AutoNation (AN) · Q3 2026

S1Our Overall Take

This is a read on AutoNation's U.S. vehicle-selling business only, not a rating on the company. AutoNation runs franchised dealerships across the United States (management's own disclosure). Our data follows what happens on its lots: how many cars sell, at what price, and how long they wait for a buyer. This report covers July through September 2026.

New sales moved with the market. New-vehicle sales in our data fell 3.1% from a year ago (5.2% at the stores AutoNation owned through both years), while the whole new market fell 3.0%. The monthly path matched the market almost exactly: −6.9% in July, −3.1% in August and +0.8% in September, against a market at −7.0%, −2.4% and +0.8%. That is what management said to expect: unit sales tracking the markets and brands it serves.

Used sales fell faster than the market, but inside our measurement error. Used sales fell 7.2% (8.1% at the stores AutoNation owned through both years) in a used market down 3.5%. In the first and second quarters of 2026 our used change came in 5.3 and 4.3 points below the change AutoNation reported, and the 3.7-point shortfall this quarter is smaller than either gap (our interpretation). Vehicle revenue in our data was $6.6 billion, down 0.5%.

New cars held price better than the market. New cars sold 2.6% below sticker, against 3.4% a year ago; franchise dealers overall moved only from 3.4% to 3.2% below. Toyota, 32.0% of AutoNation's new sales, sold 0.6% above sticker. This is a pricing reading, not a profit reading.

Used cars came off their asking price as much as a year ago. The average discount per car from the first listed price was 4.6%, against 4.5% a year ago, while franchise dealers overall narrowed from 3.5% to 2.8%. More of AutoNation's used cars were cut (65.1% against 59.1%).

The lots are heavier. 32.0% of new stock had waited more than 90 days at its selling store, against 26.0% a year ago, with Mercedes-Benz new stock up 28.2% and its Market Days Supply at 138% of the franchise benchmark for the brand. Used stock rose 10.8% from a year ago to 30,811 cars, after setting aside 1,370 cars that had been listed more than 180 days at the same store, which we treat as service loaners (Section 12), and used Market Days Supply of 40 sits at the top of its typical range.

Why neutral: new volume tracked the market, revenue was flat, and new pricing firmed. Against that, used pricing held its discount while the market's narrowed, and both lots carry more stock, new in its older tail. Confidence is Medium because our count sits above the company's own and our store-level same-store panel read the second quarter's new change 3.9 points above the company's (Section 12). Profit per car, financing, aftersales, the captive lender and the share count are outside our data; they can move the company's result a full tier in either direction.

S2The Quarter in Numbers

Every share move is below the 0.3-point level we treat as meaningful. The "company" rows compare our second-quarter change with the change in retail units AutoNation reported in its quarterly filing for the same quarter (new −4.0%, used −7.5%). Our count sits well above the company's own unit count, so the level is not comparable; the year-over-year direction is.

S3Inventory and Pricing Risk

New Market Days Supply of 75 sits inside the typical range for franchise dealers (46 to 80 days) and is 2 days lower than a year ago. The new problem is age, not size: the share of new stock listed more than 90 days at its selling store rose 6.0 points. Mercedes-Benz (11,680 cars, 38.8% over 90 days on site), Ford (9,049, 43.5%), Jeep (62.1%) and RAM (57.6%) hold most of the old stock. The asking premium on unsold new stock rose to 10.9% from 8.5%, above our 10% watch line and widening.

Used Market Days Supply of 40 sits at the top of its typical range (27 to 40 days); on the unfiltered lot it would be 41, the highest of AutoNation's 15 quarters in our data. Management said off-lease returns would rise 30% to 40% in the second half, so a larger used lot fits its plan (management's own disclosure); the share listed more than 90 days at its selling store rose only to 6.3% from 5.8%, so the build is mostly fresh stock. Used inventory figures throughout this report exclude cars listed more than 180 days at the same store, an interim rule for service loaners (Section 12).

S4Electric Vehicles, Geography, and the Two Cases

Electric vehicles. New fully electric (battery-electric) sales fell 59.8% to 2,744, after the federal purchase credit ended on September 30, 2025 and pulled sales into the year-ago quarter; the whole market's new battery-electric sales fell 52.6%. Used battery-electric sales rose 15.5% to 3,246, against 11.5% for the market, and sold in 30 days on average. Hybrids of all kinds are 41.0% of AutoNation's new sales.

Geography. Florida, Texas and California, AutoNation's three largest states, fell between 6.2% and 7.5% on new and between 8.1% and 10.9% on used, close to the company-wide figures. Georgia rose (new +21.0%) on the Toyota of Newnan purchase. Colorado, Maryland and Illinois also rose; each gain comes from stores that first appear in our data after July 2025 with no matching disclosure, a Toyota store in White Marsh, Maryland from January 2026 and two stores each in Colorado and Illinois from December 2025, which we treat as changes in how Compass attributes their listings rather than as purchases (our interpretation).

The positive case. New sales tracked the market month by month, and by September were above a year ago. New cars sold closer to sticker while franchise dealers gained less ground, and Toyota, the largest franchise, grew 6.0% and sold above sticker. Used electric sales grew faster than the market.

