Redesigning image search
Designing an efficient visual search experience for 10 million shoppers.
Context
Shopee's image search feature enables users to snap or upload a photo to find matching products across the marketplace. Despite over 10 million users relying on visual search, order conversion from this feature was more than 50% lower than the platform average. I led the end-to-end redesign of the image search experience to improve search efficiency and conversion.
Approach
Funnel analysis was carried out to understand friction points across the current image search user journey- from the home screen entry point through camera capture, recognition, and results browsing. Key datapoints were then synthesized to better understand how users were interacting with image search (e.g. more than 80% of users usually save screenshots or images in their photo albums and search thereafter, instead of directly taking a photo- which signified opportunities to streamline the experience for greater efficiency).
Competitor analysis was also conducted across other leading e-commerce platforms to benchmark best practices in visual search UX, focusing on camera interactions, loading states, and results presentation.
Solution
Targeted design enhancements were carried out that sought to increase efficiency, retention and conversion. A few noteworthy ones:
1. New image shortcut prompt to select the most recent screenshot is added to enable quicker image selection for search.
2. Clearer system status processing indicator to manage user expectations and reduce the likelihood of dropoff.
3. Instant transition to the results page, where the step of needing to 'confirm' the image selection is eliminated.
4. Same-page image editing and refresh when adjusting the image search, to reduce cognitive load so users are not venturing to a separate page for edits. Users can edit, crop and re-select predictively detected objects within the captured image for more precise matching.
Outcome
The redesign achieved the following outcomes:
1. The screenshot prompt function shortened 88% of search paths by 6%.
2. End to end funnel conversion/retention from entrypoint to search results page increased by 4%.
3. Order conversions from image search increased in countries such as Malaysia, Indonesia, and Philippines by 10%.
This also laid the foundations for subsequent image search feature optimizations.