What is a normal practice exam score?
Across 192 completed practice exam attempts from 129 candidates recorded by FigigExams between 4 March and 31 August 2026, the median score was 50.6% counting every attempt - and 68.9% counting only attempts worked at 20 seconds per question or slower. The 18-point gap exists because roughly a third of "completed" attempts were never actually sat: they were opened and submitted, at a median pace of 1.3 seconds per question. Any average practice score that does not exclude those is understating reality by nearly twenty points. Measured August 2026.
In This Report
The click-through floor
- Click-through floor
- The share of nominally completed practice exam attempts that were answered faster than 20 seconds per question - fast enough that the candidate cannot have read the items. These are submissions, not sittings, and they must be excluded before any score statistic is meaningful.
A quarter of all completed attempts in this sample ran at under 3.6 seconds per question. That is not a low score produced by a struggling candidate. It is someone opening a paper, clicking through it, and submitting - to see the format, to unlock the results screen, or because they changed their mind about sitting it.
Those attempts are indistinguishable from genuine failures in any dataset that only records a score. They are trivially distinguishable once you record elapsed time. Here is what excluding them does to the headline number:
| Filter | Attempts | Mean score | Median score | Dropped |
|---|---|---|---|---|
| No filter Misleading | 192 | 45.8% | 50.6% | - |
| At least 5s per question | 142 | 58.2% | 65.4% | 50 |
| At least 10s per question | 135 | 60.8% | 67.0% | 57 |
| At least 15s per question | 132 | 62.0% | 67.0% | 60 |
| At least 20s per question Our threshold | 127 | 62.9% | 68.9% | 65 |
| At least 30s per question | 98 | 65.7% | 68.9% | 94 |
Note that the median stops moving after 20 seconds per question - it is 68.9% at both the 20-second and 30-second thresholds. That stability is the reason we set the line at 20: past it, we are no longer removing click-throughs, we are removing real candidates who happen to work quickly.
Where the cliff is: score by pace band
The threshold is not arbitrary. Score against pace is not a smooth gradient - there is a sharp discontinuity between 15 and 20 seconds per question, which is roughly the minimum time needed to read a multiple-choice item and four options.
| Pace | Attempts | Mean score | Median score | Passed | |
|---|---|---|---|---|---|
| Under 5s per question | 50 | 10.4% | 1.0% | 6% | |
| 5 to 10s | 7 | 8.3% | 5.0% | 0% | |
| 10 to 15s | 3 | 9.1% | 11.7% | 0% | |
| 15 to 20s The cliff | 5 | 39.2% | 32.2% | 0% | |
| 20 to 30s | 29 | 53.2% | 44.4% | 48% | |
| 30 to 45s | 26 | 67.4% | 65.9% | 38% | |
| 45 to 60s | 29 | 75.1% | 80.0% | 66% | |
| 60 to 90s | 30 | 58.0% | 67.0% | 33% | |
| 90 to 120s | 4 | 70.2% | 72.7% | 50% | |
| Over 120s | 9 | 54.6% | 48.5% | 33% |
Below 15 seconds per question the mean score sits around 10% - which is close to what random guessing produces on four-option items, and is exactly what you would expect from a paper nobody read. Cross 20 seconds and the mean jumps to 53.2%, and the pass rate goes from 6% to 48%. The jump is 43 points across a five-second boundary. That is a behavioural threshold, not a skill gradient.
The distribution, before and after
The click-through floor does not merely shift the average. It changes the apparent shape of the distribution, which is what makes it dangerous to reason from.
| Score band | All attempts (n=192) | Engaged only (n=127) | ||
|---|---|---|---|---|
| 0-19% | 66 (34%) | 13 (10%) | ||
| 20-39% | 21 (11%) | 17 (13%) | ||
| 40-49% | 8 (4%) | 7 (6%) | ||
| 50-59% | 11 (6%) | 9 (7%) | ||
| 60-69% | 22 (11%) | 20 (16%) | ||
| 70-79% | 14 (7%) | 14 (11%) | ||
| 80-89% | 23 (12%) | 23 (18%) | ||
| 90-100% | 27 (14%) | 24 (19%) |
Counting every attempt, the distribution looks violently bimodal - a huge spike at the bottom, a thin middle, and a second cluster at the top. That shape invites a tempting and wrong story about candidates either "getting it" or not.
