Recommenders
Recommenders allow you to use any event properties from your datasets to predict which product, service, or other offer any individual will respond best to. For example, a recommender that uses the event property product will output the top products that an individual is likely to respond best to. This insight can help you tailor how you engage with them in order to achieve the best results. Recommenders aren't exclusive to just product recommendations–they can also be used to predict next action in a series, like next logical banking action.
Recommenders are just one of the ways that customer predictions can help you make every experience that a customer or lead has with your business exceptional. Try them in combination with persona predictions to maximize the customer experience.
Getting started
Inside Recommenders, you'll find a list of your current recommenders if you have any, with columns for:
- Event: the dataset's event that the property belongs to.
- Property: the property that you are predicting in the recommender.
- Status: whether the recommender is ready, building, or errored.
Creating a recommender
📘When should I use a recommender instead of an outcome?
Recommenders are best suited to predict events that are part of a series, such as when you're interested in seeing if individuals will purchase product B or C after having purchased product A, or if the next logical action a banking customer might take is set up auto-deposit after having created an account. For conversion predictions, such as likelihood to churn, create an outcome instead.
- Select new recommender in the upper right of the Recommenders list view.
- Select the event that contains the property you want to make a prediction on.
- Next, select the event property. This is what you want to make a prediction on, and might take the form of a product type, product category, transaction or service type, or some other offer.
📘Creating event properties
Event properties are defined when adding an event to a dataset.
- Give your recommender a unique name and click save recommender to finalize.
With your recommender built, it's now ready to be used in the payload to make predictions when creating a pipeline.
Deleting a recommender
To delete a recommender, click the options menu (three dots) on the far right of the recommender you'd like to delete, then click delete. If the recommender is in use by a pipeline, the delete recommender popup will indicate that you need to modify any pipelines using it in order to delete the recommender. Once there are no pipelines using this recommender, you can safely delete it.
📘Deleting resources
See the object preservation for info on the order in which resources should be deleted.
