UK driving test retake patterns, why 49% pass first and 4% take 6 or more tries

Section 1, The shape of the attempt-number curve

Plot UK car practical pass rate by attempt number and the shape is not the steep descent that the conventional wisdom suggests. The common story, repeated across driving-school marketing pages and the occasional newspaper feature, is that pass rates collapse after the first or second attempt: candidates who fail twice are said to enter a downward spiral of confidence and lost form. The DVSA data, which is the only source that publishes the actual per-attempt breakdown, tells a different story.

The rate is essentially flat from attempt 1 to attempt 4. The first-attempt rate (48.92%) and the second-attempt rate (49.51%) are within 0.59 percentage points of each other. The third attempt (49.06%) sits in the same band. The fourth attempt drops only fractionally. The fifth attempt is the first visible step down. The 6 or more attempt cohort is the only one that meaningfully diverges, sitting at 42.51%.

This pattern is not unique to 2024-25. Across the five-year window (2020-21 to 2024-25) the first-attempt rate has ranged from 48.08% to 51.04%, and the 6+ rate has ranged from 38.78% to 42.51%. Every year shows the same plateau-then-collapse shape. The collapse is a property of the candidate pool at attempt 6, not of the year or the policy environment.

Section 2, Selection bias and why the second attempt is the peak

The most counter-intuitive line on the chart is that attempt 2 passes at a slightly higher rate than attempt 1. By a tiny margin, but consistently across years. The first-attempt rate is 48.92%; the second is 49.51%; the difference is 0.59 percentage points. Across the five-year window, the second-attempt rate has been at or above the first-attempt rate every single year.

The mechanism is selection bias on the candidate pool. Three kinds of candidate sit a first test:

• The well-prepared. Booked when their instructor said they were ready, have done their mock tests, have a stable manoeuvre repertoire. They mostly pass on attempt 1.
• The nearly-ready. Booked under family or financial pressure to "just go and see", have one or two weak spots they have not ironed out (typically observation at junctions or response to traffic lights, the two recurring top faults in DRT121F). They mostly fail on attempt 1 with one serious fault, then book a second test after a focused recovery block of lessons.
• The genuinely unprepared. Booked because the wait was long and they wanted to lock in a slot, did not actually accumulate the practice hours by test day. They mostly fail attempt 1 badly, and a significant share never book a second test at all (or book it and cancel, then drift out of the test pool).

The first group leaves the retake funnel after attempt 1. The third group leaves the funnel without ever booking attempt 2. The second group is the dominant cohort at attempt 2. They have already failed once and they know exactly what they failed on. Their pass rate on attempt 2 is therefore enriched above the heterogeneous first-attempt rate. This is the same selection-bias mechanism that gives the COVID-2020-21 year an anomalously high pass rate in the volume-trends analysis: the candidates who sat tests during the shutdown had had three to nine extra months of practice, which lifted their group rate.

Section 3, The retake plateau holds through attempt 4

From attempt 2 through attempt 4 the pass rate barely moves. Attempt 2 is 49.51%; attempt 3 is 49.06%; attempt 4 is 48.00%. All three sit within a single percentage point of each other.

This is a non-obvious finding. The naive prediction is that pass probability should fall with each retake because the candidate pool is increasingly enriched with candidates who have persistent problems. The data does not support this until attempt 5 and 6+. Two mechanisms explain the plateau:

• The fail-and-recover cycle is short. Most candidates who fail at attempts 2 to 4 have a small number of specific faults the instructor can target. The DRT121F top-faults release shows that 80% of serious faults sit in a short list: observation at junctions, mirrors when changing direction, moving off safely, junction-turning-right, response to traffic lights. These are addressable in a focused block of lessons. The 2-to-4 attempt candidates are working through this list, succeeding once they have closed the gaps.
• The seriously-unready cohort exits the dataset. Candidates who fail badly often drop out of the test funnel entirely after attempt 2 or 3. They do not contribute to attempt 4's pool. The pool at attempt 4 is therefore a relatively select group of candidates who are still committed and still close to passing.

The plateau breaks at attempt 5, which drops to 46.76%, and breaks more sharply at attempt 6+, which sits at 42.51%. By that point the remaining pool is dominated by candidates whose persistent faults are not being resolved between tests. There is no evidence in the DVSA data that the examiners mark these candidates more harshly; the test-day standards are the same. What changes is the input.

