Cracking the GMAT Focus: 6 Principles + The Tactics That Actually Move

Question

Most GMAT advice is content advice. Learn this formula, memorise that rule, drill 400 more questions. That advice is fine, and it stops working somewhere around the 655 mark.

What breaks after that point isn't knowledge. It's decision-making - what you choose to read, what you choose to skip, when you decide your method is wrong, and how you handle the ninety seconds after a question goes badly.

This post is a write-up of a strategy session I put together after a personalised tutoring block with a well-known GMAT tutor. It is deliberately unglamorous. There is no promise of a 705+ in 30 days here. There are six principles and about fifteen tactics, and most of them contradict what you have been told.

Keep a pen handy. Some of this is worth writing down.

The Problem: Why DI Breaks High Scorers

There are three ways people fail Data Insights, and almost everyone misdiagnoses which one they have.

You can't finish.  You average 15–17 of 20 attempted, and the last four get random Cs. 
  You finish, but you err.  Clean prep, clean pacing, and you still bleed easy points to procedural slips. 
  You panic mid-section.  One brutal MSR at Q7 and the next three questions are collateral damage.

Here is the part that matters:  none of these is a timing problem. 
 
All three are decision-making problems wearing a timing costume. And that distinction is expensive, because if you diagnose "I need to go faster," you will apply speed to a broken process - and a broken process executed faster produces wrong answers faster.
 
  Going faster on the wrong process makes things worse. Change the process first. The speed is a side effect.

Part 1: Six Principles

These six sit underneath everything else in this post. If you only take one thing away, take these.
 
  Principle 1: There are no trick questions. Ever.  
 
This is the single most important mindset shift, and it is the one people resist hardest.
 
The official test is meticulously written. It does not deceive. It does not bait. It does not hide things in the phrasing to catch you out. Every published question means exactly what it says.
 
So when you catch yourself on a tangent where your reading contradicts the common-sense meaning of the sentence - stop. You are not being clever. You are inventing complexity that is not in the problem.
 
A concrete example: in CR, if a statement uses present tense ("they always choose X"), that means an ongoing pattern. It is not a coded hint about "a past trend that might have changed." That reading is your brain manufacturing trickery.
 
Two mid-problem checks:

"Am I interpreting this as roughly the opposite of what it literally says?" → Stop. Re-read. 
 "Am I assuming the test is hiding something?" → Stop. It isn't.

Payoff:  this principle alone saves most people 60–90 seconds on the problems where they'd otherwise spiral.
 
  Principle 2: GMAC regulates the work  
 
Every published official problem is calibrated so that the total cognitive work fits roughly a two-minute budget for a competent test-taker. There is no official problem that legitimately requires five minutes of honest work. That problem does not exist.
 
Which gives you a diagnostic. If you are sitting at 3:30 on a single question, that is not a hard problem. That is your method failing. There is a simpler path you have not seen, and continuing down the current one will not reveal it.
 
The flip side also holds: if you solve something in 35 seconds and it feels suspiciously easy - fine. Mark it, move on. Don't audit your own success.
 
There's a useful corollary here, and it comes up constantly in DS:  the wordier the prompt, the less work the statements will require.  GMAC distributes effort. They will not make you do heavy prompt analysis and heavy statement analysis on the same question.
 
  If you're past three minutes on an official problem, the problem isn't the problem. Reset and look for the simpler path.  
 
  Principle 3: Attention is finite. Allocate it deliberately.  
 
On test day your working memory will be measurably worse than it is in practice. Stress, fatigue, an unfamiliar chair, a proctor walking past. All real, all documented, all going to happen. Plan for a degraded version of yourself.
 
Three implications:

Don't spend attention on details you won't use.  DI problems are built with surplus information. The large majority of MSR text, table rows and graph segments are irrelevant to the specific question in front of you. 
  Don't read passages before you know what you're looking for.  Identify the goal, then go fetch. Reading first and hoping it sticks is how you burn four minutes and retain nothing. 
  Do spend attention where the test tells you to.  A word like "type" in an MSR, "same" in a TPA, an underlined EXCEPT in CR - those are load-bearing. A wordy restatement of something already established is not. Skim past it.

This is what fast test-takers are actually doing. They are not reading faster. They are reading less.
 
  Principle 4: Find the goal before you move  
 
Most test-takers start solving before they know what they are solving for. They read the prompt, pattern-match to a method, and discover at 2:10 that they've been computing the wrong quantity.
 
