Auctions are 8.7% of what actually sells on eBay in this sample, and 26.5% of those auctions close on a single bid. That is the first thing in 897,135 sold listings that surprised us, and it is the sort of number that is hard to get anywhere else, so we are publishing the whole set.
This page is the eBay sold listings data behind ING Listing Engine, our free Windows listing tool. We keep a store of completed eBay sales to price the items it drafts. It is a working dataset, not a research panel, and the limitations section is as long as it is because of that. Every figure below is computed, stated with its sample size, and reproducible from the scripts named in the method. Nothing is modelled, projected or rounded in a flattering direction.
Method, before the findings
The sold records come from eBay’s own completed-listing pages, harvested through a commercial search API between 16 July and 7 August 2026 against 2,306 seed searches. Each row is one sold eBay item number with its price, postage, condition, seller, sale date and title. Rows priced at zero or with no price were dropped (18,268 of them). Rows in a currency other than US dollars were quarantined by an earlier hygiene pass and never entered the table.
| Field | Value |
|---|---|
| Sold listings collected | 915,403 |
| Used in this study (price above zero) | 897,135 |
| Excluded: price of zero or missing | 18,268 |
| Distinct sellers | 280,320 |
| Distinct seed searches | 2,306 |
| Sale dates covered | 2015-08-26 to 2026-08-07 |
| Median sale date | 2026-07-05 |
| Sales within 90 days of the newest | 789,175 (88.0%) |
| Median sold price | $89.95 |
| 10th / 90th percentile sold price | $13.90 / $540.00 |
| Collected between | 16 July and 7 August 2026 |
| Live listings tracked separately | 6,137 listings, 14,041 observations, 30 July to 6 August 2026 |
The sale dates run back to 2015-08-26, but the mass of the data is recent: the median sale is 2026-07-05 and 88.0% of it falls within 90 days of the newest sale. Treat this as a snapshot of mid-2026, not a longitudinal series.
A second, much smaller dataset sits alongside it: 6,137 live eBay listings observed 14,041 times between 30 July and 6 August 2026, each with its asking price and its running sold count. That is what makes the price-stickiness and zero-sale sections possible, and it is small — every claim built on it says so.
What is in the corpus
The seed searches were chosen for a listing tool, not for a survey, so the mix is what a general-goods reseller touches. The twelve biggest:
| Seed search | Sold listings |
|---|---|
| laptop | 74,129 |
| computer | 68,925 |
| coffee maker | 23,016 |
| ipad | 22,777 |
| macbook | 22,682 |
| turntable | 22,199 |
| chainsaw | 22,152 |
| game console | 22,124 |
| amplifier | 21,686 |
| handbag | 21,661 |
| earbuds | 21,660 |
| backpack | 21,513 |
Alongside those sit narrow industrial and mining searches — Fanuc drive boards, Siemens modules, ASIC miners — because that is what we personally sell. If you are citing this for a category we did not seed, don’t. There is no jewellery, no cars, no property, and trading cards appear only through the Pokémon searches.
Finding 1: nine in ten eBay sales are fixed price
| Format the item sold in | Sold listings | Share |
|---|---|---|
| Buy It Now, no offers | 387,242 | 45.9% |
| Fixed price with Best Offer enabled | 383,399 | 45.4% |
| Auction | 73,038 | 8.7% |
| All fixed price | 770,641 | 91.3% |
Base: 843,679 sold listings whose record carries a format. 53,456 do not and are excluded.
Auction is a rounding error next to Buy It Now. The finding that surprised us more is the split inside fixed price: 45.4% of everything that sold had Best Offer switched on, near enough a coin flip against plain Buy It Now. Half the fixed-price market is negotiating.
One caveat that matters if you want to go further with that number: eBay’s completed-listing page shows an accepted-offer sale at a price we cannot always tell apart from the original asking price. So this dataset can tell you how many sales had offers enabled. It cannot tell you what buyers actually knocked off, and we have not tried to guess.
Finding 2: a quarter of eBay auctions close on one bid
| Final bid count | Auctions | Share |
|---|---|---|
| 1 | 19,325 | 26.5% |
| 2 | 5,742 | 7.9% |
| 3 | 3,589 | 4.9% |
| 4 | 2,903 | 4.0% |
| 5 | 2,518 | 3.4% |
| 6 | 2,416 | 3.3% |
| 7 | 2,233 | 3.1% |
| 8 | 2,057 | 2.8% |
| 9 | 2,036 | 2.8% |
| 10+ | 30,219 | 41.4% |
Base: 73,038 sold auctions. Median 7.0 bids, mean 11.51.
