A recall is not a repair. The regulator publishes, once a year and only as a PDF, how many of the cars named in each campaign were actually fixed. It publishes the part that failed somewhere else entirely, in a database, and it never puts the two in the same sentence. So the question an owner actually has (if my car has this fault, does it get fixed?) has no published answer.
We read the PDF and we read the part. This is the two of them crossed.
Every rate here is vehicle-weighted, and that matters
One campaign covers two million cars and the next covers twelve. Averaging their percentages gives the twelve-car recall the same voice as the two-million-car one and produces a number that is arithmetically true and completely misleading. So every rate below is repaired divided by recalled, summed across the group. The column next to it is not a percentage at all. It is the count of cars still unrepaired, because that is the thing actually driving around.
By the part that failed
Sorted worst first. A part appears only where the completion list covers at least 8 of its campaigns and at least 100,000 vehicles; below that a rate is a number pretending to be a finding.
| Part | Repaired | Still unrepaired | Vehicles recalled | Campaigns | Median wait to be told |
|---|---|---|---|---|---|
| Equipment | 49.1% | 513,131 | 1,008,627 | 72 | 40 days |
| Seat Belts | 64.1% | 330,955 | 922,065 | 19 | 50 days |
| Back Over Prevention | 75.0% | 807,583 | 3,226,853 | 27 | 60 days |
| Electrical System | 75.5% | 726,450 | 2,970,123 | 115 | 49 days |
| Latches/locks/linkages | 76.4% | 117,759 | 499,257 | 11 | 56 days |
| Fuel System, Gasoline | 77.3% | 192,159 | 848,252 | 23 | 56 days |
| Engine And Engine Cooling | 78.4% | 28,957 | 134,116 | 15 | 50 days |
| Structure | 79.5% | 285,445 | 1,389,699 | 62 | 39 days |
| Air Bags | 79.5% | 269,768 | 1,314,685 | 22 | 54 days |
| Power Train | 81.4% | 172,120 | 923,725 | 47 | 55 days |
| Visibility | 81.7% | 461,108 | 2,525,547 | 22 | 58 days |
| Suspension | 81.9% | 27,826 | 153,814 | 28 | 39 days |
| Trailer Hitches | 84.6% | 108,627 | 703,519 | 9 | 42 days |
| Service Brakes, Hydraulic | 85.0% | 300,335 | 1,998,457 | 24 | 45 days |
| Exterior Lighting | 86.6% | 215,112 | 1,603,196 | 33 | 50 days |
| Steering | 90.0% | 113,702 | 1,140,039 | 33 | 42 days |
The single largest unrepaired population in each of the three worst
- Equipment: 23V420, Ford Motor Company: 420,209 of 691,133 vehicles still unrepaired (39.2% completed).
- Seat Belts: 23V158, Honda (American Honda Motor Co.): 234,176 of 448,613 vehicles still unrepaired (47.8% completed).
- Back Over Prevention: 23V431, Honda (American Honda Motor Co.): 267,338 of 1,209,673 vehicles still unrepaired (77.9% completed).
Do the slow warnings and the poor repairs go together?
Everything else on this site measured how long owners waited to be told. The obvious next question is whether the faults people hear about slowest are also the ones least likely to be repaired. Ranking the 16 parts above by median wait and by completion rate gives a rank correlation of -0.20.
The sign points the way intuition expects (later warning, more cars left unrepaired), but it is far too weak to lean on. At this strength, across 16 groups, the pattern would appear and disappear on the reclassification of one or two parts. We publish the number because we ran the check, and we are telling you not to build anything on it.
Rank correlation rather than a straight line, because neither column is normally distributed and one long-waiting outlier should not be allowed to write the headline.
By what the fix actually is
A software update can reach a car without anyone booking a day off work; a replacement part cannot. Same weighting, same floors.
The obvious expectation is that the more work a remedy asks of an owner, the fewer of them do it. The table says something stranger. The remedy with the worst completion rate is inspection at 50.4 per cent (the one that asks the least) while install reaches 86.3 per cent. Effort is not what decides it. A recall that ends in a new part gives an owner something to show for the trip; a recall that ends in somebody saying “that one is fine” gives them nothing, and 208,886 cars have not made it.
| Remedy | Repaired | Still unrepaired | Vehicles recalled | Campaigns |
|---|---|---|---|---|
| inspect | 50.4% | 208,886 | 420,826 | 35 |
| other | 62.3% | 785,409 | 2,082,772 | 64 |
| repair | 65.1% | 197,503 | 565,975 | 28 |
| replace | 77.2% | 1,871,773 | 8,195,769 | 193 |
| inspect-then-replace | 79.2% | 904,181 | 4,350,746 | 139 |
| software | 84.2% | 981,815 | 6,219,068 | 101 |
| install | 86.3% | 206,478 | 1,505,702 | 72 |
How this page is built, and what it leaves out
The completion figures come from the regulator’s annual list for 2025, read out of the PDF it is published in and cross-checked row by row against the bulk file. The part comes from our own reading of the authority record for each campaign. The wait comes from the same single measurement every other page here uses, so no two pages can disagree about the same campaign.
- 633 campaigns in the completion list.
- 633 of them we also hold a component for: those are the ones every figure above is drawn from.
- 0 had no component we could match, and 0 carried no usable rate or vehicle count. Both are excluded and counted rather than quietly dropped.
- 5,156,457 vehicles across the matched campaigns were still unrepaired when the list was published, out of 23,341,287 recalled.
The list is a slice, not the register. It covers the campaigns the regulator chose to report on that year, not the fifteen thousand we hold, and every rate is as it stood on the day the list was published rather than today. A car repaired last month is still counted here as unrepaired. That is a limit of the source, and it is stated rather than smoothed over.
The completion ledger, campaign by campaign · What failed, and what happened next · Sixty days to tell you · Check your car