The fix takes four seconds: before any algebra, before any table-scanning, write the goal on your scratch pad as a single phrase. What exactly am I solving for?
 
Then carry that phrase through every step and discard everything that doesn't serve it.
 
This has immediate teeth in Quant. If a problem asks for 4x2, the move is to manipulate toward 4x2 directly - not to solve for x and then square it and multiply by four. More on this below.
 
  Principle 5: Read literally. Read precisely.  
 
The test is written with surgical precision. Single words flip answers. Three classic cases:

"Those same X"  - the word "same" ties the reference back to a specific subset already mentioned, not to all X. Miss it and your answer inverts. 
  "Type of"  - vague on its own, and always defined somewhere in the data. If an MSR asks about "type of artwork," does type mean medium or era? The passage will tell you. Go look. 
  Present tense  - "they always choose," "the train runs every hour." Ongoing pattern. Not a past trend with a hidden complication.

And the meta-rule:  if GMAC wants you to focus on a word, they make it visible.  They underline it, capitalise it, or italicise it. EXCEPT and NOT in CR are always signposted. If you see formatting, that is the test pointing at something.
 
  Principle 6: There are no silly mistakes  
 
Stop using the word "silly."
 
"Silly" implies the error fell out of the sky and landed on you. That framing quietly damages your prep, because you cannot fix something you didn't cause. It disclaims responsibility while sounding humble.
 
The accurate word is  procedural . You misread a sign. You transcribed 8.05 where the number was 7.05. You dropped a "not." These have causes, they have patterns, and they have fixes.
 
Here's the test for it - and I'll come back to this one later, because it's the most useful mental tool in this whole post:
 
  Imagine you'd be paid $50,000 to finish the GMAT with zero procedural errors. No score requirement - just zero careless mistakes. Could you do it? Almost certainly. So what would you do differently?  
 
Whatever your answer is, that's your fix list. And you should be running it on the real test anyway.

Part 2: Data Sufficiency

Five tactical shifts. Each one is worth roughly 20–30 seconds a problem, which across a section is the difference between finishing and guessing.
 
  DS 1: Wordy prompt means the work is upfront  
 
Total work per problem is roughly constant, so GMAC redistributes it. When a DS prompt is heavy with repeated structures - "the closing balance," "the 15th of each month," "as of the end of the last fiscal year" - that is a signal. The work lives in the prompt. The statements will be quick.
 
 Trap:  skim the prompt, jump to statement (1), attempt all the analysis there. You'll burn three-plus minutes and probably still get it wrong.
 
 Fix:  compact the prompt first.

Identify repeated phrases and collapse them into one symbol or one line. 
 Derive the obvious implications before you look at either statement. 
 More often than not you'll discover the answer hinges on one specific unknown - and then each statement is a five-second check for whether it supplies that unknown.

OG2526 Q391 (the deposit/withdrawal problem)  is the canonical case. Long prompt, trivial statements, and everyone who reads it in the wrong order loses three minutes.
 
  DS 2: Solve for what's asked  
 
This is the most consistently violated principle in both DS and Quant.
 
When a problem asks for an expression - 4x2, or (a+b)/c, or x2 - y2 -> the path almost always runs through manipulating toward that expression. Not through finding the individual variables first.
 
 Trap:  you see x in the equation, so you instinctively hunt for x. Then square it, plug it back, and hope. Slow and error-prone.
 
 Fix:  ask "what algebraic operation gets me from the form I have to the form I need?"
 
A useful structural tell: if the variable only ever appears inside radicals and the question wants x2, expect to square twice. Once to clear the radicals and produce x-terms, again to produce x2-terms. Knowing to expect two squarings stops you from panicking after the first one leaves you with a mess.
 
  DS 3: The two statements never contradict each other  
 
This is a free verification tool and almost nobody uses it.
 
Every DS problem is constructed so statements (1) and (2) share at least one common solution. They cannot conflict. They cannot each be independently impossible. So if your analysis says statement (1) gives n = 24 and statement (2) gives n = 21 - you have misread one of them. Go back.
 
There's a sharper version for D answers: if each statement individually pins down a unique value, those two values must be identical. If yours aren't, you've made an error somewhere.
 
This isn't a shortcut for skipping work. It's a guardrail that catches procedural errors before you submit them.
 
  DS 4: Knowing one thing ⇔ knowing the answer  
 
This is the highest-leverage DS move in this post, and most people never internalise it.
 
