About IPX Recommendations (IPXRECS)

Coveo Machine Learning (Coveo ML) IPX Recommendation (IPXRECS) models provide recommended content to users in a Coveo In-Product Experience (IPX) search interface based on previous user activity.

The IPXRECS model learns from repeated occurrences of search and click events from previous user sessions in the IPX interface. The IPXRECS model then generates relevant document recommendations based on the commonly clicked documents of users visiting certain URLs.

Example of Coveo IPX recommendations under a searchbox

The recommendations appear when the IPX search interface loads on the web page or application. This is usually when a user opens an IPX interface, but it can also be when the IPX interface reloads. The list of recommendations is based on the URL at the time when the IPX interface loads. When a query is performed, the recommendations are replaced by search results, and the recommendations will only reappear if the IPX interface is reloaded.

Automatic model configuration

When you create an IPX search interface, an IPX Recommendation (IPXRECS) model is automatically created and associated with a query pipeline that the IPX interface uses specifically for recommendations. The IPXRECS model is preconfigured to rebuild and learn from new data on a weekly schedule, based on when the model is created.

Notes
  • The recommendation query pipeline and the associated IPXRECS model are already fully optimized for use with the IPX search interface. No further optimization is required.

  • IPXRECS models can’t be added manually. To create a new IPXRECS model, you must create a new IPX search interface.

  • A newly-created IPXRECS model may display an error status. This is expected because it doesn’t have enough click events to generate recommendations yet. Once 5 click events are available and the IPXRECS model rebuilds, the error disappears and the model will begin to provide recommendations.

    IPXRECS showing error status | Coveo In-Product Experience