How to cost-effectively acquire more high LTV customers with Predictions

Learn how Predictions can help you improve ad targeting to decrease your cost per acquisition (CPA) while also converting more customers with high lifetime value (LTV).

Made by Megan DeGruttola

What do you need?

  • Twilio Engage

  • Any ad platform (Facebook Custom Ads is specifically called out in this recipe)

Easily personalize customer experiences with first-party data

With a huge integration catalog and plenty of no-code features, Segment provides easy-to-maintain capability to your teams with minimal engineering effort. Great data doesn't have to be hard work!

On this page

In order to grow, businesses need to attract new customers — particularly ones that will keep coming back to buy again and again. However, it can be difficult and time consuming for marketers to figure out which people are not only ready to buy but will also become loyal, repeat customers. 

To avoid wasting time and ad dollars targeting the wrong people, Predictions is here to help. 

With Predictions, we’ve put the power of predictive AI directly in the hands of marketers to help quickly identify customers with high predicted LTV — without needing the help of technical teams. Then, we make it easy to use those Predictions to create lookalike audiences on your preferred ad platforms for a greater return on ad spend (ROAS) and higher conversion of high LTV customers.   

Step 1: Create the Prediction

To create a Prediction in Segment, you’ll need to enter the Predictions builder by going to Unify > Traits > Create Computed Trait > Predictions > Select a Predictive LTV.

Next, you need to select the event that represents a purchase for your company, as well as the event property that represents the total order amount. For a typical retail business, you’d select “order completed” as the purchase event and “price” as the property for order amount.

Screen showing setup fields for Purchase Event and Order Amount to configure predicted lifetime value trait.
Interface for setting up and previewing the predicted lifetime value (LTV) trait in an analytics platform.

Then, click “calculate” on the right hand side of the screen to see how many users fit this Predictive LTV model. 

Click next until you get to the “Review and Create your Trait.” Name your Prediction something descriptive, like “High Predictive LTV Trait” and click “Create Trait.”

Step 2: Build your targeted audience

Once your Prediction is computed, the next step is to build an audience. Click into the Predictive Trait you named “High Predictive LTV Trait” from your Traits view under Unify in your workspace, then click into the Prediction tab in the top left corner.

Here you’ll see the “Explore your prediction” interactive chart. At the top, there are pre-defined audience cohorts for you to select from. Or, you can choose to create a custom cohort using the slider feature on the chart.

Graph showing probability of user conversion by percentile, divided into top 20%, bottom 80%, and top 50% groups.
Probability graph of user conversion rates with top 20% and top 50% users having higher likelihood to convert compared to the bottom 80%.

For this use case, let’s select the top 20% cohort, which will give us an audience of the users who are in the top 20th percentile most likely to have high LTV. Click the “Create audience” button on the right side of the chart.

Step 3: Send audience to your preferred ad platform

Once you’ve previewed your audience of the top 20% high LTV, you’ll select a destination to send it to. Since we want to use this audience to create lookalike ad audiences for our acquisition campaigns, you’ll want to send it to your preferred advertising platform

For this example, let’s select the Facebook Custom Audiences destination

Step 4: Use your high predicted LTV audience to create your lookalike audience in Facebook Ads

Facebook Ads allows you to create a lookalike audience based on the source “custom audience.” Facebook users with similar demographics, interests, and locations will be selected for advertising. 

By using your top 20% high predicted LTV audience to power your lookalike audiences in Facebook, you should dramatically increase the return on investment from your Facebook Ads spend — gaining higher quality customers at a lower cost.

If this top 20% LTV audience performs well, try expanding it to the top 30% high LTV to increase the potential impact of your acquisition campaign. 

Watch this webinar to hear how Twilio Segment customer, The Motley Fool, used Predictions to reduce lead acquisition costs on Facebook ads by 34% — and acquire customers with a 9% higher LTV than those acquired with non-AI audiences.

Getting started is easy

Start connecting your data with Segment.