19,325 of 73,038 sold auctions ended with exactly one bid. Nobody competed. The item sold for the number the seller typed into the starting-price box, which makes a one-bid auction functionally a Buy It Now with a week’s delay and a worse price floor.
The distribution is genuinely two-humped: 26.5% at one bid, then a long thin middle, then 41.4% with ten bids or more. Auctions either get a fight or they get nothing. There is not much in between.
Finding 3: auctions usually close lower — but not always, and the gap is noisy
To compare like with like we grouped sales into cells — one seed search, one eBay condition — and only kept cells with at least 20 auction sales and 20 fixed-price sales. That left 225 cells across 113 searches, covering 61,041 auction sales and 627,142 fixed-price ones.
The median cell’s auctions closed at 81.5% of its fixed-price median — 18.5% lower. The middle half of cells ran from 66.1% to 110.0%, and auctions were the cheaper side in 68.0% of cells. Raising the minimum cell size barely moves it: 81.2% at 30 sales a side (186 cells), 82.0% at 50 (136), 84.8% at 100 (88).
| Seed search | Condition | Auction median | Fixed median | Auctions | Fixed | Index |
|---|---|---|---|---|---|---|
| laptop | Pre-Owned | $155.00 | $199.99 | 9,012 | 41,451 | 77.5 |
| macbook | Pre-Owned | $306.17 | $329.94 | 2,869 | 11,211 | 92.8 |
| ipad | Pre-Owned | $135.50 | $189.99 | 2,286 | 11,838 | 71.3 |
| game console | Pre-Owned | $102.50 | $128.18 | 2,235 | 14,265 | 80.0 |
| handbag | Pre-Owned | $91.00 | $57.02 | 2,113 | 12,243 | 159.6 |
| graphics card | Pre-Owned | $151.00 | $129.99 | 1,846 | 13,374 | 116.2 |
| iphone | Pre-Owned | $190.00 | $219.94 | 1,825 | 10,413 | 86.4 |
| laptop | Parts Only | $66.30 | $89.94 | 1,608 | 7,028 | 73.7 |
| pokemon | Pre-Owned | $8.00 | $6.82 | 1,473 | 13,816 | 117.2 |
| lego | Pre-Owned | $24.50 | $22.00 | 1,216 | 7,959 | 111.4 |
| bike | Pre-Owned | $400.00 | $550.00 | 1,154 | 5,501 | 72.7 |
| smartwatch | Pre-Owned | $66.00 | $85.00 | 1,142 | 6,910 | 77.6 |
The twelve cells with the most auction sales, not the twelve that make the point best. Index 100 would mean auctions and fixed-price sales closed at the same median price.
Read the limit before you quote the 18.5%. A seed search plus a condition is a coarse control. Inside “chainsaw / Pre-Owned” the fixed-price listings are mostly chains, bars and carburettors while the auctions are whole saws, and that cell inverts hard. Roughly a third of cells run the other way for reasons like that. The honest version of this finding is: auctions are usually the cheaper way to sell the same thing, by something in the region of a fifth, and the exceptions are not rare.
Finding 4: most sellers waste the title box
eBay gives you 80 characters. The median sold listing used 73, and only 12.0% used all 80. 21.4% of sold listings came in at 60 characters or fewer — a fifth of the market leaving a quarter of the field empty.
| Title length | Share of sold listings | Median price index | Cells |
|---|---|---|---|
| 1-40 | 4.5% | 93.6 | 119 |
| 41-50 | 6.0% | 89.6 | 131 |
| 51-60 | 10.9% | 99.1 | 160 |
| 61-70 | 19.8% | 100.0 | 173 |
| 71-79 | 46.7% | 100.7 | 178 |
| 80 (the cap) | 12.0% | 105.5 | 160 |
Index 100 = the median price of its own cell (one seed search, one condition, fixed-price sales only). 168 cells qualified.
Longer titles go with higher prices, and the effect is small and not clean: the 80-character band sits 105.5 against its own cell median, the 41–50 band sits 89.6. This is a correlation and we cannot make it a cause. A $900 workstation needs more characters to describe than a $9 cable, so item value drives title length as surely as title length could drive price. Anyone telling you that padding your title to 80 characters is worth 5% is over-reading this table, including us.
