Industries / Retail, hospitality and personal services / Retail and e-commerce / Thrift and resale
Pricing
Pricing is putting a price on items that are each one of a kind, then marking them down or pulling them from the floor if they do not sell.
5
Milestones achieved, from 1909 to 2023.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1909
Prices fall on a timetable
Filene's opened its "Tunnel Bargain Basement" in Boston in 1909. Each item was marked down automatically the longer it went unsold, and anything left after 30 days was given to charity. Thrift stores run the same kind of clock: Deseret Industries in the US prints on each tag the week the item reached the floor, uses a four-week rotation of colored tag fasteners, and pulls items after four to seven weeks.
Credit
Filene's
1999
Auctions for rare donations
Goodwill Industries of Orange County, California, set up an online auction site in 1999 to sell valuable and unusual donated items, so that bidders rather than a store's price list set their price; other Goodwill organizations paid it a fee to sell through the site, the Chronicle of Philanthropy reported in September 1999. Now ShopGoodwill.com, it is used by more than 130 Goodwill organizations and sold $341 million of goods in 2023, by its own figures as reported by The NonProfit Times.
Credit
Goodwill Industries of Orange County
2018
A public contest on price
Mercari, the Japanese resale app, ran the Mercari Price Suggestion Challenge on the data-science contest site Kaggle. It published data on items sold through its US app and asked entrants to predict each sale price from the listing. The winning team scored 0.3875 on root mean squared logarithmic error (RMSLE), a measure of how far predictions fall from real prices in proportion, where lower is better. Mercari held the closing ceremony on 9 May 2018.
LessonMercari's engineers found the winning model did not reach the same score on its Japanese app, so a price model built on one market's listings does not carry over to another unchanged.
Credit
Mercari; the Kaggle entrants
2022
Shop floors bend the list
Triona Fitton, a sociologist at the University of Kent, watched staff, volunteers and customers in UK charity shops and found that prices varied in practice and responded to haggling and to staff's knowledge of their customers, despite each charity's set pricing structures.
LessonA central price list does not settle the price on its own; people on the shop floor change it.
Credit
Triona Fitton, in JOMEC Journal
2023
A price from one photo
On 7 September 2023 eBay began rolling out a tool in its iOS app that drafts a listing from a single photo: title, description, category, and a suggested list price and shipping cost. TechCrunch reported the same day that sellers on eBay's forum and on Reddit were complaining that eBay's earlier AI description writer produced untrue text, such as saying a Pentax camera came with a lens kit.
LessonAI drafts of listings still need the seller to check each claim before the item goes on sale.
Credit
eBay
Watch for
AI prices by brand
Savers Value Village, which runs more than 380 thrift stores in the US and Canada, says its ThriftIQ pricing system was live in 108 stores in November 2026, had priced more than 40 million items, and that stores using it show faster gross profit growth. On its August 2026 earnings call the company said staff no longer grade each garment's condition and quality: they identify the brand, and the system sets the price from brand, season and sell-through (the share of items put out that sell).
Achieved whenan outside party, such as researchers given the store data, compares stores with and without ThriftIQ and confirms the gain.
Claimed by
Savers Value Village, August and November 2026
Open
01
Pricing tools, tested in shops
A six-week trial in a charity shop in which alternate incoming garments are priced either by a tool or by staff as usual, comparing the share sold within four weeks, money taken per garment and seconds spent pricing each one. The tool can be an open-source helper that reads the brand label from a phone photo and looks up the shop's own past sale prices for that brand and category, or a commercial app. The team handles no money; the shop sells as usual.
The briefCredit · unclaimed02
Test the color-tag clock
A trial of markdown schedules across the stores of one thrift chain. For 12 weeks, stores chosen at random keep their usual rotation, for example four weeks with a half-price week at the end, while the rest switch to another, such as six weeks or a smaller discount. Compare the money taken per item put out, the share sold before items are pulled, and the share sent to salvage.
The briefCredit · unclaimed
Open 01 · the brief
Pricing tools, tested in shops
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A six-week trial in a charity shop in which alternate incoming garments are priced either by a tool or by staff as usual, comparing the share sold within four weeks, money taken per garment and seconds spent pricing each one. The tool can be an open-source helper that reads the brand label from a phone photo and looks up the shop's own past sale prices for that brand and category, or a commercial app. The team handles no money; the shop sells as usual.
Why it is still open
Savers reports gains from ThriftIQ only from its own comparison of stores, and the system is not sold to other shops. Thriftly, an app that prices items from photos, said in April 2026 that it was used in 72 stores, without published results from a controlled comparison. Fitton's study found charity-shop prices drift from set structures in practice. A search found no randomized trial of any pricing method in a thrift or charity shop.
Built with what exists
A phone camera; a vision-language model that reads labels, such as Claude or Gemini through its API, or open OCR software such as PaddleOCR; the shop's sales records; Python; a label printer.
First step
Export three months of a shop's sales with brand and category, and check how many of 100 brand labels the helper reads correctly.
Done looks like
At least 2,000 garments over six weeks, about half priced each way, with share sold within four weeks, money per garment and pricing time reported for both groups with confidence intervals, and the code and anonymized data published.
Open 02 · the brief
Test the color-tag clock
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A trial of markdown schedules across the stores of one thrift chain. For 12 weeks, stores chosen at random keep their usual rotation, for example four weeks with a half-price week at the end, while the rest switch to another, such as six weeks or a smaller discount. Compare the money taken per item put out, the share sold before items are pulled, and the share sent to salvage.
Why it is still open
Thrift stores run fixed color-tag clocks, such as Deseret Industries' four-week rotation with pulls after four to seven weeks, and search found no published trial comparing schedules. Research on when to mark down comes from sellers of new goods, such as Moon, Bimpikis and Mendelson's 2018 study of a specialty retailer in Management Science, where many units of a style are alike and stock can be reordered.
Built with what exists
The stores' till data by tag color and week; a randomization script; Python with pandas and statsmodels; an analysis plan registered in advance on the Open Science Framework (OSF).
First step
Ask one multi-store thrift operator for a year of sales by tag color and week, and estimate how large a difference eight stores could detect.
Done looks like
At least eight stores over 12 weeks, with money per item put out and share salvaged for each schedule and their confidence intervals, the plan registered before the start and the results published whatever they show.