The available models for Recommend currently have the following upper limits for input:
- For Frequently Bought Together and Trending, there is an input limit of 3M events from insights fed into the model. We are only able to generate recommendations for products found within these events.
- For Related Products we have the same input limit of 3M events from insights for the Collaborative Filtering model. If no key attribute is setup, then we are only able to generate recommendations for products found within these events.
If key attributes are setup, then the Content Based Filtering model is used, which can provide recommendations for a maximum of 1.5M items with attributes, regardless of their presence in insights.
If both insights and key attributes are set, Collaborative Filtering and Content Based Filtering recommendations are merged.
If your number of items exceeds the upper limit for a particular model, not all of your items will be taken into account by the model during training.
These limits are set for scalability purposes. Our product roadmap does include work to expand these limitations.