Where our prices actually come from
A pricing app is only as good as its data. Here's exactly how ours gets collected, checked, and corrected — no black box.
From catalogue to your list
Collect
Each week, we gather every supported retailer's published brochure or digital catalogue — either photographed in-store or, for retailers we've cleared for it, fetched directly from their public website.
Extract
Each page is sent to Claude, Anthropic's AI model, with instructions to read product name, brand, size, regular price, special price, and deal type from the image.
Review
A person checks every extracted batch side-by-side with the source image before anything goes live. Anything the model flags as unclear — or a multi-buy deal like "2 for R60" — gets extra attention rather than a guess.
Publish
Approved prices are matched to existing products or added as new ones, each carrying a confidence score and a last-verified timestamp.
Delete the source image
The photo or downloaded catalogue page is never kept long-term — it's deleted after a short review window. Only the structured price data survives.
Good to know
How often prices update
Weekly, in line with each retailer's own specials cycle — not real-time, and we're upfront about that.
Automated fetching, done carefully
Where we fetch a retailer's site automatically, we respect their robots.txt and rate limits, and fall back to manual collection for any retailer that hasn't been cleared for it.
Future retailer integrations
We'd rather have a direct data-sharing relationship with retailers than rely on catalogues at all — see our Partners page.
Tell us — it helps everyone.
Extraction from real-world brochures isn't perfect, especially on crossed-out prices or dense layouts. Report anything that looks wrong and a person will check it.