Finding 5: free postage is not free
48.3% of sold listings shipped free (407,330 of 843,679). Where postage was charged, the median charge was $11.25 and the 90th percentile $47.49.
Comparing the two inside the same cell — same seed search, same condition, fixed price only, 393 cells qualified — free-shipping listings sold at 110.3% of the price of paid-shipping ones, and were the dearer side in 64.6% of cells. That premium is almost exactly the median postage charge on the median item. Sellers are not eating the postage; they are moving it into the price, and buyers are paying it either way. Useful mostly as a reminder that a “free shipping” comp is not comparable to a “+$12 shipping” comp until you add the two together.
Finding 6: eBay prices barely move
2,511 eBay item numbers appear in both halves of this dataset: recorded as a completed sale, and separately observed as a live listing with an asking price. In 98.8% of them the sale came first — these are multi-quantity fixed-price listings that sold one and stayed up. So we can ask a question you cannot normally ask: a median 23 days after making a sale at a known price, what is the seller asking?
| What the seller had done to the price | Listings | Share | Median size of the move |
|---|---|---|---|
| Nothing — same price it had already sold at | 1,689 | 67.3% | — |
| Cut it | 386 | 15.4% | 6.2% lower |
| Raised it | 436 | 17.4% | 8.3% higher |
Base: 2,511 eBay item numbers that appear on both sides of this dataset. Median gap between the recorded sale and the price observation: 23 days (10th–90th percentile 14–61).
67.3% had not touched it. Same number, three weeks and a sale later. Of the ones that moved, cuts and rises were near enough even (15.4% versus 17.4%) and both were small. Within the seven-day observation window itself, only 13.9% of the 1,345 listings we saw more than once changed price at all.
Two limits. These listings are, by definition, the ones that survived long enough to be seen twice — repeat-sale inventory, not one-off items. And a listing that never sold again would not appear here at all. The direction is still hard to argue with: on eBay, the default is to leave the price alone.
Finding 7: half of tracked listings sold nothing in a week
Of 2,562 live fixed-price listings we watched for at least five days (median window 7.0 days), 50.9% sold nothing at all. The 1,259 that did sell moved a median 0.57 units a day, and 18.1% of all tracked listings managed a unit a day or better.
Caveat, and it is a real one: eBay displays high sold counts as “500+”, so 9.1% of these items have a sold count that is a floor rather than a number. Those cases understate movement. The zero-sale share is not affected by it — a listing showing no change from a hard number is a genuine zero.
What we could not measure, and one method that failed outright
A reference page that only lists what worked is not a reference page.
Active asking prices versus sold prices: the comparison does not work
The obvious study to run on this data is the one everybody wants: how far below the asking price do things actually sell? We ran it — median active ask against median sold price, keyword by keyword, 84 keywords with at least 15 active listings and 100 recent sales each — and the result is noise. Across those keywords the sold-to-ask ratio ran from 0.227 at the 10th percentile to 9.717 at the 90th, and 28 keywords produced a sold price more than double the asking price, which is impossible as a market fact.
| Seed search | Median active ask | Median sold price | Sold as % of ask |
|---|---|---|---|
| texa marine service adapter | $64,048.60 | $267.18 | 0.4% |
| flygt compressor controller | $4,077.00 | $280.00 | 6.9% |
| generator | $489.99 | $35.00 | 7.1% |
| genetec card reader lot | $12.48 | $175.98 | 1,409.5% |
| avigilon door controller | $12.87 | $299.50 | 2,327.1% |
| jasminer x4 brick | $1.39 | $78.50 | 5,647.5% |
Three worst and three best of 84 keywords that met the sample minimums. This table is published as evidence that the method fails, not as a finding.
The reason is simple once you see it. A keyword’s active listings and its sold listings are different items. Search “jasminer x4 brick” and the live results are dominated by cheap brackets and cables; the sold results are dominated by the miners. Search “generator” and the live pool skews to whole generators, the sold pool to parts. The ratio is measuring pool composition, not discounting.
We are publishing the failure because the method is everywhere. If a tool or a blog post tells you an item sells for X% below asking based on comparing search results, ask whether it compared the same items. Ours could not, so we are not claiming a number.
Condition premiums: no stable answer
We also tried to price the eBay condition ladder — what Pre-Owned fetches against Brand New for the same thing. It will not hold still. At a minimum of 30 sales per cell, Pre-Owned indexes at 100.0 against Brand New across 349 searches; raise the minimum to 100 and it drops to 83.7 across 54. Open Box moves the opposite way, 105.5 to 126.8. An index that swings that far on a sample-size threshold is not measuring condition. It is measuring the fact that inside one search, the new items and the used items are frequently different products. No condition-premium figure is published here.
