Overview

Follow this guide to learn how Elevar generates and delivers Comeback Score as a Boosted Event. This article explains the data used to train the predictive model, how customers receive a prediction, and how the signal is delivered to Meta, GA4, and Klaviyo.

Required Inputs for Comeback Score:

Comeback Score is generated using historical Shopify data, including enriched customer- and order-level information. Elevar securely uses this data to train a machine learning model that identifies purchasing patterns unique to your brand. As additional customer and order data becomes available, the model continues learning over time to improve the accuracy of future predictions.

How Comeback Score Works in Elevar

How Elevar Generates Comeback Score:

Elevar uses a predictive classification model to estimate whether a first-time buyer is likely to make another purchase within 90 days. The model produces a return probability, which is then translated into one of two predictions:

Yes – The customer is predicted to return within 90 days.

No – The customer is not predicted to return within 90 days.

How Comeback Score Is Delivered as a Boosted Event

After the prediction is generated, Elevar delivers Comeback Score as a Boosted Event to supported marketing and analytics platforms. Each platform receives the prediction in a format that can be used for campaign optimization, audience creation, reporting, or customer segmentation.

Meta: Elevar sends the comeback_score parameter on the Purchase event with a value of Yes or No. In addition, dedicated Comeback Score Yes and Comeback Score No events are sent, allowing marketers to build audiences and optimize campaigns based on a customer's predicted likelihood to return.
Google Analytics 4 (GA4): The prediction is available as the comeback_score custom dimension with a value of Yes or No. Elevar also sends comeback_score_yes and comeback_score_no events, making it possible to report on, analyze, and build audiences around predicted repeat purchasers.
Klaviyo: The customer's profile is updated with a Comeback Score value of Yes or No, allowing marketers to segment customers and trigger personalized lifecycle marketing campaigns.

How Platforms Use Comeback Score:

Comeback Score allows platforms like Meta, GA4, and Klaviyo to act on a customer's predicted likelihood of returning.

In Meta, marketers can suppress customers with a Comeback Score of No from retargeting audiences to reduce spend on buyers who are unlikely to return. Customers with a Comeback Score of Yes can also be used as seed audiences for prospecting lookalikes based on shoppers who demonstrate a higher likelihood of repeat purchasing.

In GA4, marketers can analyze customers based on their Comeback Score by using the custom dimension and dedicated events to build audiences, compare behavior, and measure campaign performance across predicted customer segments.

In Klaviyo, customers with a Comeback Score of Yes can enter dedicated post-purchase flows designed to encourage a second purchase through loyalty incentives, product recommendations, or early access offers. Marketers can also reduce unnecessary discounts for customers who are already likely to return while reserving stronger win-back offers for customers with a Comeback Score of No.

Use Cases:

Performance Marketing

Stop spending on buyers who are unlikely to return: Suppress customers with a Comeback Score of No from retargeting audiences to avoid spending budget on one-time buyers with a low likelihood of returning. This allows marketers to redirect spend toward higher-potential customers.
Seed lookalike audiences with your best customers: Use customers with a Comeback Score of Yes as the seed audience for prospecting lookalikes. These customers already demonstrate behavioral patterns associated with long-term value.

Lifecycle Marketing

Nurture potential buyers toward a second purchase: Trigger a dedicated post-purchase flow for customers with a Comeback Score of Yes that includes upsells, loyalty incentives, or early access to new products. These customers are statistically more likely to purchase again.
Protect margins by reducing unnecessary discounts: Customers with a Comeback Score of Yes are already likely to return, so offering a large discount immediately after purchase may reduce margin unnecessarily. Suppress deep discounts for high-likelihood returners and reserve stronger offers for customers with a Comeback Score of No in win-back campaigns.