When an MOT records a lower mileage than the year before: what 28.7 million MOT pairs show

About 1 in 570 year-on-year MOT pairs recorded a mileage reading more than 1,000 miles lower than the previous test. Around 7 in 10 of those match a simple recording pattern, such as kilometres entered as miles. That leaves about 1 in 1,830 with no simple pattern in the data.

Data: DVSA anonymised MOT test results for cars and light vans, tests from 1 January 2024 to 30 December 2025 (files published by DVSA on 22 June 2026). Analysis by CarCheckHQ. Not endorsed by DVSA.

Published by CarCheckHQ.

---

Key findings

• We looked at 28,748,864 pairs of consecutive MOT tests on the same vehicle, roughly a year apart, covering 28,239,999 vehicles. The typical car added 5,698 miles between tests.
• 50,613 pairs (17.6 per 10,000, about 1 in 570) had a recorded MOT reading more than 1,000 miles lower than the previous test.

• 34,932 of those (69.0%) fit a simple recording pattern:

  • 30,938 look like a kilometres and miles mix-up
  • 3,561 look like a dropped or added digit
  • 433 look like an old five-digit odometer rolling over past 99,999
  • 15,681 (5.5 per 10,000, about 1 in 1,830) had no simple pattern. We call this the strict measure.
  • Most of those still look like one-off slips. In 13,467 strict cases we could also see the test before. In 8,638 of them (64.1%), the lower reading was at or above the reading from two tests back. So the odd one out was the higher reading in the middle, not the latest one.
  • At least 4,829 pairs had a latest reading below both earlier readings.
  • High-mileage cars are far more affected. Cars with 150,000+ miles at the earlier test had a strict rate of 32.0 per 10,000. Cars under 50,000 miles had 0.2.
  • The Toyota Prius stood out: 31.8 per 10,000 on the strict measure (283 of 89,003 pairs). That is 2.2 times what its mileage profile alone would predict.

Most lower readings are not what they seem

A recorded MOT reading that is lower than the previous one is a reason to ask questions. It is not, on its own, a sign that anyone has done anything wrong. Common innocent explanations include:

• Typing errors by the tester. The tester types the reading into the MOT system by hand. DVSA's own MOT blog says it gets "over 14,000 requests to correct MOT mileage readings" every year, and testers describe catching two keys at once or reading the trip meter by mistake (DVSA Matters of Testing blog, April 2023).
• Kilometres and miles mixed up. Imported cars and some dashboards show kilometres, and readings can be entered in the wrong unit. A kilometre figure entered as miles is about 1.6 times too high, which makes the next correct reading look lower. This is by far the biggest pattern in the data.
• A replaced instrument cluster or speedometer. A replacement unit may start from zero or from the donor car's mileage.
• Odometer rollover on older cars. Some older odometers only show five digits and go back to zero after 99,999.
• A test recorded against the wrong vehicle, for example when a reading is keyed into the wrong record.
• Number plate transfers. In consumer checks that work from the number plate, a private plate moved between cars can make one car's history appear on another. In this dataset DVSA links tests by registration and VIN together, so a plate transfer usually starts a new vehicle record rather than producing a lower reading. It still matters when you look up a car by its plate.

What we measured

Recorded MOT reading lower: the mileage recorded at an MOT test is more than 1,000 miles lower than the reading recorded at the same vehicle's previous MOT test, 240 to 500 days earlier.

We then set aside lower readings that match one of three simple recording patterns. Pairs that fit none of them make up the strict measure. The three patterns are:

The 20,000-mile allowance sits between the 95th percentile (16,169 miles) and 99th percentile (25,415 miles) of annual distance for cars whose reading went up.

How we did it

• Source: DVSA anonymised MOT test results extracts for 2024 and 2025, plus 2023 for the history check.

• Included:

  • class 4 vehicles (cars and light vans)
  • normal tests only (no retests)
  • results of pass, fail or pass after rectification
  • a mileage above zero
  • Pairs: each test paired with the same vehicle's previous qualifying test 240 to 500 days earlier. Each pair counts once.
  • Vehicle identity: DVSA's anonymised vehicle ID, which DVSA bases on registration and VIN. We never saw or used registration numbers.
  • Make and model: DVSA's make and model text, cleaned into model families (for example "C 220 AMG LINE" and "C" both become Mercedes-Benz C-Class). The full mapping is available on request. Vehicles listed as unclassified are left out of the model tables.
  • Minimum sizes: makes, models and postcode areas are shown only with at least 100,000 pairs.
  • Rates: per 10,000 pairs, with 95% Wilson confidence intervals.
  • "Expected given mileage": the rate a group would have if each of its cars had the national rate for its mileage band (under 50k, 50-100k, 100-150k, 150k+). A ratio above 1 means more lower readings than mileage alone explains.
  • History check: for strict pairs, we looked for the same vehicle's test before the earlier test (back to 2023), to see whether the latest reading fits the older history.

Results

1. From all MOT pairs to the strict measure

2. The kilometres and miles signature

For every pair more than 1,000 miles lower, we divided the later reading by the earlier one. If the earlier reading was really kilometres, the result clusters just above 0.62 (1 divided by 1.609). The data shows a sharp spike exactly there.

A second, smaller spike sits at 0.10 to 0.15 (1,766 pairs). That is the lost-digit pattern.

3. Size of the drop (strict measure)

Median drop 10,409 miles (quarters: 3,532 and 45,998).

4. By mileage at the earlier test

Broad measure: more than 1,000 lower, excluding only exact kilometre and lost-digit matches (44,264 pairs, 15.4 per 10k). See "Changes from our first pass".

Read the under-50,000 row with care. At low mileage, most drops of up to about 40% also fit the kilometres and miles pattern, so the strict measure sets them aside by design.

