Skip to main content
Before you use a recommendation in a live campaign, test it. Testing shows you exactly what a real customer would see, so you can catch a zero-result setup, a bad filter, or a mapping issue before it reaches your audience. Testing works the same way for Basic recommendations, which cover Item Attribute, User Actions, and Catalog Alerts, and for Advanced recommendations, which cover all the AI-powered models.

​
Test Recommendation Panel

You can open the test panel in the following two ways:
  • While you create a recommendation. The last step of the setup includes the Test recommendation panel, so you can test the recommendation before you save it.
  • From the Recommendations list. Open an existing recommendation and click Test recommendation.
To open the test panel for a recommendation that already exists, perform the following steps:
  1. On the sidebar menu in MoEngage, hover over the Content menu item Content icon. The Content menu appears.
  2. Click Recommendations. The Recommendations page appears.
  3. Find your recommendation in the list and click the edit icon under Actions. The recommendation opens. The Recommendations page listing the recommendations in the workspace
  4. Scroll down to the Test recommendation section, and then click Test recommendation. The test panel opens as a popup. The Test recommendation panel open on a recommendation
While the panel opens, MoEngage displays “Calculating the eligible users for this recommendation.” After the calculation finishes, the Random eligible user option shows the size of the qualifying pool, in the form “Test 10 users randomly preselected from N users who qualify for this recommendation.”

​
Panel Layout

The test panel has the following three areas: The panel also has the following controls:

​
Test Modes

MoEngage offers the following three ways to preview results:
By item is available only for Similar Items, Frequently Bought Together Items, and Frequently Viewed Together Items, because these models take a product ID as their input. Recommended Items takes a user ID, so test that model by user instead.

​
Test by Specific User

To test a recommendation for a specific user, perform the following steps:
  1. In the test panel, click By specific user.
  2. In the Select user by drop-down list, select ID, Email ID, Mobile Number, or MoEngage ID.
  3. In the User identifier value box, enter the value for the identifier you selected in the Select user by list.
  4. Click Show result. The recommended items appear under Test result.
Testing a recommendation for a specific user by choosing an identifier type, entering the value, and viewing the results These are the same identifier options available in campaign personalization preview, so you can test with whichever ID you already have. You do not need to look up a MoEngage ID first.
If more than one profile matches what you entered, MoEngage shows how many profiles it found and which one it selected. Use the drop-down list next to that message to switch to any of the other matched profiles, and click View user to open the selected profile and confirm it is the right one.
Searching by ID, Email ID, or Mobile Number can match several profiles, because one person can hold more than one. MoEngage tests the profile named in the message, so switch profiles if that is not the user you meant. The matched profiles list open, showing the MoEngage IDs that share the identifier you searched for

​
Test with a Random Eligible User

To test a recommendation with a random eligible user, perform the following steps:
  1. In the test panel, click Random eligible user. MoEngage samples 10 users at random from everyone who qualifies for the recommendation, so you do not need to enter anything.
  2. Click a user to see the recommendation results for that user.
  3. Click Test for another user to switch to a different user from the sample.
  4. Click View user to open the selected user’s full profile.
Testing a recommendation with a random eligible user and switching to another sampled user A sampled user can return no results. This means the recommendation produced nothing for that particular user, not that the recommendation is broken. Click Test for another user to check a different one.
If Random eligible user returns no users at all, a filter on the recommendation excludes everyone. To widen the eligible pool, clear the filter in step 2 of the recommendation setup, and then test the recommendation again.
Clearing or changing a filter in the Recommendation conditions pane does not refresh the random eligible users. That list is built when the panel opens, and the panel’s filter controls only affect the results shown on the right.

​
Test by Item

To test a recommendation against a specific product, perform the following steps:
  1. In the test panel, click By item.
  2. In the Product ID box, enter the product ID.
  3. Click Show result. The items MoEngage recommends alongside that product appear under Test result.
Testing a recommendation against a specific product by entering a product ID and viewing the results The product ID and the item ID are the same value. The product must exist in the catalog the recommendation uses. If it does not, MoEngage displays the error “The product ID does not exist. Please check the product ID.”