The negative case. Used sales fell faster than the market, though inside our measurement error. AutoNation held its used discount while franchise dealers cut less. Nearly a third of new stock has waited more than 90 days at its selling store, Mercedes-Benz new supply stands at 138% of the franchise benchmark for the brand, and the used lot sits at the top of its typical range. Our data cannot say what any of this does to profit per car.

01Overall Verdict and Confidence

This is a read on AutoNation's U.S. vehicle-selling business only, not a rating on the company. Finance & Insurance income, aftersales, AutoNation Finance, overhead, share repurchases and the gross profit AutoNation earns on each car are outside our data.

Why neutral: all-store new sales fell 3.1% (same-store −5.2%) in a new market down 3.0%, and the monthly path matched the market. Used sales fell 7.2% (same-store −8.1%) in a used market down 3.5%. Vehicle revenue in our data fell 0.5% to $6.6 billion. New cars sold 2.6% below sticker against 3.4% a year ago. On the other side, used cars sold as far below their first asking price as a year ago (−4.6% against −4.5%) while franchise dealers overall moved to −2.8% from −3.5%; the share of new stock over 90 days at its selling store rose to 32.0%; and used stock rose 10.8% to a Market Days Supply of 40.

How much weight the volume figures carry: our count of vehicles sold has run below AutoNation's reported change in both quarters of 2026 for which the company has reported, by 2.3 and 1.1 points on new and 5.3 and 4.3 points on used. Those gaps point the same way, so the declines shown are more likely too steep than too shallow. The new figure, already in line with the market, carries medium weight. The used shortfall against the market (3.7 points) is smaller than the measured gap, so it carries low weight.

Confidence is Medium, set by two things in the quality-control table: a capture rate well above the company's own unit count and a store-level same-store panel whose second-quarter new reading sat 3.9 points above the company's own (Section 12). Management guided to more stable new and used profit per car, growth in finance and aftersales income and a lower share count in the second half. Those sit outside our data and can move the company's result a full tier in either direction.

02Signal Summary Table

03Detailed Sales Numbers

Why it matters: units sold times average price is a direct shadow of the vehicle revenue AutoNation will report. AutoNation bought and sold stores over the past year, so same-store figures, now built store by store from the seller records carried in Compass, are shown first. The all-store figure is the revenue base, and our measured gap to the company's own reported change sits beside it.

New vehicles

Our count against the company's. In the second quarter our all-store new change was −5.0%, against −4.0% for the retail units AutoNation reported; in the first quarter it was −10.1% against −7.9% (management's own disclosure). Both times our figure was the steeper one, so this quarter's −3.1% should be read as an upper bound on the decline, not a forecast of the company's figure. Same-store new sales fell 5.2%, with the same-store average price up 2.1%.

Month by month, AutoNation's new sales were down 6.9% in July (market −7.0%), down 3.1% in August (market −2.4%) and up 0.8% in September (market +0.8%). The gap to the market stayed within one point in every month.

The average new price rose $788. Higher prices within the same brands added about $1,424, and a mix shift toward cheaper brands (more Toyota, fewer BMW and Chevrolet) took away about $636.

Used vehicles

Used sales fell in every month: −4.4% in July, −11.9% in August and −4.6% in September, against a market down 7.5%, down 2.8% and up 0.5%. In our second-quarter count, the used change ran 4.3 points below the company's reported −7.5%; a gap of that size this quarter would put AutoNation's used decline close to the market's (our interpretation). Same-store used sales fell 8.1%, one point more than all stores, with the same-store average price up 9.3%.

The used-to-new check. AutoNation sold 0.863 used cars for every new car in our data, against 0.901 a year ago and 0.831 last quarter.

What drove the used price rise. The used average transaction price rose $2,398. Brand mix explains only about $103 of it; the rest, about $2,295, came within brands. Within brands, AutoNation sold younger, lower-mileage, more expensive cars: 30.7% were two model years old or newer, against 27.4% a year ago, average mileage fell by 3,665 miles, and cars under $20,000 fell to 33.2% of used sales from 40.0%. The price gain is therefore largely a change in which cars sold, and we do not read it as pricing power. Management said lower-priced used cars were hard to source (management's own disclosure).

