BRENNEN

Industries / Retail, hospitality and personal services / Retail and e-commerce / Pawn shops

Valuing items

Valuing is deciding what a pledged or offered item is worth, and so how much to lend or pay for it, usually within minutes at the counter.

4

Milestones achieved, from 1990 to 2025.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1990

    Loans at half of resale

    A pawnbroker lends only against the pledge, so the work goes into valuing the item rather than checking the borrower's credit. John P. Caskey, an economist at Swarthmore College, and Brian J. Zikmund, a Swarthmore student, wrote in the Federal Reserve Bank of Kansas City's Economic Review that the amount lent is "typically" 50% to 60% of the item's resale value and that a typical pawn loan takes less than ten minutes; at one northeastern shop whose police records they examined, 68% of 221 pledges taken over eight days were watches and jewelry.

    kansascityfed.org

    Credit

    John P. Caskey and Brian J. Zikmund

  2. 2005

    Auction sites join price guides

    In its annual report for 2000, First Cash Financial Services, a US pawn chain now called FirstCash, said it set an item's likely sale price from catalogs, blue books (printed price guides) and newspapers, plus recent selling prices of similar goods from its own stores' computer network, and that each employee, not a fixed percentage, decided how much to lend. Its report for 2004, filed in March 2005, added "on-line auction sites" to the list. Its report for 2025 names precious-metal spot prices, catalogs, blue books, online marketplaces, auction sites and retailer advertisements alongside its own sales history, and says it sets no minimum or maximum loan-to-value.

    sec.govsec.govsec.gov

    Credit

    FirstCash, in its annual reports to the US Securities and Exchange Commission

  3. 2017

    Cash first, item later

    On 28 June 2017 BANK, a Tokyo start-up, launched CASH, an iPhone app that priced an item from its brand, category, condition and a photo and paid at once. The user then had two months to ship the item free or to repay the money plus a 15% fee, which made it work like a pawn loan, ITmedia reported. BANK stopped making offers after 16 hours 34 minutes, when users had cashed in about 72,000 times for about ¥366 million and asked for pickup of more than 7,500 items, by the company's figures as reported by BRIDGE and ITmedia. It reopened on 24 August 2017 with a ceiling of ¥10 million a day on payouts, two weeks instead of two months to ship, and no fee for handing the money back, which ITmedia described as a move from pawnshop to plain second-hand buying.

    LessonPaying on a photo before the goods arrive is an unsecured loan until they do, so the payouts had to be capped and the shipping window cut.

    itmedia.co.jpthebridge.jpitmedia.co.jp

    Credit

    BANK and its chief executive Yusuke Mitsumoto; ITmedia and BRIDGE (reporting)

  4. 2025

    India standardizes gold valuation

    The Reserve Bank of India (RBI) issued directions on 6 June 2025, to be followed from 1 April 2026, for loans against gold and silver jewelry. Lenders may count only the metal, not stones, priced at the lower of the past 30 days' average closing price and the previous day's close for that purity, as published by IBJA or a regulated commodity exchange. On loans for spending, as against business, they may lend at most 85% of that value up to ₹2.5 lakh (₹250,000) per borrower, 80% up to ₹5 lakh and 75% above, must test purity in front of the borrower, and must post their method for net weight and price on their website. Manappuram Finance, a large gold lender, posts such a method dated April 2026.

    rbi.org.inmanappuram.com

    Credit

    Reserve Bank of India

Watch for

  1. AI prices from a photo

    Bravo Store Systems, a US maker of pawnshop point-of-sale software that bought the PawnMaster products from Data Age in January 2024, launched its "Shopkeeper AI Estimator" as a beta in April 2025, saying it identifies items from photos "with remarkable accuracy", detects scratches and wear, and suggests prices from current market data. Its product page now says the Estimator suggests buy, loan and resale values inside the point of sale, and that more than 3,300 stores use Bravo. Neither Bravo nor anyone else has published error figures.

