How to stop losing GMAT and GRE points to careless mistakes: a system for catching your own errors

By David Baum, founder of Roughwork.

Careless mistakes are methods spoiled by a wrong step you do not notice: a dropped sign, an arithmetic slip, a misread question, a miscopied number. You catch them by making your work visible, checking every step against a fixed routine, and keeping an error log.

This is what separates your practice score from your test score. You know how to solve the problem. You lose the point anyway, on a line you would have caught if you had looked. The good news in that sentence is the whole method: a careless error is a step you already know is wrong, which means it is catchable by a habit rather than by learning more mathematics. Below is the habit, error type by error type, and the routine that makes catching them automatic instead of a matter of luck.

Why do I keep making careless mistakes?

Because a careless error is a lapse of attention, not a gap in knowledge, and attention is exactly what time pressure and mental arithmetic drain first. You are holding several sub-results in your head, moving fast, and skipping the work you would write out if you had time. The mistakes cluster where you cut corners.

There is a precise name for what is happening, and it comes from the study of human error rather than from test prep. The psychologist James Reason, whose book "Human Error" is the standard reference on the subject, separates two kinds of failure. A mistake is a flaw in the plan: you did what you meant to do, but the plan itself was wrong. A slip is a flaw in the execution: the plan was right, and the action went wrong anyway. A careless quant error is a slip in exactly Reason's sense, right intention and wrong execution, which is why studying more content does not fix it. You did not plan to drop the sign. Your hand dropped it while your attention was three steps ahead.

Three conditions produce most of these slips. First, invisible work: steps done in your head leave nothing to check, so an error has nowhere to be caught. Second, speed without a checkpoint: under a per-question clock you move to the next line before verifying the last one. Third, a familiar problem type: confidence lowers your guard, and the questions you find routine are where you stop reading carefully. None of these is fixed by studying harder content. They are fixed by changing how you work.

What are the most common GRE and GMAT quant errors?

Four types account for the large majority of lost-point mistakes on quant: sign errors, arithmetic slips, misreading the question, and copy or transcription errors. A fifth, order-of-operations, shows up in dense expressions. Each has a specific tell and a specific check, so it is worth treating them separately rather than as one blur of "careless."

Sign errors

A sign error is losing or flipping a negative: distributing a minus across a bracket, moving a term across the equals sign, or squaring and dropping the sign of a root. They are the most common single category because the minus is small, quick to skip, and changes the answer completely.

The check is mechanical. When you distribute or move a term, say the sign out loud or write it before the number, not after. On any step that crosses the equals sign, confirm every term flipped, not most of them. Sign discipline is a habit of writing the sign first, and it is learnable in a week of deliberate attention. For the deep version — three worked examples and the checks step by step — see how to stop making sign errors on GMAT and GRE quant.

Arithmetic slips

An arithmetic slip is a wrong calculation in an otherwise correct method: seven times eight recalled as fifty-four, a carry dropped in a subtraction, a fraction added across the top and bottom. The method is sound; the number is wrong.

Two checks catch nearly all of them. Estimate first, so you know the size of the answer before you compute it and a slip that lands far from the estimate announces itself. Then, on any multi-digit step, redo it a second way: multiply in the other order, or subtract to confirm an addition. This costs seconds and returns points.

Misreading the question

Misreading is answering a different question from the one asked: solving for x when it wants x plus two, giving the perimeter when it wants the area, missing the word "not" or "least" or "must." The mathematics is correct and the answer is wrong, which is why this one hurts.

The check happens at both ends. Before you start, underline what the question actually asks for and any constraint words ("integer," "positive," "not"). When you finish, read the question again and confirm your answer is the quantity it named. On data sufficiency in particular, re-read what is being asked before you judge the statements.

Copy and transcription errors

A transcription error is a number that changes as it moves: 17 copied from the stem as 71, a term dropped when you rewrite an equation on the next line, a figure read off the wrong row of a table. Nothing in the method is wrong. You solved a problem that was not on the page.

The check is to copy once and verify once. When you bring a number down from the question or from a previous line, look back at the source and confirm the digits before you keep going. Most transcription errors happen in the handoff between lines, so that handoff is where the glance belongs.

How do you catch your own math mistakes?

You make your work visible and then check it against a fixed routine, the same routine every time, so that checking does not depend on whether you happen to feel careful that day. This is the core advice every serious prep source gives, and it is correct: write your steps out, review each one, and verify the answer before you commit to it.

The self-checking techniques worth building into that routine are few and reliable. Use the opposite operation: substitute your answer back into the original equation and confirm it holds. Use estimation: form a rough size for the answer before you compute, and distrust any result that lands far away. Re-solve the fastest step a second way when the arithmetic mattered. And read the question a final time against your answer. None of these require new mathematics. They require doing the check every time rather than only when something already feels off, because the errors you most need to catch are the ones that did not feel wrong.

The limit of manual self-checking is honest and worth naming: you are checking your own work with the same attention that produced the error. When you are tired or rushed, the review is exactly as careless as the step was. That is the gap the rest of this system is built to close.

How do you keep an error log?

An error log is a record of every mistake you make in practice, written down with enough detail that you can see the pattern. For each miss, note the question, the specific wrong step, and which of the four types it was. Reviewed weekly, the log turns a vague sense of "I make careless mistakes" into a short, specific list of the two or three that actually cost you points.

Test-prep instructors who work with this segment are close to unanimous that the log is the highest-return habit available, and that its real value is diagnostic. Elaine Loh, a Manhattan Prep GMAT instructor, puts the point directly:

The awareness is the part that matters for careless slips, because the trap with these errors is treating them as beneath attention. Mike McGarry, writing for Magoosh, names that trap:

The log exists to stop you dismissing the miss. McGarry's own remedy is to write it out in full: forcing yourself to articulate, in words, exactly what the mistake was and exactly what you need to remember next time, which is why a log entry is a sentence about the step, not a tally mark.

