Apparel visual search

Use case

A user sees an item worn by a friend or on display in a store, perhaps while comparison shopping.

Challenges

• Visual search has limits, so identification has to fail gracefully.
• The app has to beat simply searching Amazon or Google, especially when an item is misidentified.
• With several items in frame, users need a way to select one. Results are layered over the camera like pins on a map, without adding friction to checkout.

Key features

• Visual search: the camera is instantly on and searching at launch, with text and voice as alternate inputs.
• Shopping results: one-click purchase straight from results, with receipt and shipping info.
• Supporting pages: search history, preferences and failure states.

Results hierarchy

Keep users browsing even when the match is wrong or unwanted:

1. Hero image for the most likely result, for a quick purchase
2. Other likely results
3. Visually similar results
4. Recovery when the app misidentifies the item
5. Alternatives, such as a less expensive item
6. Related categories
7. Items people also buy