Sherpa AI is crafted with a no-code-like platform experience. It only requires you to define the name, and description for your understanding and choose the events you would like to feed to the state-of-art algorithm.
However, there are some other basic pre-requisites that you need to meet before setting it up.
The recommendations need the following two things to be configured as pre-requisite.
- Catalog: The item catalog should be updated in the MoE portal.
- Map user actions: Set of events having user-item interaction. AI is fed with the last 2 months of such data points.
The followings are the criteria for the recommendation to be usable.
- At least 1 type of user-item interaction event having the relevant data points should be mapped to the User Action
- 1000 user-item interaction events in the last 2 months
- 25 unique users with at least 2 interactions each.
However, the more data, the better the recommendation will result in performance.
Create AI recommendations
As you navigate to the Recommendations landing page, you will see the ‘Create recommendation’ button.
You will need to select the AI ‘Recommended Items’ model in step 1 and define the ‘User Action’ in step 2 to complete the AI recommendations setup.
Select the AI recommendations in step #1 of configurations.
Choose the set of user actions you want to generate the recommendations with. You can also define the name and description from this page for your own ease.
Now sit and relax with ease, we got you! The Sherpa AI will start learning the user behavior and the status on the landing page will be shown as ‘Processing’ till then. In less than 24 hours, the model will be active and ready for you to use.
Note: You are allowed to have only one active AI recommendation at a time
There can be times when you want to add more or remove some events to optimize or experiment with AI recommendations. Your recommendation settings will be editable by clicking the pencil icon from the Recommendations landing page.
From the edit recommendations page, you will be able to change the name, description, catalog, or alter any User Action.
Note: If you are editing Recommendations to add more events, please ensure to define them in some User Actions.
If the model is in an active state the recommendations will be ready for each and every user. Click on the active AI recommendation and you will see the test screen at the end of the page.
You will be able to test the recommendation results by passing the MoEngage user id for any sample users.
As the users keep on doing activities, the algorithm keeps updating and refreshing recommendations. So you may see different results at different times for the same user.
Note: The edited recommendations can take up to another 24 hours to be Active for usage. The updated recommendations results will be shown only then.
You have an active AI recommendation, and you have tested the results - what’s next? It’s time to personalize your campaign with the recommended items. You can personalize Email, SMS, Push, In-App, On-site messages, and Cards using the recommendation feature.