Keep it simple enough that you will maintain it. A row per error with four columns is enough: the problem, the step where it went wrong, the error type, and the correct step. The value is not in writing it down once. It is in the weekly read, where you find that most of your lost points come from a narrow set of repeated slips, and that a targeted drill on those beats another full practice set. The log is the difference between making the same sign error for three months and retiring it in two weeks.

Should you write out quant by hand?

Yes. Writing quant by hand is the single change that makes every other check possible, because a step done in your head cannot be reviewed and a step on paper can. The prep advice to keep neat, legible work is not about tidiness. It is about leaving a trail an error can be caught on.

Handwritten work does two things at once. It slows you at the exact moments where careless errors happen, the distribution, the sign, the carry, so the step gets the half-second of attention it needs. And it produces the visible record that your end-of-problem check and your error log both depend on. Neat is not the point; visible is the point. If your work is a scatter of half-steps across the page, the review has nothing solid to run against, so structure your steps down the page, one step per line, the way you would want to read it back.

How do you stay accurate under time pressure?

You keep accuracy and speed from trading against each other by building the checks into the work itself, so that checking is not a separate step you skip when the clock is tight. Careless errors spike under time pressure precisely because the review is the first thing to go. If the check lives inside how you write each line, there is nothing left to cut.

In practice this means checking as you go rather than at the end: write the sign before the number, estimate before you compute, confirm the transcription in the same glance that copies it. A per-line habit survives time pressure because it is not optional and it is not extra. Then time your practice from the start, so that the routine is trained under the same conditions you will face on the test. Accuracy that only appears when you have unlimited time is not accuracy you can use on exam day.

The system in 5 steps

1. Write every step by hand, one step per line, so your work is visible and reviewable.
2. Estimate the size of the answer before you compute it.
3. Check as you go: sign before the number, transcription confirmed in the same glance, opposite operation on the step that mattered.
4. Re-read the question against your answer before you commit.
5. Log every error by type, and review the log weekly to drill the two or three slips that actually recur.

What marking handwritten work reveals about careless errors

Roughwork's whole job is to read a student's handwritten steps and mark them, and building a tool that has to look at the actual page, line by line, changed how I see careless errors.

The errors that cost points are almost never in the steps people write down. They are in the steps people do in their head and never commit to the page: the sign distributed mentally, the multiplication not worth writing, the number carried across a line from memory. To the person solving, those steps feel done, so there is nothing to look back at. On the page it is different. The line before is right, the line after is wrong, and nothing in between explains the jump. The error is not hidden because it is subtle. It is hidden because it was never written.

That is the honest reason "check your own work" fails on exactly the errors you most need to catch. You can only review what you wrote, and the slip lives in what you did not. Reviewing an hour later makes it worse, because by then you reconstruct the missing step the same way you did it the first time, with the same wrong reflex, and it reads as correct again.

The conclusion I kept arriving at is that careless errors need a second set of eyes on the steps at the moment they happen, not later and not your own. That is the specific job we built Roughwork to do, and it is worth being exact about what it does and does not do. It catches a wrong step while the page is still in front of you, and re-serves a step you missed so you can practice it again. It does not do the thinking for you, it does not write the answer onto your page, and it does not fill in the steps you left out. It does not keep your error log either; the record stays yours to keep and to read for patterns, because the weekly read is where the value is and it only works if it is yours.

Where Roughwork fits

Everything above is manual. It works, and it is what every prep course teaches, but it asks you to review your own work with the same attention that produced the mistake, and to keep the error log by hand.

There is a reason that is hard, and it is not a lack of discipline. In his paper on how errors are best managed, James Reason argues that the durable fix is rarely to try harder to be careful:

That is the principle Roughwork is built on. You cannot reliably will yourself into catching your own slips, so you change the conditions: you put a second reader on the steps as you make them.

Roughwork is practice with instant feedback on your handwritten work. You solve a problem by hand with Apple Pencil, and when a step is wrong, Roughwork catches the mistake and explains what is wrong in the app. You make the correction yourself. It does not do the thinking for you, it does not write the answer onto your page, and it does not add the missing steps. It shows you where your work goes wrong and tells you what is wrong, which is the review step you would otherwise have to run on yourself.

Because it checks the page while the work is still in front of you, Roughwork does the hardest part of the review for you: spotting the slip you did not feel, with attention that is not already tired. It also brings a mistake back to practice until you have put it right. Keeping the error log and reading it weekly for patterns is still yours to do; what Roughwork removes is the catching, not the record-keeping.

FAQ

References and further reading

• James Reason, "Human error: models and management," BMJ, 2000;320:768–770. The source of the principle that the reliable fix is to change the conditions people work under, not to demand more care. DOI: 10.1136/bmj.320.7237.768.
• James Reason, Human Error, Cambridge University Press, 1990. The standard reference on the slip versus mistake distinction used above. Publisher page.
• "Human Error and Patient Safety," in Textbook of Patient Safety and Clinical Risk Management (NCBI Bookshelf). A concise account of Reason's slips, lapses, and mistakes framework. Read it.
• Mike McGarry, "Careless Mistakes on GMAT Math," Magoosh. On why dismissing a miss as "careless" is a way of refusing to learn from it. Read it.
• Elaine Loh, "Error Log: The #1 Way to Raise Your GMAT Score!!," Manhattan Prep. On the error log as the habit that makes you aware of careless errors. Read it.