DS asks about sufficiency, not computation. You don't need to find the answer. You need to know whether the answer is uniquely determined. So the efficient technique is to identify a single piece of information that is  equivalent  to "the answer is determined," and then check each statement against only that.
 
 Worked pattern - OG2526 Q407, Jane's credit card 
The question wants the average daily balance over a 30-day cycle. The balance sits at $600 and drops to $300 when the payment lands. So the average is a linear function of the payment date, and nothing else. Which means:
 
  Know the payment date ⇔ know the average. Sufficiency collapses to a single question: does this statement give me the date?  
 
Statement (1) states the payment was credited on day 21. Done, sufficient, no arithmetic. You never compute the average at all.
 
Every DS problem has some version of this reduction. Find it before you start grinding.
 
  DS 5: When algebra fails, grind the cases  
 
Algebra is powerful and it has blind spots. It has no native concept of:

whole numbers only 
 positive values only 
 "at least 10 people" 
 any real-world constraint at all

When a problem carries constraints like these, elegance is a trap. Drop to enumeration. Check whole-number solutions. Verify the boundaries. Don't skip steps to feel clever.
 
The reassurance:  GMAC knows the case count.  They do not write problems requiring you to check 50 cases. It's usually 2–4. If you find yourself on case 6, that itself is a signal you're on the wrong path - go back to Principle 2.
 
The classic instance is a linear equation in two variables with an integer constraint. The equation alone has infinite solutions; the integer constraint narrows it to a handful. There is no shortcut. Grind it, and trust that the grind is short.

Part 3: Two-Part Analysis

TPA is always about the link between the two blanks. Name the link and the answer usually falls out.
 
  TPA 1: The two blanks share a relationship - name it first  
 
Every TPA has some relationship binding the two answers. Your job is to state that relationship in one sentence  before  you look at the answer choices.
 
Typical shapes:

Blank 1 must be greater than Blank 2, with no further constraint. 
 Blank 1 implies Blank 2 through direct cause and effect. 
 Both must satisfy the same condition set but differ along one dimension.

Once you have the sentence, test choice-pairs against it. Most TPAs collapse fast, because only one pair survives cleanly.
 
 Trap:  solving each blank independently, as if it were two separate questions. That's the single most common TPA failure and it's usually slower than the intended path.
 
 Diagnostic:  if multiple pairs satisfy your stated relationship,  your relationship is incomplete.  Don't start guessing between them - go back and re-read for the extra constraint. It's almost always a single word or phrase, and it's often something like "those same X."
 
  TPA 2: "All X are Y" means draw a Venn  
 
The moment a TPA gives you multiple statements shaped like "All X are Y" or "Every X is a Y" - stop reasoning verbally. Draw circles.
 
Each "All X are Y" becomes an X-circle nested inside a Y-circle. Chain them. The diagram then shows you mechanically which inferences are valid and which aren't, and you stop trying to hold four nested conditionals in working memory.
 
A 30-second sketch routinely saves three minutes of verbal contortion. When you see three or more statements of this shape, just draw it.
 
 Watch the quantifier:  "SOME X are Y" is not a nested circle. That's an overlap. Different diagram, different valid inferences. Read the quantifier before you draw.
 
  TPA 3: Rows and columns for anything with units  
 
When a problem has multiple scenarios sharing quantities - rate × time = distance, $/hr × hours = total, gallons/hour × hours = gallons - draw a three-column table with the units written across the top.
 
Don't write "rt = d" and then start substituting. Write the actual units. The table becomes self-documenting.
 
Why it works:

It separates the data from the arithmetic. 
 It prevents transcription errors - the 3a/140-when-you-meant-a/140 class of mistake. 
 It makes "what's the same and what's different across scenarios" visible instead of remembered.

Make this a default habit. The moment you glimpse "X per Y" units in a prompt, the table goes on the pad. It eliminates an entire category of procedural error in both TPA and Quant.

Part 4: Multi-Source Reasoning

MSR is where most people lose whole minutes, and the fix is counterintuitive: read less, not more.
 
  MSR 1: Find the vague word, then go make it un-vague  
 
MSR is dense, and the instinct is to read every tab carefully before touching a question. Resist it. You'll spend four minutes and retain very little that's relevant.
 
 Better sequence:  read the question first. Identify which words in it are vague - terms that could mean two different things and would flip your answer. Then go to the tabs and look up exactly those.
 