​
Recommendation Results

The Test result pane, headed Preview of [X] items, shows the items MoEngage recommends for the user, item, or sample you selected. In this pane, you can perform the following actions:
  • Search the results by title or ID in the Search by title and ID box.
  • Switch between grid view and list view using the view icons at the top right of the pane. Grid view shows the items as cards, and list view shows them as rows.
  • Review each item’s ID, name, short description, and price, exactly as a campaign displays them.
Searching the recommendation results and switching between grid view and list view

​
Dropped Items

Dropped items are items that the model recommended but that have since been deleted from the catalog. When this happens, the preview shows an expandable notice telling you how many items are missing and listing the exact missing catalog item IDs. To resolve dropped items, check that those IDs exist in an active feed, and that they match the attribute mapped as the item ID. This is the most direct way to find a catalog or event ID mismatch. Expanding the dropped items notice to see the catalog item IDs that could not be found For Trending Items recommendations, the test panel also displays the trend configuration in use, as described in the following table: For attribute trends, the By item and By attribute value options appear only when you select User Property in the value drop-down list. If you select Custom value instead, only the standard trending options appear. Selecting User Property to reveal the By item and By attribute value options

​
Recommendation Conditions

The Recommendation conditions pane has two parts. Generate Recommendation shows the settings the model uses to produce results, and is read-only in the test panel. Filter lists every filter currently applied, and you can change these while you test. MoEngage groups the filters under User action, Exclude user action, Item attribute, each marked AND or OR, and Sort Recommendation. To see how a condition affects the results, perform the following steps:
  1. Click the arrow next to a group to expand its individual conditions.
  2. Select or clear the checkbox for a group or an individual condition. The preview on the right updates immediately.
  3. Click Reset changes to revert everything to the saved configuration. This link becomes active only after you change something.
Selecting and clearing condition checkboxes while the results update alongside To find out why a particular item is missing from the results, turn conditions off one at a time until the item appears.

​
Apply or Discard Filter Changes

After you finish testing, click Done. If you changed any conditions, MoEngage asks whether to apply the configuration to the recommendation, and warns that applying replaces the saved configuration.
  1. Click Apply filters to save your test-time changes as the recommendation’s new live configuration.
  2. Click Don’t apply to close the test view and keep the recommendation exactly as it was saved.
The prompt asking whether to apply the modified filters to the recommendation
Apply filters overwrites the saved recommendation for everyone, not only for this test. If you are testing to diagnose a problem rather than to make a permanent change, click Don’t apply.

​
Troubleshoot Zero Results

A recommendation can return no results even when it is configured correctly, usually because of timing or a data mismatch. Check the following causes, in order:
  1. The interaction happened before you mapped the event. MoEngage starts counting an event toward recommendations from the moment you save its user action mapping, and does not backfill anything from before that point. If your test user’s only relevant activity predates the mapping, test again with more recent activity.
  2. The item ID in your interaction event does not match any item ID in your catalog. Check that the attribute mapped as the item ID has identical values in both the event and the catalog feed, with the same casing, no extra spaces, and the same string or number format. This is the same mismatch that produces dropped items, but when every interacted item is mismatched you get zero results instead of a partial list.
  3. No user action is mapped yet for the model you are testing. At least one of the four user actions must be mapped. Map user actions before you test a User Actions model or an AI-personalized model.
  4. The test user no longer exists, or has been merged into another profile. Test with the surviving MoEngage ID instead.
  5. You searched with an identifier type your account has no data for. Confirm that the type you chose in the Select user by list actually has activity for this user.
  6. The catalog used by this recommendation was deleted. Reconnect a catalog before you test again.
If none of these explain the result, contact MoEngage Support with the recommendation name and the identifier you tested with.

​
Related Articles