Adding AI Re-Ranking to Category Pages
You can leverage Algolia’s AI Re-Ranking feature to find trends in your users’ behavior. Based on users’ interaction throughout search and browse, AI Re-Ranking improves the relevance of your category pages by boosting results that are trending or rising in popularity.
Once you set up the category pages and successfully send click/conversion events (requires coding) to Algolia, Algolia starts using this data to detect trends in your user behavior. There is no coding required to enable AI Re-Ranking on category pages.
Learn more about AI Re-Ranking’s effect on relevance.
There are four steps to implement this solution:
- Build the category pages using Algolia
- Send click and conversion events to Algolia
- Enable AI Re-Ranking on category pages
- A/B test AI Re-Ranking on category pages
Step 1: Build the category pages using Algolia
To apply AI Re-Ranking on category pages, you must first create those Algolia-powered category pages.
Step 2: Send click and conversion events to Algolia
You must complete this step before enabling AI Re-Ranking.
AI Re-Ranking refreshes each query’s ranking every 24 hours. Re-Ranking always takes the last 30 days into account, meaning it’s based on a sliding window of the most recent analytics data.
Step 3: Enable AI Re-Ranking on category pages
You can configure the following options from the AI Re-Ranking page in the dashboard.
- Opt in an index for AI Re-Ranking.
- Select the correct events source index (optional).
- Pick your category page identifier by choosing a Browsing facet.
- Adjust weight to recent events (optional)
- Simulate AI Re-Ranking.
- Activate AI Re-Ranking.
You can see re-ranked results only after Algolia has received enough search traffic and click and conversion events. For a record to be promoted for a given query, Re-Ranking requires it to gather at least 20 clicks or two conversions during the 30 days time window.
Opt in an index for AI Re-Ranking
Navigate to the AI Re-Ranking page in the dashboard and select an index you want to use for AI Re-Ranking.
Optional: Select the correct events source index
Under AI Re-ranking, click Events source index and update the source index where you send your click and conversion events in production. If you need to use a different source for your events, you can select the correct events source index.
A main use case for selecting an alternative source index is when you have a staging index that replaces the production index every day for atomic reindexing. In this case, you should enable re-ranking on both the staging and target production indices as source events, given that the production index already receives events.
Pick your category page identifier by choosing a “browsing facet”
Under AI Re-Ranking, click Browsing facet and update the browsing filter to the category page identifier you indexed in the dataset.
A category page identifier is a unique identifier that references a particular category page.
Possible identifiers include
Please check with your team to make sure the browsing filter is referencing the correct filter used in the query to generate the category page in the front-end UI.
Optional: Adjust weight to recent events
By default, the event freshness setting is enabled to handle seasonality and shorter trends. You have the option to turn off this setting, but it’s recommended to leave it on.
Simulate AI Re-Ranking
Once Algolia receives enough search traffic and click and conversion events, you can assess the effect of re-ranking on your search results/browse page in the simulator. You should use the simulator before activating the feature in production.
Activate AI Re-Ranking
Before activating AI Re-Ranking, you can turn off re-ranking for specific records.
To prevent promoting popular items that are out of stock or no longer listed, select a filter value, for example
Turn off Re-ranking for specific records:
A/B test AI Re-Ranking on category pages
To measure performance of the Re-Ranking approach, it’s important to launch the A/B test directly from the Re-Ranking section of the UI and not from the A/B test tab in the dashboard. This ensures all the settings are configured correctly and that you are measuring the performance only on the pages with enough traffic and where Re-Ranking rules apply.