    Achieved whenan outside party publishes how far the Estimator's suggested prices fall from the prices the same items actually sold for.

    bravostoresystems.combravostoresystems.com

    Claimed by

    Bravo Store Systems, April 2025, repeated on its product page as read in November 2026

Open

  1. 01

    Photo price tools, tested blind

    A public test of tools that price a used item from photos. Photograph 300 items as a cooperating pawnshop or resale store puts them on sale, record the price each actually sells for, and ask each tool for a price from the photos alone: Bravo's Estimator if Bravo grants access, Google Lens followed by a fixed look-up of sold listings, a general vision-language model, and two consumer "what is it worth" apps. Publish each tool's median error and how often it was more than 25% off, by category.

    The brief
    Credit · unclaimed
  2. 02

    India's gold-loan methods compared

    Collect the valuation method that each Indian bank and finance company lending on gold must now post, turn each into a formula, and value the same 20 sample ornaments (described by weight, purity and stone weight) under every method, to show how much the same ornament could borrow from each lender.

    The brief
    Credit · unclaimed

Open 01 · the brief

Photo price tools, tested blind

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public test of tools that price a used item from photos. Photograph 300 items as a cooperating pawnshop or resale store puts them on sale, record the price each actually sells for, and ask each tool for a price from the photos alone: Bravo's Estimator if Bravo grants access, Google Lens followed by a fixed look-up of sold listings, a general vision-language model, and two consumer "what is it worth" apps. Publish each tool's median error and how often it was more than 25% off, by category.

Why it is still open

Bravo's launch release claims "remarkable accuracy" but gives no figure, and its product page gives none either. Research models that price second-hand goods from photos, such as one trained on a Japanese retailer's used jewelry (Yamaura and colleagues, 2019) and one built for the Chinese marketplace Xianyu (Han and colleagues, 2020), were trained and scored on private data, and none tests the tools sold to pawnshops.

bravostoresystems.comarxiv.orgarxiv.org

Built with what exists

A phone camera and a light box; Google Lens; a vision-language model such as Claude or Gemini through its API; the consumer apps as sold; the store's own sale records; Python with pandas for scoring. The team makes no loans and handles no money; the store sells as usual.

  1. First step

    Ask one resale store or pawnshop to let the team photograph 50 items going on sale this month, and price them with two tools to time the work.

  2. Done looks like

    300 items in at least six categories (phones, game consoles, power tools, guitars, gold chains, watches), each priced by at least four tools, with photos, sale prices and errors published under an open licence.

Open 02 · the brief

India's gold-loan methods compared

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

Collect the valuation method that each Indian bank and finance company lending on gold must now post, turn each into a formula, and value the same 20 sample ornaments (described by weight, purity and stone weight) under every method, to show how much the same ornament could borrow from each lender.

Why it is still open

Paragraph 22 of the RBI's 2025 directions requires lenders to post their method for net weight and the price used. Manappuram Finance's method shows choices that can differ between lenders: it values purities between 22, 18 and 14 carat at the next lower published rate with a weight cut (4.5% for 21 carat), and adds 1% to 8% to its usual purity deduction when stones make up more than 5% of the weight. Searches found the rules and lenders' own explainers, and no comparison across lenders.

rbi.org.inmanappuram.com

Built with what exists

Python (requests, pdfplumber) to collect the posted pages and PDFs; IBJA's daily rates; pandas for the calculator; a document-reading model such as Muse Ember to draft each lender's formula from its text (see the Note "The jobs agents repeat"), checked by hand.

  1. First step

    Download the posted methods of five large lenders, starting with Manappuram's, and value three sample ornaments under each by hand.

  2. Done looks like

    Methods from at least 25 lenders, including the ten largest gold lenders, each turned into a tested formula; the 20 sample ornaments valued under all of them; the spread published, with each lender sent its row for comment before publication.