Resources Administration & Configuration
AI Re-Ranking
Dynamic Re-ranking analyzes which search results users click and ranks them higher.
Why AI Re-Ranking?
AI Re-Ranking uses user behavior to adjust search results and category page ordering to drive sales. As customer behaviour changes with new products of seasonality the search results react automatically.
It’s worth noting that AI reranking overrides any other rules or settings except for pinned products. If you have pages that you want to merchandise for visual appeal or brand experience then disable re-ranking for that page.
NOTE: Before you can use AI Re-ranking, you have to track click and conversion events with Algolia. Coordinate with your data or front-end engineering team to implement event tracking.
From the Merchandising Studio homepage:
- Click the Merch tools link at the bottom of the left-hand navigation.
- Continue to Step 1 of the Enabling Dynamic Re-Ranking tutorial below.
- Go to the AI re-ranking page in the Algolia dashboard.
- Select the index for which you want to enable Dynamic Re-ranking.
- Click Start simulating and wait for the simulator to be ready.
- Change the Events source index if you want to use a different index to source your user events. By default, Dynamic re-ranking uses the events for the re-ranked index.
- Select the Browsing facet option to turn on re-ranking on browsing pages. By default, Algolia doesn’t dynamically re-rank empty queries. Setting a browsing facet enables re-ranking on empty queries with facet filters for that facet. If you want to re-rank category or collection pages, you will need to select a browsing facet to let Algolia know which facet defines your category structure. You will also need to ensure the site is set up correctly to use Algolia for Browse pages
- Enable Event freshness to give more weight to recent events. This setting improves seasonality and short-term search trends.
- Simulate the effect of re-ranking by typing sample queries in the Re-ranking Simulator and review the results. On the left of the Re-ranking simulator, you’ll see the results without re-ranking. In the middle, you can see how the order of the search results for a given query changes due to re-ranking. On the right, you’ll see the results after AI Re-ranking is applied. By default, Algolia will re-rank first 20 results, you can either reduce or increase that.
- Before fully activating AI Re-Ranking, it is highly recommended to launch an A/B test for AI Re-ranking from the AI Re-Ranking configuring page.
- Select Activate Re-Ranking to turn on AI Re-ranking in production.
- In the Launch Re-ranking A/B Test dialog, select the traffic split, test duration, and an optional Re-ranking filter. For example, if you want to turn on re-ranking only for products that are in-stock, select
instock = trueas re-ranking filter.