills

  Choosing your section order for good

  Section order is a one-time, irreversible decision made before the clock starts: three sections of 45 minutes each, one 10-minute break after the second, and no adapting once you begin. That structure — not the content of any section — is what you are actually choosing between.

  The question rarely arrives as which order? It arrives as its absence. Most candidates run the order their last practice test happened to use and mistake that for a decision. The costlier version: you rehearse in one order for eleven weeks, then switch on test day because Quant "feels better first", and your pacing — which is a habit calibrated to position, not a skill — drifts in the final section.

  Three decisions separate a good order from a defensible one. Decide where your weakest section sits, not your strongest; you are allocating attention, and a section you find easy will look after itself. Treat the break as a hard reset and place your most draining section immediately after it. Then give the final slot to whichever section degrades least under fatigue — for most people that is not Data Insights, the most reading-dense and least forgiving of a wandering mind.

  The mechanical failure is building pacing norms on an order you do not use. If you change order, you must have sat at least one full-length test in the new one, break included.

  No formula here. What you memorise is the constraint: 45 minutes a section, one break after the second, nothing renegotiable afterwards.

  You are ready to move on when

  • You can state your chosen order in one sentence, naming which section sits last and why that section tolerates fatigue better than the other two
  • Your most recent full-length practice test used that order exactly, break included, and your final-section accuracy was within your normal range
  • You fixed the order at least a week before test day and have not changed it since

  Exam skills

  Test day: logistics and the first ten minutes

  Most test-day losses are administrative rather than mathematical, and they are settled in the ten minutes before question one.

  At a centre, arrive thirty minutes early: lateness forfeits the appointment, not a slice of it. Check-in runs photograph, palm scan, signature, locker — phone, watch, notes and water all go in. You get a fine-tip marker and a laminated booklet, so request a spare marker at the desk rather than raising a hand mid-Quant. Online, the room scan and ID capture consume the same window with none of a proctor's goodwill, so run the system test the day before.

  The untimed screens before section one — non-disclosure, instructions — are free minutes. Write your timing landmarks on the booklet there. Cap the marker between uses; a dried tip costs a raised hand and two minutes of clock. Do not hoard booklet pages.

  The mechanical way strong candidates drop points is overrunning the ten-minute break: the next section's clock starts without you. Budget eight minutes, keep two as buffer, and read the section clock the instant a section opens, not at question three.

  Disguised as strategy, the opening questions reward restraint. The first item is often a deliberately unfamiliar-looking one, and treating that as a signal to recalibrate is the error. Keep one landmark in memory: in a 45-minute section, the halfway question count should already be behind you at the halfway clock mark — question $10$ by minute $20$ in Quant. It requires no arithmetic and catches a slow start within two questions.

  You are ready to move on when

  • You can list the check-in sequence and exactly what goes into the locker without consulting notes
  • You wrote your timing landmarks on the laminated booklet during the untimed preamble in your most recent full mock
  • You took a break of eight minutes or less in a full mock and resumed with no section time lost

  Open week 12 on its own page →