The definition is always there. An MSR asking about "type of artwork" - does type mean medium (painting, sculpture) or era (16th century, 20th century)? There will be a list above the data table defining the four types. Now you know which categorisation to apply, and you read nothing else.
 
The discipline is simple:  if a word literally stops you from solving, that's the signal to go fetch its definition.  Otherwise, don't pre-read.
 
  MSR 2: Eliminate first, then plug in  
 
Standard sequence for MSR multiple-choice, and it's mechanical enough to run under pressure:

Read the question. Note the constraints. 
 Eliminate choices that violate any obvious constraint. Usually 2–3 of 5 die immediately. 
 For the survivors, plug each into the full constraint set. Whichever holds is the answer.

This beats solving forward, which on MSR tends to require chaining several tabs and gets tangled fast. The plug-in step also surfaces hidden constraints - you'll find that a fact like "the Davis Institute borrows in periods II and III" forces a particular configuration once you combine it with a candidate answer.
 
If you're staring at the screen with no move - start eliminating. Motion beats paralysis.
 
  MSR 3: Sometimes you really do have to connect tabs  
 
Default assumption: each question draws on one tab. Most of the time that's correct.
 
But when a tab is unusually sparse - three sentences, a small table, one chart - that's a signal rather than filler. Sparse tabs are high-signal, and they exist to be combined with something else.
 
  Scan the relative tab sizes before you start solving. If they're imbalanced, expect cross-tab work.  
 
This is maybe one MSR set in five. But when it happens and you don't notice it, you will spin for three minutes on a question that looks unanswerable - because from any single tab, it is.

Part 5: Table Analysis and Graphics Interpretation

TA/GI 1: Use the sorted view. Every time.  
 
Every Table Analysis problem has clickable column headers, and every click re-sorts the table. GMAC built this in deliberately. Use it religiously.
 
For any "extremes" question - highest, lowest, most, least - sort by the relevant column first. Your candidates are now at the top, and you test two or three instead of nineteen.
 
Concrete pattern, using a hockey-stats-style table:

Question: which player had the most penalty minutes per game? 
  Wrong:  compute PIM/GP for all 19 players. 
  Right:  sort by a correlated per-game metric, eyeball the top few, run the calculator on those two or three only.

Target: never more than 2–3 ratio computations per Table Analysis sub-question. If you're doing five or more, you aren't using the sort.
 
  TA/GI 2: Practise the on-screen calculator  
 
The DI calculator is not your phone's calculator. Different button layout, different cursor behaviour, and order-of-operations handling that can surprise you at exactly the wrong moment.
 
 The fix is embarrassingly simple:  30 minutes of pure calculator practice the day before your test. No problem-solving. No thinking. Just bang numbers into it until your hands know where things are.
 
This is the lowest-cognitive-load prep activity available to you, and it eliminates the panic-typing that produces "3.65 + 3.4 = 8.05" on test day.
 
One specific pattern worth killing: entering every number twice "to be safe." That doubles your time on every ratio. Practise enough that you trust your first entry, then move.
 
  TA/GI 3: The $50,000 check  
 
Back to Principle 6, because this is where it cashes out.
 
Imagine you'd be paid $50,000 to finish the GMAT with zero procedural errors. No score requirement. Just zero careless mistakes. Could you do it?
 
Of course you could. So what would you do differently?

You'd re-check signs. 
 You'd re-read the question stem before marking an answer. 
 You'd verify calculator entries instead of assuming. 
 You'd underline NOT and EXCEPT in your scratch work. 
 You'd write out 3.65 + 3.40 = 7.05 instead of doing it in your head.

Whatever's on that list - do it on the real test. Over a career, a 70-point score improvement is worth considerably more than $50,000.
 
  Procedural errors aren't accidents. They're decisions you already made about how much attention to spend - 
 you just made them unconsciously.

Part 6: Quant

Less philosophy here. Three tactics and a grind.
 
  Quant 1: Plugging in answer choices is clean, not cheap  
 
Most students treat plugging in as a fallback for when the "real" method fails. Wrong model. Plugging is frequently the cleanest available path, particularly when the answer choices are expressions rather than values.
 
Why it's clean:

It reduces the variable count to zero. 
 Every step is concrete arithmetic - no algebraic identity you might misremember. 
 The correct choice is simply whichever one matches your number.

