Canadian used vehicles that cut their price take 118% longer to sell

Across 210,345 used Canadian dealer listings that we watched appear and then disappear between 25 June 2026 and 26 September 2026, the median time on market was 19.3 days. A vehicle whose asking price was cut before it left took 33.1 days, against 15.2 for one that was never discounted — 118% longer. 31.7% of vehicles were discounted at all, by a median of $1,096. These are asking prices, not transaction prices: we measure what sellers ask and how long they ask it for. Used vehicles only — new and used are never blended in this series, because they are different markets and the mix would move every figure above.

A discount is a symptom, not a cure

The vehicles that were discounted did not leave faster — they took markedly longer. Read causally in the obvious direction this is backwards, and that is the point: a price cut is what happens to a vehicle that is already not moving. It identifies a car in trouble rather than rescuing one. For anyone pricing inventory, the operative number is not the size of the discount but how early the market told you one was coming.

Where inventory moves, and where it sits

A vehicle in Alberta takes 5 days longer to leave the market than the same class of vehicle in Quebec — 22.1 days against 17.1. Provinces below 200 completed listings are omitted rather than shown with a thin number.

The fastest and slowest vehicles in Canada

Every province

Every model with 150 or more completed used listings

The data, free

The full series behind every figure above — every province, make, model and model year this edition can measure — under CC BY 4.0. Copy it, chart it, publish it, sell work built on it; attribute OmniaAuto and link this page. No account, no email. The licence covers these aggregate files, not the underlying listings corpus, which is licensed under our terms.

Methodology

We observe Canadian dealer inventory daily. A completed spell is a vehicle we watched arrive and watched leave: time on market is the interval between the two. Three rules govern every figure.

Used vehicles only. New and used are never blended. 52% of the listings we watch are new vehicles, and they behave differently on both axes this series measures: they sit longer and they are discounted less often. A blended median is therefore not a compromise between two markets, it is a number describing neither, and it moves with the new/used mix rather than with prices. Every figure in this edition is restricted to condition = Used. (Editions before September 2026 were blended; that is why they were never published.)

Cross-listed vehicles are removed. The same car sits at several locations of a dealer group — 27% of listing rows are duplicates of another row. Each duplicate carries its own first-seen date, so leaving them in does not merely inflate counts, it moves old vehicles into young age bands and corrupts the shape of any duration measure. We collapse to one row per VIN and treat a car as gone only when it has gone from every location; a vehicle moving between stores of one group is not a sale.

Only completed spells are measured. A vehicle still listed has no time on market yet — it has an age, and an unknown future. We do not model it, estimate it or treat it as censored data; we exclude it. Figures are further restricted to spells that also began inside our observation window, so both endpoints are observed facts rather than artifacts of when we started looking.

Medians throughout, never means: asking prices have a long right tail and one $180,000 truck moves an average. A cell with fewer than eight vehicles produces no number at all.

Limitations

• Asking prices, not transaction prices. We measure what sellers ask and how long they ask it for. We do not see what a vehicle sold for, or whether it sold at all — a delisting is a vehicle leaving the market, which is usually a sale and sometimes not.
• A 93-day window. Figures are conditioned on having left within it, so vehicles that sit for many months are under-represented and the true median is higher than the one printed here. This corrects itself as the series lengthens, and we would rather print the bias than model it away.
• No trim-level options control. Two vehicles of the same year, make and model may differ by thousands of dollars of factory options.
• Listing coverage varies by province. Provinces below 200 completed listings are omitted rather than shown thin.
• Each edition re-measures a longer window. An edition is cumulative — every completed spell observed up to its as-of date, not the spells of that month. Consecutive editions therefore overlap heavily and are not independent observations: do not difference two of them and call the result a monthly change.
• Not an appraisal. Nothing here is a valuation for lending, insurance or legal purposes. See terms.

Omnia Index, September 2026 edition. Data as of 2026-09-26. The figures on this page are permanent. This edition is minted once and never overwritten — a later edition gets a new URL, so a citation made against this one keeps resolving to the numbers it quoted. Cite as: OmniaAuto, Omnia Index, September 2026 edition, https://omniaauto.ca/research/2026-09 (accessed date). Licence: CC BY 4.0.