Artsy: Search
Artsy is an online art marketplace that aggregates artworks from leading galleries around the world. Users can shop for art, browse shows and auctions, and learn about the art world at large using Artsy’s online platform or iOS app.
Part I: Auto-Suggest
Onsite search traffic was especially significant to Artsy, with one of the highest conversion rates to commercial action (inquiry, bid, buy, offer). Users who searched were generally more targeted in their pursuit. Therefore, it was critical that search performed with accuracy and expediency.
Auto-suggest selection rate was already high, at about 75%, but we saw an opportunity to turn more of these sessions into commercial action. We also planned to introduce this functionality to mobile users for the first time, which was likely to help boost conversions overall.
First AB Test
The goal in our initial AB test was two-fold. From a user perspective, our goal was to give users more confidence in making selections from auto-suggest. From a business perspective, our goal was to see selection rates increase while additionally increasing conversions to artwork pages.
In our test we monitored four key metrics—conversion from search to artwork pages, sessions involving search, selection rate from auto suggest, and conversion from search to artwork to commercial action.
Revised AB Test
Our first attempt introduced two new elements to the experience: entity titles and an artwork preview pane. When we rolled out the test, we saw a significant drop in selection rate and only a 1% increase in artwork page views.
We realized we had too many new variables to which we could attribute these results. We scaled back the design to only introduce one new element from the previous experience and ran the test once more. This time, we were able to pinpoint which changes were successful and which were not. In our revised AB test we monitored only one key metric—selection rate from auto-suggest.
Results
We saw a 15.2% lift in item selection rate overall with no significant shift in artwork page views. Since artwork page conversion rates remained relatively stable, the test was considered considered successful. We shipped this new design to all users.
Part II: Search Results Page
The previous search results page delivered results in a view that mixed entities into a single list. There was no way for users to refine the results and they were forced to rely on our sorting logic to provide relevant results.
Similar to auto suggest, we wanted to give users confidence and control over their experience by distinguishing entity types. From a business perspective, we wanted to optimize this experience for our marketplace.
We A/B tested a new inventory-driven search results page that introduced an artwork grid and distinct tabs for specific entity types. For this test we monitored two key metrics— conversion from search to artwork page, and overall selection rate.
Previous Design
New Design