Same-store bridge. This quarter's same-store figures compare stores. Using the seller records carried in Compass, each sale is tied to the website of the store that sold it, and a store counts as comparable when AutoNation had been selling through it since before July 1, 2025 and it was still selling in the quarter just ended: 250 stores, carrying 96.4% of AutoNation's new sales and 97.0% of its used sales this quarter. Ten stores whose first AutoNation sales in our data came after that date contributed 2,937 new and 2,041 used units this quarter with no year-ago equivalent. Four of them are disclosed purchases: Toyota of Newnan in Georgia (1,276 units, first seen July 2026) and the three Fremont, California stores (Mercedes-Benz, Audi and Porsche, 608 units, also first seen July 2026) (management's own disclosure). The other six have no matching disclosure: AutoNation Toyota White Marsh in Maryland (853 units, first seen January 2026) and, all first seen on December 23, 2025, AutoNation Ford and Mazda Arapahoe in Colorado and Audi and Mercedes-Benz of Chicago in Illinois. Stores at several dealer groups first appear in our data on that same December date, so we treat those four as a change in how Compass attributes their listings rather than as purchases, and keep them out of the comparable set either way (our interpretation). Twenty-three stores that sold a year ago but not this quarter, led by AutoNation Chrysler Jeep Arapahoe in Colorado (632 units, last seen May 2026), AutoNation Nissan Las Vegas, Jaguar Woodland Hills, AutoNation Lincoln Clearwater and AutoNation Ford Memphis, carried 925 new and 1,002 used units in the year-ago quarter; AutoNation disclosed the sale of two domestic stores in the first half, and the rest are small. Three Porsche stores (Bellevue, Irvine and Orlando) whose volume swung more than fourfold after their websites were re-platformed were left out of both quarters; eleven other re-platformed websites, mostly Porsche stores whose records changed in December 2025 and January 2026, were matched to their stores by seller name. Same-store and all-store differ by 2.1 points on new and 1.0 point on used, consistent with the purchases, so the panel has not leaked. Sales that Compass records under AutoNation's two group-level websites, the AutoNation USA record and autonation.com, carry the zip code of the store that fulfilled them and are counted with that store, so a sale moving between a store's website and a group-level record does not change the same-store total; the few that match no store (970 used units this quarter, 1,912 a year ago) are set aside as one unit in both periods. Store-level same-store approximation built from Compass seller identifiers, not the company's certified same-store comparison; stores acquired part-way through the year-ago quarter can add partial-period noise.

The group-level records are taking share of the credit. Inside the comparable stores, used sales recorded on their own websites fell 23.3% while used sales recorded under the two group-level websites rose 96.1%, taking their share of comparable used sales to 28.3% from 13.6%; on new, store-website sales fell 6.0% and group-level sales rose 15.5%. AutoNation USA, the company's standalone used-car chain, is growing, but not at that rate, and a shift this large between two records that cover the same stores fits a change in which listing Compass credits with the sale better than a change in where cars are sold (our interpretation). It does not move the same-store total, because both records carry the store's zip code, and it is the reason the store-level panel is used in place of the zip-code panel of earlier drafts.

By state

The three largest states, which AutoNation says generate about 64% of its new retail units, fell broadly in line with the company total, so the decline is not concentrated in one market (our interpretation). Georgia's gain matches the Newnan purchase. Illinois and Maryland also rose from small bases, on the stores first seen after July 2025 described above; Washington fell 20.3% on new after a Spokane store left the network.

04Top Brands Sold

Why it matters: Toyota is about a third of AutoNation's new sales, and the luxury brands carry the highest prices.

New

Toyota is 32.0% of new units, sold in about a month, and sold above sticker. Toyota, Honda, Lexus and Subaru grew; Chevrolet, Jeep and GMC fell by double digits. The import brands sold closer to sticker than a year ago (Toyota +0.6% against −1.5%, Honda +0.1% against −2.3%), and so did the German luxury brands (Mercedes-Benz +1.1% against −0.2%, BMW +0.7% against 0.0%). Ford sold further below sticker (−9.7% against −7.3%).

Against the company's own segments. AutoNation reports new units in three segments: domestic, import and premium luxury. Grouping our brands into those segments as closely as we can, our second-quarter changes were −22.9%, +6.2% and −6.8%, against the company's −11.7%, +1.0% and −3.9%. The order matches but our magnitudes run two to three times the company's, so the segment split is shown as an internal observation and does not feed the positive or negative case. This quarter the same grouping reads domestic −14.8%, import +4.0% and premium luxury −4.3%.

Used

Luxury used held up best (Mercedes-Benz −0.9%, BMW −2.2%, Audi +11.6%), while the mainstream brands Toyota, Ford and Chevrolet fell 11% to 12%. Luxury makes are 31.2% of used sales, against 30.0% a year ago. Used Mercedes-Benz cars sold after 37 days at their selling store but 73 days on the market, the widest gap of any brand; Section 7a takes up what that difference means.

05What's Sitting on the Lots (Inventory)

Why it matters: a car that sits costs the dealer interest every day, and its eventual selling price tends to drift down. A growing pile of older stock is tomorrow's clearance problem.

New: the same amount of stock, older. New units fell 2.9% from a year ago, but the share listed more than 90 days at its selling store rose to 32.0% from 26.0%, and from 31.2% in June. The aging runs through both luxury stock (34.7% over 90 days on site, against 27.6% a year ago) and mainstream stock (30.5% against 25.3%). Mercedes-Benz new stock grew to 11,680 cars from 9,112 (+28.2%) and 38.8% of it had waited more than 90 days at its store, against 28.8%. Ford new stock fell 29.4% to 9,049, but 43.5% of what remains is over 90 days on site, against 33.2%. Luxury brands are now 36.7% of new stock, against 31.3% a year ago. The sport-utility share of new stock is 55.6%, pickups 19.4%, and 56.8% is priced above $50,000.

