Product reading

Product reading is the reader inside Wellsourced that turns a vendor product page, a tearsheet PDF or the page open in the Clipper into a product record — fields read and checked against the category's own specification, and photos chosen by looking at them.

A product page is written for a shopper, not for a specification. Reading it means finding the trade price beside the retail one, the width that is a width and not a chain length, and the photograph that is this product rather than the one beside it. The reader works in two passes: the first reads the product's spine — name, vendor, SKU, prices, dimensions, lead time — and proposes a category; the second re-reads the same page against that category's own fields, so a sconce gets a projection and a rug gets a pile height. Photos are judged by sight: one hero, alternates, details, dimension drawings and lifestyle shots kept, related-product tiles rejected, anything under print quality flagged. Every read value is marked as read until you change it. Reading is included in the seat and never counted.

What it does

  • Read name, vendor, SKU, trade and retail price, dimensions, weight, lead time, origin and description from a page or a sheet
  • Propose a category and subcategory, and write the category only when the read is confident; otherwise the board asks
  • Re-read the page against the category's own fields — canonical inches and pounds, the finishes and specifications the trade expects for that kind of product
  • Choose product photos by looking at them: one hero, then alternates, details, dimension drawings and lifestyle shots; tiles of a different product are rejected
  • Flag a hero photo below print quality, 1200 px on the long edge
  • Mark every read value as read, with its confidence, until a member edits it
  • Read a fabricator's estimate, quote or invoice into custom work: the partner, a working title, the amount, the document number and date, for the designer to confirm before the item exists
  • Serve the board's link and tearsheet intake and the Clipper's Read page from one reader

What it doesn't do

  • The reader has no login. A page shown only after signing in is read through the Clipper, or printed to PDF and added as a tearsheet
  • No browser rendering. A page that draws its product only in JavaScript may return nothing to read
  • A read is a snapshot at the moment it runs. Nothing is monitored afterwards
  • Read values are proposals for review, not verified facts. The item opens for the designer to check
  • Photos come from the page or the sheet. The reader does not look for better ones elsewhere

Questions this page answers

  • does wellsourced use ai to read product pages
  • can it pull the price and dimensions off a vendor page automatically
  • how does it pick which photos to keep
  • will it grab the photo of a different product by mistake
  • does it work on sites that need a trade login
  • how accurate is the extraction
  • is there a limit on how many products i can read
  • does it read metric dimensions
Every feature is included in the seat — nothing here is an add-on.
esc
Type to search project names and every binder in the studio.