You can combine several RS using the Advanced template in Exponea. This option allows you to create a customized model by combining more models from and setting additional options.
The following engines are available in Exponea for you to combine
| Name | Which model it is |
|---|---|
| Content based | Similar items template |
| Preferences of many users | Personalised recommendations for you template (includes an arbitrary number of events with custom weights (weight is a positive number, where higher the number, higher the "weight" or "importance" of the event) to set their strength for indicating user similarity) |
| Manual selection | Manual selection technique |
| New in stock | New items template |
| Previously seen items | Customer recent interactions template |
| Chosen by metric | Chosen by metric template |
You can add any model from the list by navigating to “Choose and set up the recommendation models” and hitting the “+ Add engine”:
Example of combining two models (Similar items + Personalised recommendations for you) with priorities set as 1 and 2 (described below). In this example, all of the results from model Similar items will be returned at first (priority 1) and then all of the results from the model Personalized recommendations for you (priority 2) will follow:
Example of Advanced template with two models.
After you pick preferred engines, you can also choose a combination strategy for building the recommendations:
| Combination name | What does it mean |
|---|---|
| Only the best one | We will pick the best model according to the performance. Performance is measured after at least 1000 hits on each engine by the live traffic. After this process, only the best engine will serve. |
| Combine multiple | Recommendations will be mixed from all selected engines results. |
| Order by priority of models (you can pick the priorities in the previous step while designing the engines) | Recommendations will be composed using engines ordered by their priority (1 is the highest one). Practically it means that we will primarily use the engine with priority 1 and if this engine will not be able to deliver enough results, we will use the engine with priority 2, and so on after the required number of items can be returned. It is useful to define custom fallbacks if any recommendations model is not able to personalize. |
Combination strategies explained visually.
Combining models
It is recommended, especially for email recommendations, to combine different engines in a model, so that customers get the most relevant recommendations. As the last priority (the highest number), we recommend having a manual selection engine that the model will default to if all the previous ones would fail to deliver any or enough results. Please note that more recommendation models you combine, the API latency is proportionally increasing.
Updated 17 days ago
What´s next?
| How to setup pre-built templates in Exponea |