5. By vehicle age at the later test

6. By fuel type

DVSA notes that fuel type is not always recorded accurately.

7. London and the rest of Great Britain

Highest postcode areas with at least 100,000 pairs: E 9.2, HA 7.8, BD 7.8, L 7.4, N 7.3, M 7.3, B 7.3 per 10k.

8. By make (at least 100,000 pairs)

9. By model (at least 100,000 pairs): highest and the most common

Highest strict rates:

Most vans and big estates on this list have high typical mileage, so their ratio against expected is close to 1. Toyota Yaris, Nissan Note and the Ford Transit sit well above what their mileage predicts.

The 10 most common models:

The full model table is in the download.

10. The Toyota Prius

Prius profile compared with all other cars:

Part of the Prius figure comes from high mileage. But even at the same mileage the rate is higher, between 1.5 and 2.5 times the rate for other cars in each band above 50,000 miles. These data cannot tell us why. Instrument cluster replacements and recording errors on very high readings are possibilities. We have not verified either.

11. Does the latest reading fit the car's older history?

When the latest reading sits above the reading from two tests back, the most likely explanation is that the middle reading was recorded too high. That is a slip at the earlier test, not a lower mileage now.

12. How sensitive are the numbers to our choices?

So the honest range is roughly 3.6 to 15.4 per 10,000 pairs depending on definition, with 5.5 as our central figure.

Limitations

• This is recorded data, not the car's true mileage. DVSA describes the reading as "as entered or validated by the tester". We cannot tell what a dashboard actually showed.
• Pattern rules are approximations. The kilometres and miles rule also sets aside some genuine lower readings, especially on low-mileage cars (see table 4). Some unit slips with unusual driving distances will stay in the strict measure. The sensitivity table shows the effect of the main choice.
• Units in the source. DVSA states that from the 2022 data onwards, readings recorded in kilometres are converted to miles. It is not possible to tell from these files which readings were originally kilometres.
• Two years of tests. Pairs need two tests 240 to 500 days apart, so the earlier test is in 2024 or 2025 and the later test is in 2024 or 2025. The history check reaches back to 2023 only.
• Vehicle identity. DVSA's anonymised ID is based on registration and VIN. Errors in either can split one car into two records or join two cars together.
• Make, model and fuel come from DVSA's vehicle data and are sometimes missing or wrong. Some vehicles are "unclassified". DVSA notes that fuel type is not always recorded accurately. Vehicles with an unknown build date carry a placeholder first-use date of 1 January 1971, which affects age bands.
• Location is the postcode area of the test station, not where the owner lives. DVSA merges areas with fewer than five test stations into "XX".
• Association, not cause. Differences between makes or areas may reflect who owns and uses the cars (for example taxis and vans), not the cars themselves.

Changes from our first pass (23 October 2026)

An earlier internal draft used the broad measure (44,264 pairs, 15.4 per 10,000, about 1 in 650). It also quoted much higher rates for London, hybrids and some models, including the Vauxhall Grandland and Crossland. Two problems came up on review:

• That draft only set aside exact kilometre matches. The ratio histogram shows that most kilometres and miles slips include some driving between tests, so they were missed.
• Model names were not cleaned, so large models such as the Mercedes-Benz C-Class were missing or split.

With both fixed:

• The Grandland (broad 51.5, strict 0.8 per 10k) and Crossland (41.7 to 0.6) drop to among the lowest. 96% of their lower readings fit the kilometres and miles pattern.
• London falls from 23.5 to 6.7, and hybrids from 26.5 to 5.5.

This page supersedes those figures.

What a buyer can do

1. Check the MOT history for free on GOV.UK: Check the MOT history of a vehicle. You only need the number plate. Look at the recorded mileage across every test, not just the last one.
2. If one reading is lower, look at the readings either side. If the next reading goes back above the earlier ones, a one-off recording slip is the most likely explanation.
3. Check the units on the dashboard and whether the car is an import.
4. Ask the seller about the reading. Ask whether the instrument cluster or speedometer has ever been replaced, and for paperwork showing mileage, such as service records and old MOT certificates.

5. Wrong reading on your own car? GOV.UK explains how to get the mileage corrected:

   • within 28 days of the test, through the MOT centre
   • after that, by reporting it to DVSA with evidence such as an MOT invoice, emissions printout, service receipt or job card dated within one day of the test

Source, licence and attribution

• Data: "MOT Anonymised Data", published by the Driver and Vehicle Standards Agency (DVSA): https://open.data.dvsa.gov.uk/mot-anonymised/index.html (listed on data.gov.uk: https://www.data.gov.uk/dataset/c63fca52-ae4c-4b75-bab5-8b4735e1a4c9/anonymised-mot-tests-and-results).
• Licence: Contains public sector information licensed under the Open Government Licence v3.0.
• Independence: CarCheckHQ is independent. This analysis is not endorsed by DVSA, and nothing on this page suggests any official status.
• Privacy: the data is anonymised. We did not try to identify any vehicle or person and we publish no registration numbers.
• Method notes: the model-name mapping and run order are available on request.

Download

Summary table (CSV): headline, mileage band, age, fuel, London, make, model and Prius rows. Each row has pairs (n), broad and strict counts (k), rates per 10,000, 95% confidence intervals and ratio to expected where calculated.

For MOT fail rates by age, fuel, mileage band and defect area across 66.3 million initial tests on Great Britain cars in 2024 and 2025, download the open MOT dataset.

More used-car buying guides

For practical advice on checking mileage, MOT history and paperwork before you buy, see our used-car buying guides. The free GOV.UK check above is all you need for MOT history.

---

Published by CarCheckHQ. Last updated: 23 October 2026.