Used: a bigger lot, mostly fresh. Active used units rose 10.8% from a year ago (3,011 cars) and 5.5% from June (1,595 cars), after setting aside the 1,370 cars listed more than 180 days at the same store. The share listed more than 90 days at its selling store rose to 6.3% from 5.8% a year ago and 5.4% in June, so the build is mostly recent stock; on the unfiltered lot the over-90 share would be 10.3%. Luxury cars are 32.2% of used stock; their over-90-day share on site is 9.0%, against 5.0% for mainstream stock.

06Price Position: Sticker, Asking Premium and Discounts

Why it matters: these measures show where AutoNation prices its cars against the sticker, against what its cars sell for, and against its own first asking price. None of them is a profit measure: they say nothing about what AutoNation paid for a car or earned on it.

New, against sticker. New cars sold 2.6% below the manufacturer's suggested retail price, against 3.4% a year ago and 2.5% last quarter. Franchise dealers overall moved from 3.4% to 3.2% below, so AutoNation gained 0.8 points while franchise dealers gained 0.2. The gain came from the import and German luxury brands; Ford, Nissan, Hyundai and GMC sold further below sticker.

Asking premium on unsold stock. This compares the average list price of cars still for sale at September 30 with the average price of cars sold during the quarter. The two groups are different cars, so the trend matters more than the level.

The new premium is above our 10% watch line and widening: unsold new cars are listed further above what sold cars fetched than a year ago, which fits the aging new stock. The used premium narrowed because sold prices rose faster than the lot's list prices, and it sits below the watch line once the long-listed cars are set aside (10.3% on the unfiltered lot).

Discount from first listed price. This compares each sold car's final price with the first price MarketCheck saw on its listing.

More of AutoNation's used cars came off their first asking price than a year ago (65.1% against 59.1%), though each cut was a little shallower, so the average discount per car was almost unchanged. Franchise dealers' average discount narrowed by 0.8 points, from −3.5% to −2.8%. Against the second quarter, AutoNation's discount deepened by 1.7 points and franchise dealers' by 0.4 points. The monthly path was steady: −4.5% in July, −4.7% in August and −4.4% in September. On new cars, the average discount from first ask was −2.2%, against −2.4% a year ago and −1.8% for franchise dealers.

6dListing Price Position (used)

Listing Price Position compares each used car's opening asking price with MarketCheck Price, MarketCheck's model estimate of the car's market value. A ratio of 1.00 means listed at our estimate of market value; 0.95 means 5% below.

A year ago AutoNation opened its used listings 0.011 below the franchise median; this quarter it opened them 0.002 above it, with fewer cars listed well below market value and more listed well above. Its cars still sold within 30 days on site a little more often than franchise dealers' cars (57.4% against 55.8%), and that rate fell by about as much as theirs. Higher opening prices followed by more frequent cuts after listing is consistent with the discount table above. This section is descriptive and does not feed the signal.

6eSelling Margin Pressure (used)

Selling Margin Pressure measures markdowns beyond the usual on cars that took longer than usual to sell. For each price band and vehicle age, AutoNation's own previous 12 months set the usual time to sell and the usual markdown on cars that sold on time. A car counts when it took longer than usual and was marked down by more than usual; the excess, in dollars, is averaged over all used cars sold.

AutoNation's Selling Margin Pressure was $1,321 per used car sold, up $120 from a year ago, against a rise of $3 for franchise dealers. Against the second quarter it rose from $932, as the market's did ($805 to $1,050). This describes selling behavior only. What AutoNation paid for its cars and spent reconditioning them are not in our data, so the measure is not a gross profit figure and does not feed the signal.

07Days to Sell and Market Days Supply

7aDays on Site and Days on Market (how long each car takes)

Two clocks run on every car in our data. Days on site counts only the days at the current selling store or website and restarts whenever the car moves; it is the clock a store manages its own pricing and markdowns against, and it is the aging measure this report leads with in every section. Days on market counts from the first day the car was advertised anywhere and keeps counting if the car moves to another store or another website; it is the listing's full exposure to the market, and it is analyzed here and only here. Both are shown for cars sold during the quarter and for cars still unsold on September 30.

On the days-on-site clock, new cars that sold took about as long as a year ago (64 days against 65), and used cars one day longer (32 against 31), though the typical used car took three days longer (median 23 against 20). The unsold new lot is much older: its average age on site rose 12 days and its median 7 days. The unsold used lot, after setting aside the cars listed more than 180 days at the same store (Section 12), aged two days on both measures. The days-on-market rows tell the same story with larger numbers: cars sold took a day less on new and two days more on used, and the unsold stock aged 12 days on new and 2 on used over the year.

