Build a recommendation

This article will walk you through the steps of creating a new recommendation to be used in your personalization campaigns.

Build a new recommendation

To create a new recommendation, open Amplitude Audiences, navigate to the Recommendations tab and click Create Recommendation. Then follow these steps:

1. The first step in the process is defining your outcome. The outcome is the goal you are trying to reach, or the metric you are trying to improve.
   In the Select Starting Cohort section, choose the cohort this recommendation will train its algorithm on. By default, Amplitude Audiences selects all users active in the last 90 days, but for best results, you should use a more narrowly-tailored cohort. To do that, click Define Your Own and define your cohort following the process described here.
2. In the Define Your Outcome section, choose the desired outcome event for this recommendation.
3. Next, specify a time frame for the user to undertake this outcome event. The default setting is one hour. Then click Next > to move to the next step.

NOTE: There is a limit of recommending on cohorts < 20 million users

4. The second step is creating your item catalog, which is choosing the items you want to recommend in order to reach your desired outcome. Under Define Items To Be Recommended, click Select Event… to choose the exposure event.
   
   
5. Click Select property... to designate the item to be recommended to the user. You won’t select the item itself; instead, you’ll choose an event property associated with the exposure event. The recommendation will choose the recommended item from the values of this event property.
   
   For example, a music app might want its users to buy concert tickets from within the app. It might show users who followed a playlist a concert popup, based on the genre of the playlist they followed. In this case, the event property selected here would be genre, and it would be attached to the follow_playlist event.
   
   More generally, the event property will often be something like “SKU”, “ID”, “Name”, etc.
   
   
6. Next, specify the number of items to be recommended to each user. By default, Amplitude Audiences will choose from the 50 most-frequent (based on 30-day uniques) values of the event property you selected in the previous step.
   
   However, you can also set your recommendation to work from a static list of property values instead. To do this, toggle the Create with Static List toggle and select the candidates from the list of options.

You can exclude specific values from a dynamic recommendation as well. Then click Next > to move to the next step.

7. In the New Recommendation tab, give your new recommendation a name and description.
8. Finally, use the Define Holdout slider to configure the percentage of users to include as a control. These users will be randomly selected to receive a random set of items as a recommendation, as a control group against which you can measure the lift this recommendation generates.
9. When you’re finished, click Build. It’ll take about an hour to generate your recommendation. You’ll receive an email when it’s ready.

Understand your recommendation

Once your recommendation is complete, you can view some basic information about it by clicking on it. It will open to the Overview tab.

The first thing you’ll notice is the confidence score. This represents Amplitude’s confidence that this recommendation will generate statistically-significant life, relative to a random list of items.

NOTE: If the confidence score is less than 60, you should not use the recommendation.

Below that, you’ll find a list of items ranked by their frequency of appearance in the recommendation. The item at the top of the list is the one most commonly suggested to users by this recommendation, and is the one most likely to result in a conversion.

Common mistakes in creating a recommendation

• Using the wrong cohort. While recommending all users can be sufficient, sometimes you need to be more specific. Think about your goal and which users are the best candidates to achieve it. If your goal is to optimize for a second purchase, for example, select users who have already purchased once as the starting cohort.
• Specifying the wrong outcome. Your outcome event will dictate the rankings of the selected items. So if you optimize for “product purchased WHERE amount > $100,” the recommendation will prioritize expensive items. Be sure that’s what you want before launching the recommendation.
• Wrong exposure event timing. If you choose an exposure event that occurs after the outcome event, the recommendation will not train properly.
• Wrong exposure event context. The context for the event property depends on the exposure event it’s configured from. The event property’s name can easily mean different things on “product clicked” vs “button clicked”.
• Wrong event property. For example, Product Name and Product ID convey the same basic information, but in very different formats. Make sure the one you select matches the way data is stored in your CMS.