Section 4, How much of total UK test volume is retakes?

The pass-rate chart above gives the success-rate story. The share-of-volume chart gives the population story. About half of all car tests in 2024-25 were first attempts (48.9%); the rest were retakes of various depths.

The shape is a sharp decay. About 48.9% of all sittings are first attempts. By attempt 3 that has fallen to 12.4%. By attempt 5 it is down to 3.4%. The 6+ bucket holds 4.4% of all sittings, which is a meaningful share given that it represents the long tail of candidates accumulating multiple retakes each.

Read together with the pass-rate chart, the volume chart says something useful for system planners. DVSA's test slots are not being clogged by a small group of long-tail candidates. The 6+ bucket is only 4.4% of volume; the dominant share is going to first and second-attempt sitters. If the practical wait time is to come down, the lever that matters is not "how to discourage long-tail retakes" but "how to add throughput": more examiners, longer days, weekend tests. The retake pattern itself is not the bottleneck.

Section 5, What we know about the 6+ attempt cohort

About 4.4% of all test sittings in 2024-25 were a 6th or later attempt by the candidate, which translates to 81,775 tests in absolute terms. The pass rate for this group was 42.51%, the lowest of any attempt cohort. The drop from the population-average of 48.67% is 6.16 percentage points.

DVSA does not publish granularity beyond "6 or more", so the candidate who is on their 12th attempt is statistically indistinguishable from the candidate on their 6th. Anecdotally, and from the (limited) freedom-of-information data DVSA has released, the deep tail has a handful of recognisable characteristics:

• Older average age. The published age cross-tab (see comparison block below) shows that the over-50 and over-60 buckets are heavily over-represented in the 6+ column. Younger candidates tend to either pass within their first 3 attempts or exit the funnel.
• Geographic clustering. The DRT122A per-centre data, which is the only public source for centre-level pass rates, shows that some hard centres (typically dense urban centres with complex traffic conditions) produce a long-tail cohort by virtue of being where local candidates have to test. Pre-COVID academic work by the Transport Research Laboratory found that candidates testing at the bottom-decile centres were 1.7x more likely to need 4 or more attempts than candidates at median centres.
• Persistent fault patterns. The DRT121F top-faults release lists the same five recurring serious faults year after year. A candidate who fails attempt 1 with "junctions, observation" and fails attempt 4 with the same fault has not had that fault closed out by their instructor. This is observable in the fault-distribution of the long-tail cohort even though individual candidate data is not published.
• Anxiety-driven recurrent failures. The DVSA has acknowledged in transport-committee evidence that a small share of long-tail candidates pass mock tests cleanly and fail real tests on observable nervous-system responses (hesitation at roundabouts, freezing at junctions). DVSA does not have a separate "anxiety pathway" for these candidates; they sit the standard test and fail repeatedly.

What the data does not let us say: how many unique candidates are in the 6+ pool. DRT121D measures test sittings, not unique candidates. A candidate on their 8th attempt contributes one test to the 6+ bucket; a candidate on their 12th contributes one too. The number of distinct people behind the 81,775 6+ sittings figure is not published.

Section 6, Age cohort comparison, first attempt vs 6+ attempt

The DRT121D release publishes single-age rows for ages 18 to 49, plus banded rows for "17 and under", "50 to 59", and "60 and over". The table below shows the first-attempt pass rate and the 6+ pass rate for each band in 2024-25. The columns are aggregated across Male and Female (Unknown gender excluded). Bands with fewer than 100 tests at either end are dropped from the comparison.

The two ends of the candidate journey diverge sharply. The youngest cohort passes first time at a rate well above the national first-attempt mean. The oldest cohort, by the time they are on their sixth or later attempt, is passing at a rate well below the national 6+ mean. The DVSA data does not allow us to attribute this fully to age (it could be cohort-specific preparation patterns, geography, vehicle type, or simply that the older cohort started learning later in life and is closer to the test in real-time terms). What it does allow us to say is that the retake pattern is not uniform across the population.