What the gap between the two clocks tells us. Days on site runs below days on market by about six days on new cars sold and seven on used, and 32.0% of the new and 34.6% of the used cars AutoNation sold this quarter were younger on their selling site than on the market (34.8% and 28.6% a year ago). Those shares are the second-highest of the eight dealer groups we cover, behind Lithia, whose gap sits in its Driveway record, and well above Carvana, which sells through one website and shows almost none. Store-to-store movement does not explain it: only 0.1% of the new cars and 0.2% of the used cars on AutoNation's lots at September 30 had changed selling store within their listing, and only 0.4% of new and 0.9% of used sales were followed by the same car reappearing at another AutoNation store within 60 days (1.2% reappeared at the same store, and 1.6% of used sales at a dealer outside the group). Nor is the gap concentrated in the two group-level records the way Lithia's is in Driveway: among cars sold, those recorded under the AutoNation USA record or autonation.com were younger on site than on the market less often (27.0%) than those recorded on store websites (37.7%), though among unsold used stock the order reverses (47.6% against 37.5%). Of the used cars on the lots at September 30, 22.1% were already younger on site than on the market on the first day we observed them, and a further 20.3% had their site clock restart without a visible change of store.

Our reading, which the Compass data cannot fully confirm, is that the gap has two sources (our interpretation). The first is cars that reached the store after time advertised elsewhere: off-lease and auction cars that a previous seller had listed, and the share of used sales that is about to be sourced from lease returns makes this the part to watch. The second is the restart of the site clock when a car's record switches between a store's own website and one of the group-level records, which the shift in credited sales in Section 3 suggests is happening more often than a year ago; a car that has not moved can then look young on site. What Compass cannot tell us is how much of each. Days on site is the clock this report leads with because it is the one AutoNation's stores price and mark down against; days on market, which neither source affects, is the listing's full market exposure and is analyzed in this section only.

7bMarket Days Supply (how long the whole lot would last)

Market Days Supply divides the cars on hand at September 30 by the average number sold per day in the 45 days ending that day (August 17 to September 30). Each period shown uses the 45 days to its own quarter-end.

New Market Days Supply of 75 sits inside the typical range for franchise dealers (46 to 80 days) and is 2 days lower than a year ago. Used Market Days Supply of 40 sits at the top of its typical range (27 to 40 days). It rose because active used inventory grew 11% from a year ago while used vehicles sold in the 45 days to September 30 fell 6%. Used stock here excludes cars listed more than 180 days at the same store (Section 12); on the unfiltered lot the figure would be 41, just above the typical range and the highest of the 15 quarters in our data for AutoNation.

Coherence check. Sold days on site divided by Market Days Supply is 86% for new and 78% for used, against the ranges in which the two normally sit, 62% to 100% for new and 76% to 114% for used, derived on sold days on site over complete quarters since 2023; both sit inside their ranges. The used ratio is taken on the unfiltered lot (Market Days Supply 41 rather than 40), because the service-loaner exclusion does not enter benchmark comparisons. On the days-on-market clock the ratios are 94% for new and 95% for used against ranges of 69% to 108% and 82% to 124%, inside their ranges, so the inventory and sales figures agree with each other on both clocks.

Against franchise dealers, brand by brand. Each figure shows AutoNation's Market Days Supply for a brand as a percentage of the per-make benchmark across franchise dealers for the same brand, at the same date and on the same 45-day window. 100% means AutoNation is in line with franchise dealers of the same makes; below 100% means it carries less supply than they do.

On new cars, AutoNation carries less supply than franchise dealers in Toyota, Ford, Honda and RAM, is level in BMW, and carries more in Mercedes-Benz, where it stands at 138% of the franchise benchmark. On used, it carries less than franchise dealers in every major brand, even after this quarter's build; the used figures exclude the same long-listed cars as the rest of the used inventory, while the franchise benchmark does not, and the exclusion is what takes BMW and Audi from about 90% of the benchmark to about 80%.

08Electric Vehicle Sales

Why it matters: the federal electric-vehicle purchase credit ended on September 30, 2025, so buyers pulled purchases into the year-ago quarter. Year-over-year comparisons are against an inflated base.

New battery-electric sales fell faster than the market's 52.6%, and AutoNation's share of the market's new battery-electric sales slipped from 2.7% to 2.3%. BMW (729) and Toyota (390) led its new electric sales. Used battery-electric sales rose 15.5% against 11.5% for the market, led by Tesla (1,207), and sold more than 11 days faster than a year ago.

Hybrids. Hybrids of all kinds were 41.0% of AutoNation's new sales in our data, and full hybrids alone 25.1%. Full hybrids sold in 38 days on average, against 76 days for gasoline cars.

09The Positive Case

New sales moved with the market. AutoNation's monthly gap to the new market stayed within one point in July, August and September, and by September new sales were up 0.8% from a year ago. That is the path management guided to.

Our count is more likely too low than too high. In both reported quarters of 2026 our year-over-year change ran below AutoNation's own, by up to 2.3 points on new and 5.3 points on used. On used, a gap of that size would bring the quarter close to the market (our interpretation).

New cars held price better than the market. AutoNation sold 2.6% below sticker against 3.4% a year ago, a 0.8-point gain against 0.2 points for franchise dealers. Toyota, 32.0% of new sales, grew 6.0%, sold 0.6% above sticker and carries less supply than franchise dealers of the brand.

