Does 700 out of 1000 mean I need 70 percent of questions correct?

No. 700 out of 1000 is a scaled score, which is your performance mapped onto a fixed reporting range the vendor defines. It is not a percentage of questions answered correctly. The raw number of questions you need right varies by exam form, so you cannot convert the passing mark into a fixed percentage.

What is the difference between a raw score and a scaled score?

A raw score is the count of questions you got right, or that count as a percentage of scored questions. A scaled score transforms that raw performance onto an abstract range like 100 to 1000 so results stay comparable across different, differently difficult versions of the same exam. The exam reports the scaled score, not the raw percentage.

Why can't I calculate how many questions I need to answer correctly?

Because three things are hidden from you: the difficulty adjustment applied to your specific form, the fact that some questions are weighted more than others, and the unscored trial questions mixed in silently. Without those, any target number of correct answers you compute is just a guess.

Is AWS passing at 720 out of 1000 the same as 72 percent?

No. Like Microsoft's 700, AWS's 720 is a scaled cut point on a 100 to 1000 reporting range, not 72 percent of questions answered correctly. The share of questions you actually need right shifts depending on your form's difficulty.

Why is CompTIA Security+ scored out of 900 instead of 1000?

Because each vendor picks its own reporting range. Security+ passes at 750 on a 100 to 900 scale. The fact that it does not end in 1000 is a helpful reminder that the passing number is a point on a scale, not a percentage of your questions.

What are unscored or pilot questions?

They are trial questions vendors seed into live exams to collect difficulty data before deciding whether to score them on future forms. They do not count toward your result, and you are almost never told which ones they are, so you should answer every question as if it counts.

What practice percentage should I aim for before I book the exam?

There is no single number that converts cleanly to a scaled pass, so aim comfortably above the pass line rather than exactly at it, and look at the trend across many attempts rather than one score. Consistently scoring well clear of the minimum on timed, full-length practice gives you a buffer against a tougher form on exam day.

Can CertGrid tell me my exact scaled exam score in advance?

No, and no honest practice tool can. Vendor scaling algorithms are not public. CertGrid shows your practice percentage and a readiness estimate built from your recent measured attempts (20 or more questions) by domain, which is best used as a trend and a prompt to aim above the pass line, not as a prediction of your scaled result.

Why 700/1000 Doesn't Mean 70% (Scaled Scores Explained Simply)

Here is the most common scoring mistake in IT certification. You see that an exam passes at 700 out of 1000, and you assume you need to answer 70 percent of the questions correctly. It feels obvious: 700 divided by 1000 is 0.7, so 70 percent, right? It is wrong, and quietly believing it can distort how you prepare. This post clears up that one confusion fast and without the drama. For the deeper mechanics behind scaled scoring across every major vendor, there is a longer companion guide linked at the end.

The myth, stated plainly

A passing score of 700 out of 1000 is a scaled score, not a percentage of the questions you answered correctly. The 1000 is not a count of questions, and it is not a percentage ceiling. It is simply the top of a reporting scale the vendor invented. 700 is the point on that scale the vendor has defined as competent enough to pass. The number tells you the bar to clear. It tells you almost nothing about how many individual questions you personally have to get right.

Raw score versus scaled score

The whole confusion comes from mixing up two different numbers that both look like a score. Keeping them separate is most of the battle.

A raw score

Your raw score is the straightforward one: the count of questions you answered correctly, or that count expressed as a percentage of the questions that were scored. If 40 questions counted and you got 32 right, your raw score is 80 percent. This is the number you naturally think in, and it is the number practice tools report. It is not the number the exam reports back to you.

A scaled score

A scaled score takes your raw performance and transforms it onto a fixed reporting range, commonly something like 100 to 1000 or 200 to 800 depending on the vendor. The range is deliberately abstract. It is chosen so results can be reported consistently no matter which version of the exam you happened to sit. The passing mark is a point on that scale, not a slice of your question count. That is the entire reason 700 out of 1000 and 70 percent correct are not the same thing.

Why exams do this at all

Vendors maintain a large pool of questions and assemble many different versions, or forms, of the same exam from that pool. Two candidates rarely see an identical set, and some sets are harder than others purely by chance. If everyone had to hit the same raw percentage, whoever drew the tougher form would be punished for bad luck. To keep things fair, vendors statistically adjust for how difficult each form is, so that a given scaled score always represents the same level of ability regardless of which form you took. That adjustment is exactly why the scaled number, and not your raw percentage, is what gets reported.