Counting only engaged attempts, the bimodality mostly disappears. What is left is a broad, left-skewed distribution with its mass between 60% and 100% and a modest tail of genuine low scores. The bottom band falls from 34% of attempts to 10%. The apparent bimodality was an artifact of measurement, not a fact about candidates.
The bottom cluster is not failure
66 attempts scored under 20%. 53 of them - 80% - were answered at under 20 seconds per question, at a median pace of 1.3 seconds per question.
Only 13 of the 66 low scores came from a candidate who actually worked the paper. The near-zero cluster that dominates an unfiltered practice exam distribution is overwhelmingly abandonment, not failure.
1.3 seconds per question is worth sitting with. On a 100-question paper that is a little over two minutes from first question to submission. Nobody is reading anything. The score that comes out is a measurement of nothing, and it will sit in a dataset looking exactly like the score of someone who studied for a month and struggled.
This matters beyond our own numbers. Any published statistic about practice exam performance - a vendor's average, a forum poll, a "students score X on their first attempt" claim - is inflated by this artifact unless the author explicitly excluded it. Most do not, because most only have the score.
How long a practice exam actually takes
Among the 127 engaged attempts, the median full-length practice paper took 79 minutes across a median of 101 questions - about 50 seconds per question.
| Exam | Engaged attempts | Median score | Median time |
|---|---|---|---|
| CCNA 200-301 | 67 | 64% | 102 min |
| CompTIA Security+ SY0-701 | 36 | 80% | 72 min |
| CompTIA A+ Core 1 (220-1201) | 9 | 62% | 64 min |
| CCNP ENCOR 350-401 | 7 | 50% | 60 min |
| CompTIA A+ Core 2 (220-1202) | 6 | 93% | 67 min |
| CompTIA Network+ N10-009 | 2 | 90% | 62 min |
Only CCNA and Security+ carry enough engaged attempts to read much into. The four rows below them are reported for completeness, not as findings.
The CCNA figure is the interesting one: 102 minutes median on a 103-question paper, against the roughly 120 minutes Cisco allows for the real exam. Candidates who work our CCNA paper properly are using close to the full clock, which means pacing is a live constraint on that exam rather than a theoretical one.
Slower is not better past a point
Within the engaged group only - click-throughs already removed - the relationship between pace and score is an inverted U, not a straight line:
| Engaged candidates, split into thirds by pace | Median pace | Mean score | |
|---|---|---|---|
| Fastest third | 26s per question | 55.9% | |
| Middle third | - | 74.3% | |
| Slowest third | 70s per question | 58.5% |
The slowest third scores no better than the fastest third. The obvious reading is the right one: past roughly a minute per question, taking longer is a symptom of not knowing the material rather than a technique for recovering marks. Deliberation helps when you are choosing between two candidate answers. It does not help when you are staring at four options you do not recognise.
The practical consequence is that "slow down and read the question more carefully" is bad advice for a candidate already at 70 seconds per question. Their problem is content, not pacing. Which domain to fix is a separate question, and one we measured in where candidates lose marks.
Your free CCNA exam code
See where your own score and pace land against this distribution. A full-length CCNA 200-301 practice exam, timed, with an Explanation and an Exam Tip on every question. Create an account on the next screen, the code auto-applies.
CCNA-EXAM-FULL
Take a timed CCNA paper
SECPLUS-EXAM-FULL
Take a timed Security+ paper
Already have an account? Log in to start
Method
- Source
- Completed practice exam attempt records written at submission, each carrying a start timestamp, an end timestamp, a recorded pause duration, a question count and a percentage score.
- Window
- 4 March 2026 to 31 August 2026 inclusive - just under six months. Attempt volume by month: March 6, April 12, May 21, June 12, July 40, August 101.
- Population
- 362 raw attempt records reduced to 192 from 129 distinct candidates. Excluded before analysis: 39 belonging to FigigExams staff, test-pattern or deleted-authentication accounts; 80 still in progress; 24 capped preview attempts; 27 with unusable timing or no recorded score.