Use it when the choices are simple values or expressions, when the equation is gnarly (radicals especially) but tractable at a specific x, or when you're stuck on the algebra and the clock is moving.
 
Worked example: √(3 − 2x) = √(2x) + 1, find 4x2. Pick x = 1/8, because then 2x = 1/4 and the radical is clean. LHS = √2.5 ≈ 1.58; RHS = √0.25 + 1 = 1.5. Close enough to confirm x ≈ 1/8. Then 4x2 = 4(1/64) = 1/16 ≈ 0.06, and you test each choice at x = 1/8 to see which produces it.
 
  Quant 2: Don't approximate when the answers are tight  
 
Estimation is excellent when choices are far apart and actively dangerous when they're close. The answer-choice spacing tells you the precision the problem requires - read it before you decide.

Wide spacing  (100, 200, 400, 800, 1600) → approximate freely. 
  Tight spacing  (2.55, 2.64, 2.72, 2.90) → compute exactly.

If you can't quickly decide whether the gap is wide enough to estimate - it isn't. Grind it.  
 
And remember Principle 2. GMAC regulates the work, so the exact answer is always reachable inside the budget. Approximation is a tool for specific situations, not a general-purpose speed hack.
 
  Quant 3: Ignore difficulty labels  
 
Third-party platforms label everything - 555, 605, 655, 705, 755. Treat these as noise.
 
Difficulty is deeply individual. A problem that's routine for you may be brutal for me, and vice versa. Those labels are aggregated guesses about a population, not facts about the problem in front of you.
 
Two ways they hurt you:

Before:  "this is a 705-level problem" → you tighten up, abandon your normal process, and choke on something you'd otherwise handle. 
  After:  "wait, that was a 555 and I got it wrong?" → spiral, which then costs you the next question too.

The cure is to not look. Not before, not during, not after. Each problem is just a problem - you solve it or you don't, and then you move.
 
This matters even more on the real test, where you have no idea what difficulty bin you're in and no way to find out. Stop trying to read the adaptive algorithm. It's not information, it's anxiety with a number attached.

The Recap: Six Principles

Everything above flows from these.

No tricks.  The test means what it says. If your reading contradicts the obvious one, re-read. 
  Work is bounded.  Past three minutes on an official problem, your method is wrong - not the problem. 
  Attention is finite.  Read less. Spend it where the test points. 
  Find the goal.  Solve for what's asked, not for what's solvable. 
  Read literally.  Single words flip answers. Formatting is a signal. 
  Procedural errors are yours.  Own them, and they become fixable.

These six replace most of the "tips and tricks" content floating around online. They're not exciting. They don't promise a fast score. They're just true, and they hold across every section.

What to actually do from here

Practice with official questions only. 
 After each problem, review  what your process actually was  - not just whether the answer was right. A correct answer reached the wrong way is a future wrong answer. 
 Treat every error as procedural, and ask the specific question: what attention allocation would have prevented this one?

You don't need to out-think the GMAT. You need to stop handing it free points.

Related Guides

How to improve your DI score from D75 to D82  
  How to Manage Time on DI  
  My Approach to Solving MSR  
  How to Attempt DS  
  Simple Framework for TPA  
  Estimate vs Accurate Calculation in DI  
  Tips to a Perfect Quant Score

	Target Test Prep - Save up to $599 $400 Discount + $99 Bonus GMAT Club Tests
	e-GMAT: 60% Off on GMAT Focus Get to 735+ from any starting point
	Magoosh increases your score, guaranteed - 15% off ALL Magoosh GMAT plans with code GMATCLUB15 + Free GMAT Club Tests.
	Manhattan Prep: $99 Bonus Score Higher on Your GMAT Exam. Guaranteed.
	Experts' Global - ~~$1720~~ $960 End-to-End GMAT Prep + App Support: 150+ Videos, 2000+ Questions, 15 FLTs +
	GMATWhiz - $299 ~~$499~~ Experience the course through a free trial.
	GMAT Club Tests - Free Included with every course purchase of $149+ View all deals

Cracking the GMAT Focus: 6 Principles + The Tactics That Actually Move

Prep Toolkit

Top 5 GMAT Debrief Videos

615 to 715 in 15 Days (Ria's Strategies)

More than Quant/Verbal Abilities: Speed vs. Accuracy

745 with Only Self-Prep and GMAT Club

From 555 to 765: 210-point Improvement

Perfect 805 Score Debrief by Julia

My Rewards