83% of "was" prices were never the price. We measured it.

Dylan Ferrara — Fri, 02 Oct 2026 04:13:06 +0000

We run fazy.com, a deals site that records what products actually sold for. Last week we asked our own data a simple question: when a retailer shows a "was" price, was that ever the price?

For 83% of products, the answer was no. Not once in our tracking history did the product sell within 2% of its "was" price.

How we measured: every product in our catalog showing a reference price at least 5% above its current price, price read within the previous three days, tracked for at least 30 days with 4+ observations. Average watch time: eight weeks, 13 observations each.

By category, the worst offenders: wellness 97%, cosmetics 96%, home 91%, footwear 91%, electronics 88%. Apparel was the most honest at 68%.

43% of products showed the same price at every single check. A crossed-out price that never moves is not a discount. It is the presentation.

Full methodology: https://fazy.com/blog/was-prices-we-never-saw

Why I am posting this here: the dataset is queryable. We expose a free, no-key deals API and MCP server with price history: https://fazy.com/developers and https://fazy.com/mcp. Any agent can ask "what did this actually cost over the last 90 days" instead of trusting the tag.

I am the founder of Fazy, so read the numbers skeptically and check the methodology. That is the point of publishing it.

I wired a free no-key deals API into my agent: 7 MCP tools, verified discounts

Dylan Ferrara — Fri, 02 Oct 2026 04:02:28 +0000

I built fazy.com, a price-checked deals site. Last week I needed deal data inside an agent without scraping, so I used our own API. It is free, no key, 60 req/min, and read-only. Here is the whole integration.

The API: https://fazy.com/developers
The MCP server: https://fazy.com/mcp (7 tools: search_deals, get_deal, get_price_history, similar_deals, list_categories, catalog_slice, post_brief)

Why it is different from scraping: every deal carries a verified discount_pct plus the current and reference price, and get_price_history returns the recorded price changes with dates. So the agent can answer "is this actually a good deal" instead of trusting the store's "was" price.

We ran our own catalogue against those "was" prices: 83% of them were never the price while we watched (methodology: https://fazy.com/blog/was-prices-we-never-saw). That is the dataset your agent gets to query.

Minimal example, curl, no key:

curl "https://fazy.com/api/v1/deals/search?q=robot+vacuum"

Response includes discount_pct, price, reference_price, merchant, in_stock, and a url field per deal. For MCP clients, point at https://fazy.com/mcp and the tools show up.

Repo: https://github.com/FazyFer/fazy-mcp

I am the founder, so obviously biased, but the API is free and the data is the interesting part. Happy to answer questions about the price-tracking methodology.

DEV Community: Dylan Ferrara

83% of "was" prices were never the price. We measured it.

I wired a free no-key deals API into my agent: 7 MCP tools, verified discounts