DocsShopping
Product detail
One product's shelf performance across the AI shopping carousels it appeared in — win rate, visibility share, average slot, price-versus-shelf, and a trend.
The product detail page zooms into a single product and shows how it performs on the AI shelf: how often it wins the top slot, how much of the shelf it covers, and how its price compares to the shelf around it.
What it does
For one product, the page shows its shelf performance across the AI shopping carousels it appeared in — the metrics that describe how prominently ChatGPT featured it.
How to use it
The page has five parts:
- Five shelf tiles summarising the headline metrics.
- Matched placements — a list of up to 100 rows, each linking to its receipt so you can read the answer the placement came from.
- A price-versus-shelf card comparing your price to the shelf's typical price.
- A trend card.
How it's computed
- Win rate = slot-#1 placements ÷ total appearances.
- Visibility share = distinct carousels featuring the product ÷ distinct carousels in the window. A product holding two slots of one carousel is not double-counted — the numerator is distinct carousels, not placements.
- Average slot = mean position, and it is an em-dash when the product never appeared — there is no average of nothing.
Price handling is deliberately honest:
- When an engine sends a structured price (an exact amount plus currency), it is stored as-is; otherwise the amount is parsed once from the engine's verbatim price text. Some engines send only numbers with no display string — the shown price is then formatted from the stored amount and currency, and when the engine sent nothing at all, the price stays null, never guessed.
- Prices that can't be parsed cleanly (
Free, percentages, ranges) become null rather than a made-up number. - A product's typical shelf price is the median of per-carousel medians, so a single 12-slot carousel can't outweigh three 3-slot ones.
- A price-vs-shelf delta is only shown when both sides share a known currency; otherwise it renders as an em-dash with the exact reason stated. See how MentionFlow handles missing data.
Limits
- The matched-placement list is capped at 100 rows.
- Metrics reflect the rolling shopping window across every collecting engine — ChatGPT, Google AI Overviews, Google AI Mode, and Copilot today; see Engine coverage.