2012 Subaru Impreza: what breaks, and at what mileage

What owners report breaking

Owners have reported 44 named failure modes on the 2012 Subaru Impreza, and the three most-reported carry 30% of everything filed, which is an unusually even spread for one model year.

Airbag warning light or occupant sensor fault leads with 56 reports, about 1.0 per 1,000 vehicles and with the middle half of reports between 20k to 75,000 miles.

Other engine problem follows with 29 reports, usually reported between 10k to 70,000 miles.

Excessive oil consumption follows with 17 reports, usually reported between 6k to 40,000 miles.

By system, the reports fall mostly into airbag (74), engine (54) and transmission (27).

Severity, and what it costs

Across the named modes, 13 mention a crash, 6 a fire and 7 an injury. Those are the words owners used in their own complaints, and NHTSA does not verify them, but they are the closest thing in this data to a measure of what a failure actually did.

The average severity weight across every report on this year is 4.11, where a stalled engine or a failed airbag weighs many times more than a rattling trim panel, and a figure this high means the reports are dominated by failures with safety consequences rather than annoyances. The weighting is set out in full on the methodology page.

What the top failures cost to put right, as editorial estimates for an independent US shop rather than figures from NHTSA: airbag warning light or occupant sensor fault at $300–$1,600, other engine problem at $300–$4,000, excessive oil consumption at $900–$6,500 and other electrical problem at $250–$2,000. Dealer labour, a seized bolt or a part that is no longer made can push any of these past the top of its range.

26% of the reports on this year sit in the engine, transmission, drivetrain or cooling systems — the four where a repair bill is usually four figures.

How often, against how many were built

Scaled by an estimate of how many of these cars exist, the 2012 Subaru Impreza draws about 6.2 per 1,000 vehicles, which puts it in the 87th percentile of complaint rates among 2012 vehicles we can rate.

That denominator is the weakest number on the page and it is fair to say so plainly. Rate per 1,000 uses a fleet-size estimate derived from recall paperwork for this exact model year. It is a lower bound.

The working figure is 54,313 vehicles. If the true number is twice that, every rate on this page is half what it says; the ranking against other 2012 vehicles, which is what the risk score uses, is far more robust than the rate itself.

Recalls on the 2012 model year

NHTSA has issued 6 recall campaigns covering the 2012 Subaru Impreza, affecting an estimated 133,017 vehicles. A recall is a different kind of evidence from a complaint: it is the regulator and the manufacturer agreeing that something has to be fixed, and the fix is free for as long as the car exists.

The campaigns cover exterior lighting:brake lights:switch, engine and engine cooling:engine, electrical system and air bags:sensor:occupant classification. Whether the work was actually done on the car in front of you is a question only the VIN can answer, and the manufacturer or NHTSA will answer it free.

One caution about reading these together: a recall usually raises the complaint count for the same part, because owners who read about it go and file. The mileages stay informative; the counts get noisy.

How 2012 compares with the other years

High risk for a 2012 — 89th riskiest of the 271 2012 vehicles we can rate. Because the ranking is against the same model year only, age cancels out of it.

Across the 27 model years of this car that carry enough reports to score, 2015 comes out best at 65 and 2021 worst at 84, with 2012 itself at 73.

If the model year is negotiable, 2015, 2003, 2016, 2002, 2005, 2014, 1996, 2010 and 2013 score lower than 2012 by a margin worth acting on. Same body, same engines, different luck — usually a supplier change or a mid-cycle fix, which the complaint data can show but never explain.

Within the 2012–2019 generation specifically, 2015 is the strongest year at 65 and 2012 the weakest at 73.

Where to spend the inspection money

The mileages below are where the first quarter of reports for each mode had already been filed, which makes them the point at which a fault is worth looking for rather than the point at which it happens: excessive oil consumption from about 6k, seat belt or retractor failure from about 8k, other engine problem from about 10k and airbag warning light or occupant sensor fault from about 20k.

On a 2012 Subaru Impreza with more miles than that on it, the useful question to put to a seller or a mechanic is whether excessive oil consumption has been dealt with, and if so with what parts and at whose expense. A car that has already had the repair is in a different position from one that has not.

Treat it as where to point the inspection, not as a prediction. A well-maintained example can sail past every mileage on this page, and a neglected one can fail well before the first of them.

Where these figures stop being useful

337 owner reports; fleet size estimated from this year's own recall paperwork. Every count on this page is therefore a floor: it is the number of owners who went to the trouble of filing with a federal agency, which is a small and self-selected fraction of the owners who had the fault.

Of those 337 reports, 262 carried an odometer reading, so the mileages here describe 78% of the sample rather than all of it. Owners also round: readings pile up on multiples of ten thousand, which is why this site publishes "about 90,000 miles" and not "91,438".

The failure modes are assigned by software reading the owner's own words, which is accurate in bulk and wrong in individual cases. A mode with a handful of reports may contain a misfiled one.

And nothing here predicts what one car will do. These are other people's complaints, filed years apart, on cars maintained by strangers. Read the page as a map of where this model year has previously gone wrong, then have the specific car in front of you inspected.