The used side is rebuilding stock ahead of the off-lease returns management expects, and still carries less used supply than franchise dealers in every major brand. Used electric sales grew faster than the market.

10The Negative Case

Used pricing softened against the market. AutoNation's average used discount from first ask stayed at −4.6% while franchise dealers' narrowed to −2.8% from −3.5%. More cars were cut (65.1% against 59.1%), and the discount deepened by 1.7 points from the second quarter. Our data cannot say what this does to used profit per car.

New stock is aging. 32.0% of new stock had waited more than 90 days at its selling store, against 26.0% a year ago, and the asking premium on unsold new cars widened to 10.9%. Mercedes-Benz new stock rose 28.2% and its Market Days Supply is 138% of the franchise benchmark for the brand; Jeep and RAM have more than half their new stock over 90 days on site.

The used lot grew while sales fell. Used Market Days Supply reached 40, the top of its typical range (41 on the unfiltered lot, AutoNation's highest in 15 quarters), though the share of used stock over 90 days at its selling store is only 6.3%. If the off-lease build arrives faster than sales, this stock will need to be cleared.

Used sales fell faster than the market (−7.2% against −3.5%, and −8.1% same-store), though the shortfall is inside our measured gap to the company.

11ADid Management Get It Right Last Quarter?

Sources: AutoNation Q2 2026 earnings call transcript (July 31, 2026) and Q2 2026 Form 10-Q (quarterly report). "Fact" claims describe the prior quarter; "Guidance" claims describe this one.

The track-the-market claim is the one our data supports most directly on new. On used, our count runs too negative against the company to confirm or refute it. The used build fits the off-lease guidance, but our data cannot see where the cars came from.

11BWhat to Watch on the Earnings Call

1. Used volume against the market. Our capture shows used sales down 7.2% (same-store −8.1%) in a market down 3.5%, and our used count has run 4 to 5 points below the company's in both earlier quarters this year. What were AutoNation's same-store used units, and did used volume track its markets as guided?
2. New stock age and Mercedes-Benz supply. 32.0% of new stock had waited more than 90 days at its selling store, and Mercedes-Benz new stock rose 28.2%, to a Market Days Supply 138% of the franchise benchmark for the brand. How much aged new stock is on the books, and what happened to new gross profit per car, which our data cannot measure?
3. The used build and used discounts. Used stock rose 10.8% while the average used discount from first ask stayed at −4.6%, against −2.8% for franchise dealers. How much of the build is off-lease returns, and how is used gross profit per car holding?
4. Same-store units. Our all-store new change was −3.1% and our store-level same-store figure −5.2%, with the Newnan and Fremont purchases outside the panel; on used, −7.2% and −8.1%. What were the company's same-store new and used changes?
5. Portfolio changes. Which stores were bought or sold in the third quarter, and does AutoNation have a Toyota store in White Marsh, Maryland, which first appears in our data in January 2026 with no matching disclosure?
6. AutoNation USA and the credit of sale. Used sales that Compass records under the AutoNation USA and autonation.com records rose 96.1% inside the comparable stores while sales recorded on store websites fell 23.3% (Section 3). How much of AutoNation's used volume now closes through AutoNation USA, and are cars moved between the franchise stores and the used-car stores before they sell?

11CPost-Print Accuracy Summary (AutoNation Q3 2026 Results, October 30, 2026)

What this section is: This section was added after AutoNation reported third-quarter 2026 results on October 30, 2026. Everything above it is the report as written, unchanged.

Scorecard: Of 29 vehicle-transaction calls that could be tested against what AutoNation reported, 19 were confirmed, 5 were partly right and 5 were missed. Another 5 could not be tested: two the report itself said our data could not see (margin pressure and new-vehicle gross profit per unit), two for which AutoNation gave no figure (the White Marsh Toyota store and how much used volume went through AutoNation USA), and one that was a record of last quarter's days' supply rather than a call. The Neutral verdict held. Retail units fell 4.3%, with new units down 3.4% (same-store 5.3%) and used units down 5.1% (same-store 6.3%), and revenue from new and retail used vehicles fell 1.4%. Adjusted earnings per share were up 2% at $5.11, driven by service and parts, AutoNation Finance and share repurchases, all of which the report named as outside our data.

What we got right: The volume read. Our new-vehicle decline of 3.1% sat beside AutoNation's 3.4%, and our same-store figure of −5.2% beside its −5.3%, the closest this report has come on both measures. We called used sales a shortfall against the market but weighted it low because our used count had run steeper than AutoNation's in earlier quarters, and that is how it landed: the company's used units fell 5.1% against our 7.2%. The heavier lots were the report's clearest call, and AutoNation confirmed it: new days' supply rose to 52 from 47 and used to 43 from 37, management said new inventory was higher than it wanted and would be reduced over the coming months, and part of the used stock was held too long. The fall in battery-electric sales we measured (59.8%) matched the company's 60%. Finally, the report said the result would turn on items outside our data, and it did: earnings per share were higher on service and parts, AutoNation Finance and share repurchases while vehicle profit fell.