Why you cannot compute how many questions you need right

Once you see how scaling works, it follows that you cannot look at 700 out of 1000 and back out a fixed number of questions you must answer correctly. Three separate things break that math, and all three are hidden from you during the exam:

Form difficulty is adjusted for. On an easier form, each correct answer effectively counts for a little less, so you may need more right to clear the bar. On a harder form, fewer correct answers can reach the same passing scaled score. The raw count required floats between versions.
Not every question is worth the same. Some exams weight complex or performance-based items more heavily than simple recall questions, so 'number of questions right' and 'score earned' are not the same measurement.
Unscored questions are mixed in silently. Many exams seed trial questions (variously called pilot or beta items) that do not count toward your result at all. You are almost never told which ones they are, so the number of questions actually being scored is smaller than the number on your screen.

Put those together and the arithmetic simply is not available to you. You do not know your form's difficulty adjustment, you do not know the weighting, and you do not know which questions were unscored. Any 70 percent you compute from the passing number is a guess dressed up as a calculation.

Practical rule: answer every question as if it counts, because you cannot tell which ones do, and never try to reverse-engineer a target number of correct answers from the passing score. The information you would need is not shown to you.

Real cut scores, all scaled, none a raw percentage

It helps to see how varied these scales actually are. Notice that none of them is a plain percentage, and the top of the scale differs from vendor to vendor. The figures below reflect commonly published marks as of 2026; always confirm against the official exam page before your test date.

Microsoft role-based and fundamentals exams (the AZ, AI, SC, and similar series) pass at 700 on a scale that runs to 1000. This is the classic case mistaken for 70 percent. It is a scaled cut point, not a raw percentage of questions correct.
AWS associate-level exams, such as Solutions Architect Associate, pass at 720 on a 100 to 1000 scale. By the same logic, 720 out of 1000 is not '72 percent of questions correct' either.
CompTIA Security+ passes at 750 on a 100 to 900 scale. Because that range tops out at 900 rather than 1000, 750 obviously cannot be read as a simple percentage of questions.

The Security+ example is useful precisely because its scale does not end in 1000. It makes the point impossible to miss: the passing number lives on the vendor's own scale, and trying to read it as a share of your questions gives you a different, meaningless figure every time. The scaled cut is a bar to clear, not a percentage to hit.

What this means for your prep

The good news is that none of this needs to stress you out, and you do not have to decode the scale. You just need a preparation approach that makes the hidden math irrelevant.

The key mental shift is to stop treating any single practice score as a prediction of your exam result. Your practice percentage and the exam's scaled result are measured on different scales, built for different purposes. A 78 percent on a practice set does not convert to 780 on the exam, and it does not mean you would just squeak past a 700 cut. Use your practice percentage as a trend line, not a forecast.

Track your practice percentage over time and watch the direction. A steady climb across many attempts tells you far more than any single score does.
Aim comfortably above the pass line, not exactly at it. Because form difficulty, weighting, and unscored questions are all hidden, scraping a bare pass on a good day is fragile, and a slightly tougher form can push you under.
Practice by domain and shore up your weakest area. Most exams let strength in one area offset another, but a badly weak domain still drags your overall result down.
Move to full-length, timed practice once you are consistently scoring well clear of the minimum, to confirm your pacing and stamina hold up under exam conditions.

This is exactly how CertGrid frames its own numbers. CertGrid shows your practice percentage and a readiness estimate that blends your recent measured attempts (20 or more questions) by domain, so you can see where you are trending strong and where you are still thin. What it deliberately does not do is claim to reproduce a vendor's exact scaled score. That math is not public, and no honest practice tool can recreate it. Treat your CertGrid percentage as a trend to push upward and a signal to aim above the pass line, not as a stand-in for the scaled number you will see on exam day.

CertGrid uses original, independently written questions, not official, real, or leaked exam content and not braindumps, and it does not guarantee a pass. The free plan gives you 20 practice questions a day from a fixed sample pool per exam, in practice mode with basic explanations. Upgrading unlocks the full question bank, timed exam mode, full explanations, and domain analytics.

The bottom line

700 out of 1000 is a finish line drawn on a scale the vendor made up, not 70 percent of questions correct. You cannot compute how many answers you need right, and you do not need to. Learn the material well enough that your practice percentages sit comfortably above the pass line as a trend, and the scale, the weighting, and the silently unscored questions all stop mattering. That is the whole trick, and it is a lot simpler than the myth it replaces.

Practice with realistic exam-style questions