Photo counts
We cannot answer whether more photos sell for more. The completed-listing records carry one thumbnail each, not a gallery count, so the question is not answerable from this data at all. We would rather say that than estimate it.
Limitations, stated plainly
- Sample, not census. 897,135 sales out of eBay’s hundreds of millions, reached through 2,306 searches chosen for a listing tool.
- Category coverage is uneven and shown above. Do not project it onto fashion, collectibles, vehicles or anything unseeded.
- US dollars, eBay US. Other currencies were quarantined out.
- A snapshot. 88.0% of sales fall within 90 days of 2026-08-07. There is no seasonal or year-over-year claim here, because the data cannot support one.
- Sold price is what eBay displayed on the completed listing. It does not net out fees, and accepted Best Offer amounts are not reliably distinguishable.
- The live-listing dataset is small — 6,137 listings over eight days. Findings 6 and 7 rest on it and should be read as indicative.
- We sell on eBay and we sell listing software. That is a conflict of interest, which is why the method, the sample sizes and the two failed analyses are all on this page.
Citing this, and the data behind it
Download every table on this page as a CSV (10 KB, CC BY 4.0). One row per figure, each carrying its own sample size, so a number cannot get separated from its n. It includes the analyses that failed, marked as such. It contains aggregates only — no item numbers, no seller names, no titles, no listing URLs; republishing a scraped copy of other people’s listings would be a different decision from the one made here.
Please cite it, and please link back. Suggested form:
ING Listing Engine, eBay Sold Listings Data: 897,135 Real Sales, https://inglisting.com/ebay-sold-listings-data/ (data collected 16 July – 7 August 2026; n = 897,135 sold eBay listings).
If you want a cut that is not here — a category, a price band, a different control — ask and we will run it against the same data and publish the result with its sample size, including if it comes out boring. That offer stands for journalists, other sellers and competitors alike.
Where the data comes from
This is the same store of completed sales that ING Listing Engine prices against when it drafts a listing. The app reads it to suggest a number based on what things sold for rather than what they are listed at — which, given how rarely sellers revisit a price, is the difference that matters most. It is free, it runs on Windows, and there is no paywall: download it here.
Related reading: Terapeak and the alternatives to it covers which research tool to use; how to see sold items on eBay covers where to click now that eBay has put that view behind a sign-in; eBay auction sniping is the other side of Finding 2, from the buyer’s seat; and the 30-day window problem counts this dataset one item at a time, to show how few sales a demand metric actually has to work with.
Questions people ask about this data
What share of eBay sales are auctions?
In this sample of 843,679 sold eBay listings collected between 16 July and 7 August 2026, 8.7% sold at auction and 91.3% sold at a fixed price. Of the fixed-price sales, 45.4% of the total had Best Offer enabled.
How many bids does a typical eBay auction get?
Across 73,038 sold auctions the median was 7.0 bids and the mean 11.51. The distribution is two-humped: 26.5% closed on exactly one bid, while 41.4% attracted ten or more.
Do eBay auctions sell for less than Buy It Now?
Usually, by roughly a fifth, but not reliably. Comparing 225 cells of one seed search and one condition with at least 20 sales each side, the median cell’s auctions closed at 81.5% of its fixed-price median. Auctions were cheaper in 68.0% of cells, so about a third ran the other way, usually because the two pools contained different products.
How long should an eBay title be?
eBay allows 80 characters. The median sold listing in this sample used 73 and only 12.0% used all 80. Longer titles correlate with higher sold prices, but the effect is small and confounded — expensive items simply need more words — so this data cannot tell you that padding a title raises the price.
How often do eBay sellers change their prices?
Rarely. Of 2,511 listings observed live a median 23 days after a recorded sale, 67.3% were still at exactly the price they had already sold at. 15.4% had cut the price and 17.4% had raised it.
Can I cite this data?
Yes. Cite it as: ING Listing Engine, “eBay Sold Listings Data: 897,135 Real Sales”, https://inglisting.com/ebay-sold-listings-data/, with n = 897,135 sold eBay listings collected 16 July to 7 August 2026. If you need a cut that is not published, ask and we will run it and publish the result with its sample size.
Filed under Research & Comparisons · Tagged Sold comps, Dirty data · Browse all guides by topic