What we missed and why: Both pricing reads pointed the wrong way. New cars sold closer to sticker than a year ago, and revenue per new car did rise 1.0%, but new gross profit per unit fell 4.6%; a narrower discount to sticker was not a sign of stronger profit per car, and the same read missed last quarter. Part of the move toward sticker across spring 2026 also reflects franchise dealers adding mandatory fees to advertised prices after Federal Trade Commission enforcement, and the size of that part is not estimated. On used cars, the steady discount from first asking price did not mean weaker retail profit: used gross profit per unit rose 2.4%, and the cost of older stock appeared in wholesale, which our data does not cover. The report also said our new-vehicle decline was more likely too steep than too shallow and could be read as an upper bound; AutoNation's new units fell further than ours, because our count of its new sales held up better over the year than in the two earlier quarters. Our new-vehicle supply measure, which counts cars in transit, moved down while the company's new days' supply rose, so one inventory reading in the report pointed the wrong way even as the aging figures pointed the right way. Finally, our data put Toyota at 32.0% of AutoNation's new sales; the company's brand mix puts it at 23.3%, so the Toyota read was right in direction but overstated in size.

Bottom line: Read this report for the shape of AutoNation's vehicle quarter, with new sales in line with the market, used sales weaker and both lots heavier, which held up, and for its new and same-store new unit changes, which landed close. Read its pricing signals as descriptions of advertised and selling prices only, not of profit per car, and read its unit changes as what our data captured, not as limits on the company's own figures.

12Methodology and Glossary

Where the data comes from

Primary data source: MarketCheck Compass parquet files were the primary data source: North American new and used retail sales and the daily inventory panel, filtered to AutoNation's dealer-group ticker. Sales cover July–September 2026 (current quarter), April–June 2026 (prior quarter) and July–September 2025 (year-ago quarter). The inventory snapshot is September 30, 2026, the true quarter-end; the data run through at least October 9, 2026. Every AutoNation sales record in our data is in the United States.

Compass totals, Q3 2026: new 84,745 units and $4.4B; used 73,175 units and $2.2B; combined $6.6B.

Revenue basis. Revenue is the sum of captured selling prices and is not units times average price. The share of used sales without a captured price was 0.8% in Q3 2025, 1.4% in Q2 2026 and 0.6% in Q3 2026; for new it was 0.7%, 4.1% and 0.4%.

Basis line. Inventory basis: Compass panel for unit counts, list prices and Market Days Supply. Aging basis: days on site leads every aging figure in this report (Sections 3, 4, 5, 6d, 10 and 11); days on market is shown and analyzed in Section 7a only. Days on site: the Compass site counter, re-anchored at each observed reset. Sold and active days on market: measured from the opportunity start date. The over-90-day shares count cars with more than 90 days on site at the snapshot; on the days-on-market clock the same shares are 34.6% new and 9.8% used at September 30 (27.6% and 8.7% a year ago), and the brand and class shares are higher by a similar margin. Market Days Supply: cars on hand at quarter-end divided by average daily sales in the 45 calendar days to quarter-end (ds_denominator = t45); baselines installed 2026-10-19. AutoNation's own figures use a one-month selling rate and count stock on the lot; ours uses the 45 days to quarter-end and counts every advertised unit, and the two are compared directly with that difference stated. New-vehicle Market Days Supply counts every advertised unit, including units advertised before they reach the dealership; company-reported figures count stock on the lot and are not directly comparable.

Market share denominator. Market share is computed against MarketCheck Compass-captured North American retail transactions for the same period, not against total US registrations or company-reported deliveries. Coverage of new-vehicle retail runs approximately 80–95% of the actual market and varies by brand, so these figures are a consistent relative measure across periods, not an absolute share of the true market.

What this signal is built on, and what it does not adjust for

Volume, revenue and share in this report are what MarketCheck captured, not the company's results. The direction stated is our reading of the transaction evidence, not a forecast of AutoNation's reported numbers.

Coverage record. AutoNation reports retail new and used units in its quarterly filing, so a coverage rate (our count divided by the company's) can be computed. It sits well above 100% in every quarter we checked: our count includes sales the company does not report as retail units.

Waiver disclosure 1 (coverage). Our rules treat a new-vehicle coverage rate above 100% as a stop condition, to be investigated before publication. The analyst waived that stop for this report on October 12, 2026, on condition of this disclosure. The error runs in one direction: our count sits above the company's. The excess shrank from a year earlier, slightly on new (from 141.6% to 140.0% in the second quarter) and by more on used (from 119.6% to 114.1%), which is why our year-over-year changes have read more negative than the company's, by about 1 to 2 points on new and 4 to 5 points on used. The gap should be assumed to recur; its size should not be assumed. The new-car volume reading, already level with the market, carries medium weight; the used reading, whose shortfall against the market sits inside the measured gap, carries low weight. The pricing and inventory lenses are per-car or within-lot measures that a capture change does not move.