- Seconds per question
- Elapsed wall-clock time from start to submission, net of recorded pause duration, divided by the number of questions on the paper. Attempts longer than 600 minutes are treated as abandoned sessions and excluded.
- The 20-second threshold
- Chosen because it is where the score cliff sits, and validated by the median stabilising at 68.9% at both the 20s and 30s thresholds. Every figure in this report is given at multiple thresholds so a reader can disagree with the choice and still use the data.
- Certifications covered
- CCNA 200-301, CompTIA Security+ SY0-701, CompTIA A+ Core 1 and Core 2, CompTIA Network+ N10-009, CCNP ENCOR 350-401. Unlike our domain-level report, this analysis needs no blueprint mapping, so it covers every certification with attempts in the window.
Limits of this data
What this report cannot tell you
- These are practice results, not exam results. We do not observe testing-centre outcomes, so nothing here is a pass rate for a real exam or a prediction of one. "Passed" in the pace table means passed our paper against its configured pass mark.
- The 20-second threshold is a judgment call. It is defensible - the score cliff is there, and the median stops moving past it - but it is still a line we drew. Every table gives figures at several thresholds so you can move it.
- Elapsed time is not attention. A candidate who starts a paper, walks away for twenty minutes and comes back registers as slow rather than interrupted. Recorded pauses are netted out, but only pauses the candidate actually triggered. This inflates the slowest band and is the most likely explanation for part of the inverted U.
- Four of the six certifications are thinly sampled. Only CCNA (67) and Security+ (36) carry enough engaged attempts to read into. A+ Core 1, A+ Core 2, Network+ and CCNP ENCOR are reported for completeness.
- Self-selection. Everyone here chose to open a practice exam. Candidates who never started one are invisible, which biases the engaged figures upward relative to everyone studying for these exams.
- No rate-shaped claims about our funnel. This report describes attempts, not people. It does not say what share of candidates abandon, because a candidate can produce several click-throughs and one real sitting.
https://figigexams.com/learn/research/what-is-a-normal-practice-exam-score/
Sample: 192 completed practice exam attempts, 129 candidates, 4 March to 31 August 2026. Median score 50.6% unfiltered, 68.9% excluding attempts under 20 seconds per question.
Frequently Asked Questions
What is a normal score on a practice exam?
Measured across 192 completed attempts from 129 candidates between 4 March and 31 August 2026, the median was 50.6% counting every attempt and 68.9% counting only attempts worked at 20 seconds per question or slower. The second figure is the one that describes a candidate who actually sat the paper. Roughly a third of "completed" attempts are click-throughs.
How long should a full-length practice exam take?
Among candidates who genuinely worked the paper, the median was 79 minutes for a median 101 questions - about 50 seconds per question. On CCNA 200-301 specifically the median was 102 minutes on a 103-question paper, close to the full time Cisco allows for the real exam.
Does taking longer produce a better score?
Only up to a point. Among engaged candidates, the fastest third averaged 55.9%, the middle third 74.3%, and the slowest third 58.5%. It is an inverted U. Past roughly a minute per question, going slower is a symptom of not knowing the material rather than a way of recovering marks - so "slow down and read more carefully" is bad advice for someone already at 70 seconds per question.
Why do so many practice scores cluster near zero?
Because most of them are not failures. Of 66 attempts scoring under 20%, 53 (80%) ran at under 20 seconds per question, median pace 1.3 seconds. That is a paper opened and submitted, not sat. Only 13 of the 66 came from a candidate who worked the paper. The apparent bimodality of an unfiltered practice score distribution is a measurement artifact.
Is my score good enough to book the real exam?
This report deliberately does not answer that - it describes a population, not a booking decision, and a threshold that ignores which domains you missed is a bad threshold. For that, see am I ready for the CCNA, am I ready for the Security+ or am I ready for the CISSP, which carry the honest score thresholds and what they assume.
How do I compare two of my own attempts fairly?
Carefully - two papers of different length or difficulty are not directly comparable, and a score rise after re-sitting a paper you have already seen mixes learning with recall. Comparing practice exam attempts: when improvement is real works through what a fair comparison needs.