The liquidation cost-per-unit index

The population is the archive’s closed auctions: the 246,154 lots in the archive on 13 October 2026 that genuinely closed, one row per lot, spanning 23 April to 13 October 2026. A closing qualifies for the index when it carries both a recorded final bid and a unit count, and 214,561 of the 246,154 (87.2%) do. The per-unit figure is the lot’s final bid divided by its unit count, both read off the lot’s own closing record, so the denominator is the seller-published unit count for that lot: a 150-piece pallet counts 150, a single sofa counts 1.

Two scope notes before the numbers. A closing bid is the last bid observed before the lot closed, not a confirmed sale price; where nobody bid, it is the asking price the listing came down at. And every figure is bid only: buyer premium, tax and freight arrive on top, and none of them is in the numbers below.

The index publishes July, August, September and the first thirteen days of October 2026, and no earlier month, on purpose. The archive holds no lots that closed in May 2026 at all, and in April and June the closing cohorts carry unit counts on fewer than one lot in five (499 of 2,797 April closings, 901 of 9,232 June closings), so a per-unit median there would describe a small sliver of what actually closed. From July onward, 88% to 93% of each month’s closings qualify, which is the level of coverage an index cell needs to be honest.

The monthly read, measured 13 October 2026:

Measured 13 October 2026. "Qualified closings" is the count carrying both a recorded bid and a unit count greater than zero, the denominator every figure in the row is computed over. The middle half runs from the 25th to the 75th percentile. October is a partial month: thirteen days of closings, not a full calendar month.

The headline move is $7.00 in July down to $5.00 in August, then $5.05 and $5.36. Read beside the last two columns, though, the index is telling two stories at once: the median lot shrank from 20 units in July to a single unit by September, and the median lot bid fell from $190.00 to $35.00 over the same stretch. What closed changed at least as much as what it cost.

A median across every category at once answers "what did the typical closing look like this month", and the typical closing changed shape. July’s archive is weighted toward multi-unit pallets from the marketplace sources that were closing the most lots then; by September the single-unit local-auction stream (small lots, one item each, frequently closing near a $5.00 minimum) made up a far larger share of closings, which pulls the overall per-unit median toward that stream’s price even if no category moved at all.

The category rows below show how much of the move is mix. Home goods held inside a $3.51 to $4.39 band across all four reads; apparel sat between $2.26 and $2.78; furniture slid gradually from $18.82 to $12.50; electronics fell hardest, from $55.94 in July to $17.27 in early October, the one large category with a real downtrend in its own median rather than a composition effect. Where a category barely moved while the headline fell 24%, the fall was the mix doing the talking.

Median closing bid per unit, month by month, measured 13 October 2026. Every cell carries at least 67 qualified closings; the rightmost column is each category’s qualified total across the four reads:

Measured 13 October 2026, October partial. "Not yet categorized" is the cohort whose titles the archive’s classifier has not bucketed; it is the largest row and its median sits on the single-unit stream’s floor prices, so read it as a mix, not a category. The repeated $5.00 medians in the smaller categories are the same artifact: where the median lot is one unit closing at a $5.00 minimum bid, the per-unit median is $5.00 by construction.

Jewelry is the row that argues for reading n before reading a median: its per-unit median quadrupled between August and September on 470 and 1,237 closings, which is what a small cohort does when a few lots of a different shape land in it. The electronics slide, by contrast, sits on 27,994 qualified closings, the third-largest cohort in the index.

• A closing bid is the last bid observed before the lot closed, not a confirmed sale price. Where no bid was placed, it is the asking price the listing came down at.
• Every figure is bid only. Buyer premium, tax and freight come on top, and on a heavy lot the freight can rival the bid: what 23,105 real freight quotes show is the measured version of that line item.
• A unit is whatever the seller counted. One lot’s unit is a single phone and another’s is one piece of a 1,500-piece apparel pallet, so a per-unit median compares unlike shapes across categories and sometimes inside one.
• The index is not condition-adjusted. A month in which more salvage lots close reads lower than a month of shelf pulls, and the archive’s condition grades are the seller’s own labels, so two months with the same categories can still price differently if the grade mix shifted.
• 12.8% of the archive’s closed lots carry no bid or no unit count and appear in no cell, and April, May and June 2026 are absent for the coverage reasons above. The index describes the lots it can measure, not all liquidation.
• A median describes a cohort, not the lot in front of you. The contents, condition and unit count of any specific lot are on that lot’s own record, not in these figures.
• These figures are a snapshot of the archive on 13 October 2026. The archive grows daily and the medians drift with it.

The intent is for this index to recur monthly: each new month’s closings make one more column, recomputed from the archive and published as its own dated snapshot. The figures are computed by hand against the live database, so an update lands as a new dated post here rather than a silently changing ticker, and each edition keeps its own measurement date.

The live version of every row above is the feed itself: the sample feed shows the same cards with no account, the statistics page carries the sitewide closing figures recomputed on its own schedule, and the trends pages track asking-price movement by category on the live side. The free pallet calculator folds a lot’s bid, premium, freight and your own resale prices into a landed cost per unit on your own numbers.