Same-store panel. Stores are identified by the Compass seller website; the dealer identifier was not usable across years because its numbering changed in January 2026. A store is comparable when its first AutoNation sale in our data is on or before July 1, 2025 and it sold in the quarter just ended; stores first seen after that date are treated as acquired, stores with no sales this quarter as departed, and both are left out. Stores whose volume moved more than fourfold between the two quarters, with at least 150 units in one of them, are left out of both. Websites that were re-platformed or rebranded during the year were matched to their stores by seller name, or by store zip code and leading make where the name changed. Sales recorded under the two group-level websites (the AutoNation USA record and autonation.com) carry the fulfilling store's zip code and take that store's status; the few that match no store form one unit counted in both quarters. On the second quarter this panel read −0.8% for new and −8.7% for used, against AutoNation's reported same-store −4.7% and −8.0%: used within a point, new 3.9 points too high. The zip-code panel used in earlier drafts read −2.5% and −8.6% for the same quarter. The store-level panel is adopted because it compares what the company compares, stores, and because it is unaffected by a sale moving between a store's website and a group-level record, which the zip-code panel could not separate from a store entering or leaving; its new reading for the second quarter is the reason confidence stays at Medium. The company's own same-store comparison is built store by store on its own ledger; our quarter-level rule approximates it.

What our quality checks found

Inventory snapshot date. September 30, 2026, the true quarter-end; Compass inventory quarter-end: 71,550 new and 32,181 used, of which 30,811 used after the service-loaner exclusion.

Days-on-site basis for sold within 30 days. The Section 6d speed figure counts used listings started at their selling store between the first day of the quarter and September 3, so that every listing could be followed for 30 days before the data edge. A listing is sold within 30 days when its sale record carries 30 or fewer days on site, and not sold when the car reached a 31st day on site in the inventory panel; listings that disappeared unsold before day 31 are in neither count. Because the site clock restarts when a car's record switches between a store's website and a group-level record, a car that switched inside its first 30 days starts a new site listing, which can flatter this measure slightly relative to dealers that sell through one website. The franchise comparator is built the same way over all franchise dealers. This cohort is not the listing-day cohort used for the price-position figures above it.

Service-loaner exclusion (interim rule). Dealers keep some cars for service customers to drive and advertise them while they do, and those cars should not count as stock for sale. Pending a company-wide definition of a service loaner, every used inventory figure in this report (units, list prices, aging, asking premium, Market Days Supply and the brand-level supply comparison) excludes cars that had been listed more than 180 days at the same store on the snapshot date, read from the days-on-site counter. That removed 1,370 used cars at September 30, 1,012 at June 30 and 742 a year ago. The cars set aside look more like a loaner fleet than aged stock: at September 30 they averaged 310 days on market and a $44,279 list price, 58.2% had under 10,000 miles, and the group averaged 39,042 miles. The used aging figures are correspondingly gentler than on the unfiltered lot (active days on site 33 against 44; over-90-day share on site 6.3% against 10.3%; Market Days Supply 40 against 41). New inventory is not filtered: the 9,636 new units over 180 days on site average 355 miles and 265 days on market and are aged stock, not loaners. Sold figures are unchanged.

State notes. No top-three state fell by more than 15%. Washington fell 20.3% on new after a Spokane store left the network; same-rooftop continuity and MarketCheck coverage stability there have not been independently checked.

Listing Price Position. Listing Price Position compares each used car's opening asking price with MarketCheck Price, MarketCheck's model estimate of the car's market value, at the listing's first recorded status change. It describes where the dealer set the price, not what the car sold for. The year-ago quarter lacks MarketCheck Price for its first four days. This quarter's position (+0.002) sits above AutoNation's observed range (−0.011 to −0.007).

Directional read for grading. AutoNation's used listing position against franchise dealers moved by +0.013, from −0.011 to +0.002. Under the grading rule, a group listing further above its comparator than a year ago is logged as a read for higher used gross profit per car. It is recorded for the post-earnings accuracy review only and does not feed the signal.

Lenses. Selling Margin Pressure (Section 6e) is descriptive and does not feed the signal. Brand equity is applied to Toyota, which is about a third of new sales. The company-segment reconciliation in Section 4 did not tie out, so the brand-mix lens is not in the signal table. Overhead is outside our data; AutoNation's adjusted overhead rose to 68.2% of gross profit in the second quarter from 66.2%, and a flat vehicle revenue line neither helps nor hurts that ratio on its own, so we take no view.

What we cannot see

Outside MarketCheck's scope for AutoNation: Finance & Insurance income; parts, service and collision work; AutoNation Finance; wholesale and auction sales; gross profit per vehicle; floor-plan interest, overhead and share repurchases; and fleet sales not advertised at retail.

Sources of figures. Every figure in this report comes from MarketCheck data unless the text names another source, such as a company filing or management's comments on its earnings call. Only figures from those other sources carry a source label.

Plain-English Glossary

Prepared by MarketCheck Research using the Anthropic Claude LLM. This report describes North American retail vehicle transaction activity only and is not investment advice.

We welcome your feedback. To offer comments, suggestions for future coverage, or to inquire about licensing this report or the MarketCheck data behind it, including vehicle-level sales, inventory and pricing data, contact Bryant Sheehy at bryant.sheehy@marketcheck.com.