Pricing government surplus clothing: 3,772 sold lots vs. Poshmark prices

Clothing was one of our worst-priced categories, and outside data helped. We ran all 3,772 clothing, shoe and bag lots that sold on GovDeals since May through SoldStack, an API of cleaned Poshmark sold prices. On lots where the listing names the items, its value came in below the winning bid on 9% of lots. Our own estimates did that on 34%. That's why we're adding SoldStack as a data source in our quest to find the real arbitrage in everything government auctions sell.

Why clothing is hard to price

Clothing is a small corner of government surplus: about 2% of sold GovDeals lots (3,772 of roughly 180,000 we tracked through closing, May 22 to October 23, 2026). That still comes to around 750 lots a month, mostly uniforms, university athletic gear, lost-and-found and seized property.

It's also hard to price. There are no UPCs, sizes and condition swing the price, fakes exist, and lots range from one Coach purse to a 500-pound pallet of "assorted clothes." When we tested our AI valuations against real outcomes, bulk clothing came in at about 0.48× the winning bid, the worst category we measured. eBay comps don't help much either, because clothing resells on Poshmark and Mercari.

The test

SoldStack builds its prices from Poshmark sold listings. It matches on brand and item, drops wrong models, bundles, kids' sizes and replica wording, trims outliers, and splits prices by condition. For each of our 3,772 sold lots, it read the listing into items, priced each item from those sales, and added them up. We then compared that value, and our own estimate, with what the winning bidder paid.

The results, by lot type

The lot type mattered more than anything else. Itemized lots name their items; counted bulk lots are many of one thing ("61 coveralls"); uncounted lots are pallets and boxes sold by weight. "Value ÷ bid" is the median SoldStack value divided by the winning bid.

Another 55 lots were jewelry, armor or collectibles and were left out. For uncounted lots SoldStack gave per-item prices only, since nobody knows how many items are in the box.

Two harder cuts on the itemized lots:

• On lots our own feed had actually scored (341 lots): SoldStack was below the winning bid on 9%, we were on 34%.
• On lots with 3 or more bidders (420 lots), where competition makes the winning bid a firmer floor: SoldStack 20%, us 51%. Our estimates missed the floor on half of them.

What we learned

1. Itemized lots are where the money is

A lot of 38 sneakers listed pair by pair, 10 pairs of customer-return boots with tags, a bag of Vuori and Lululemon. When the listing names what's inside, each item can be priced from its own sales. The median itemized lot was worth about 2× its winning bid at Poshmark prices, before fees. That's a real flip margin, and it's the case our old estimates handled worst.

2. Single branded items sell close to resale

One pair of Nike Pegasus, one Tech Fleece hoodie, one Coach purse: these went for about what they resell for (1.1× median), and nearly half sold above their Poshmark value. Single branded items are easy to look up, so other bidders price them correctly. The spread is thin.

3. Bulk lots need a steep discount

Poshmark prices are for one item sold online. Multiply that by 2,100 briefs or 200 belts and you get a number nobody could realize: 17% of counted bulk lots came out at 10× the winning bid or more. In a hand-checked sample, the winning bid was about 7% of the single-item value × count. It ranged from roughly a quarter to a third for shoes and boots down to under 3% for belts and basics. Bulk is worth something, but nowhere near retail × count.

4. Pallets are our blind spot

More than half of all clothing lots are pallets, gaylords and tri-walls described only by weight: "Pallet of clothes, approx. 540 lbs." Our estimates came in below the winning bid on 88% of them. Per-item prices don't fix that without knowing how many items are in the box, and GovDeals never reports what a pallet held after it sold. If you buy these, count and weigh what you get. That data is rare.

What we're doing with it

We're adding SoldStack as a data source for clothing, shoes and bags. It's in testing first: every day it values live itemized lots before they close, and we compare those values with how the lots actually sell. If the live results hold up like the backtest, it starts feeding GavelGap's clothing estimates for itemized lots. Bulk and pallet lots stay out until we have a bulk discount we trust. Same rule as everywhere else in GavelGap: real sold prices beat guesses, and a new source has to prove itself on real outcomes first.

Tips for pricing a clothing lot yourself

1. Price from sold listings, not asking prices. Poshmark and Mercari sold listings for clothing; eBay sold for sneakers and workwear works too.
2. Price the items the listing names, then add them up. One "lot of clothes" query prices the whole lot as one cheap item.
3. Match brand, item and condition. New with tags and used can be twice apart for the same item.
4. Watch kids' sizes. A youth 6Y Air Jordan sells for a fraction of an adult pair.
5. Discount bulk hard. Price one unit, then assume you'll realize a small fraction of that per unit across the lot, not the full price.
6. Skip what you can't identify. Unbranded single items had the weakest values in the whole set.

Methodology

The data covers 3,772 lots that sold on GovDeals between May 22 and October 23, 2026 in the Clothing; Shoes; Clothing, Jewelry, and Accessories; Handbags, Purses, Totes; Sports Jerseys; and Military Clothing & Field Gear categories. The winning bid is the final hammer price, without buyer's premium, tax or shipping. SoldStack classified each lot by type, read itemized, counted and single lots into items with a language model, and valued each item as the median of matching cleaned Poshmark sales, by brand and item where it had at least 5 sales, otherwise by category, department and condition. Lot values are before Poshmark fees and shipping. Uncounted lots received per-item values only and are not in the multiples above. "Our estimate" is GavelGap's resale estimate for the same lot from our existing sources (eBay sold and active comps, past GovDeals prices and AI valuations). A language-model reading of a listing can be wrong, so we report medians and shares, not individual lots. SoldStack provided API access for this test. GavelGap is not affiliated with Poshmark.

Related: Why AI alone can't price your flips · Why we read every sold comp twice · GovDeals vs. eBay prices · SoldStack →