Section 7, What this means for booking your next test

Three implications stand out from the pattern, in priority order for a learner who has just failed a test:

• The second test is statistically a fine outcome. Pass rate at attempt 2 (49.51%) is not lower than attempt 1; it is fractionally higher. If you failed your first test with one or two serious faults that you and your instructor have identified, the data says you have roughly the same chance of passing your second test as a first-timer.
• The plateau is real to about attempt 4. Pass rate stays in the 47% to 50% band through attempt 4 in 2024-25. The mental story of "I have failed three times, I am doomed" is not supported by the population data. The candidates passing attempt 4 are passing at a rate roughly the same as first-timers.
• By attempt 5 to 6 the recovery loop is not working. Pass rate at attempt 5 dips to 46.76%, then 6+ to 42.51%. If you find yourself booking a fifth or sixth test with the same approach that has not worked four times, the data says continued same-method testing has a lower expected return than a change of approach: a different instructor, a structured re-assessment with mock tests at a different centre, or a gap of several months for the anxious-driver cohort.

None of this is medical or coaching advice; it is what the DVSA-published frequencies imply about expected pass probability conditional on attempt number. Individual candidates vary substantially; the figures here describe the average, not any single person.

Section 8, Methodology and limitations

Data source. DVSA DRT121D, "Car driving test number of attempts by financial year, gender, age, conducted, passed and pass rates: Great Britain". The release covers every car practical test sat in Great Britain from April 2011 onwards. We use the five most recent financial years (2020-21 to 2024-25). All figures are licensed under the Open Government Licence v3.0. See the methodology page for the broader aggregation rules used across this site.

Headline figures. The hero stat-grid, the attempt-level bar chart, and the share-of-volume chart all use DVSA's published Total/Total row at every attempt level: every test sat in the year, every gender (Male, Female, Unknown), every age. No further filtering. This is the same set of figures DVSA itself uses in its annual statistical commentary.

Per-age aggregation. The age cohort comparison block aggregates Male and Female rows across the published age bands. Unknown gender is excluded (about 0.05% of test volume). DVSA's per-age rows summed across Male+Female do not always equal the published Total/Total row exactly (typically within 1 to 5% at any single attempt level), because DVSA's underlying disclosure-control rounding is applied at the per-row level not at the aggregate level. This is the same kind of rounding noise you see in the age-cohort analysis on /research/pass-rate-by-age. The pass-rate figures quoted in the cohort table are computed from the per-age aggregation; treat them as accurate to within roughly one percentage point.

Attempt-number definition. DRT121D records the attempt number of the test sitting itself, not the unique-candidate attempt count. A candidate who books a test then cancels it does not register a sitting at all. A candidate who books a test then does not arrive registers a "no-show" in DRT122B (cancellations) but is not counted as an attempt in DRT121D. So the "1st attempt" row is the candidate's 1st test sitting, which is what the candidate themselves would call their first attempt.

Average attempts per pass. The 2.05 figure is total tests divided by total passes for 2024-25. This is the volume-weighted mean, which slightly over-estimates the mean for candidates who do eventually pass (because never-pass candidates contribute to the numerator but not the denominator). DVSA does not publish a unique-candidate-passing-eventually denominator that would let us compute the corrected figure directly. The honest interpretation is that 2.05 is an upper bound on the mean attempts-per-successful-pass, and the true figure is somewhere between 1.9 and 2.05.

What this analysis cannot do. The DRT121D release does not carry lesson-count data, instructor data, cancellation history, or geographic data per candidate. We cannot test hypotheses about lesson hours per attempt, examiner stringency by attempt number, or post-fail recovery patterns at centre level. The DVSA per-centre release (DRT122A) carries centre-level totals but no attempt-number breakdown; the two releases cannot be cross-tabulated. The DRT121D also stops at "6 or more"; we cannot say anything about the candidate on their 12th test except that they are in that bucket.

Cite this page: passrates.uk research/retake-patterns v1.0 (2026). Underlying data: DVSA DRT121D (OGL v3.0), financial years 2020-21 to 2024-25.

Frequently asked questions

Related on PassRates.uk

• How we calculate pass rates
• Seasonality of UK pass rates
• Why men pass more often than women
• London vs UK pass rate
• Pass rate by age cohort
• Test volume trends 2019 to 2025
• Best first-time pass-rate centres
• What happens if you fail your test