Understanding Repeat Purchase Prediction Scores

Repeat Purchase Prediction is a filter on the Segments Plus dashboard that gives the likelihood a consumer will make a purchase within the next 90 days. The score is based on a machine learning model that is trained per store. Meaning, it knows to take the information that is specific to your customers and build a predictive model unique to your brand. 

This article will explain what each score means and about the algorithm behind the predictions.

How it works

Read on to find out what’s behind the algorithm and how it works.

More than just segmenting. Repeat purchase prediction is a predictive model that is far more effective than simply segmenting based on the last purchase date. Repeat Purchase Prediction uses machine learning to take into account a customer’s purchase frequency and the average spent per purchase, then compares it relative to the other customers in your store.

The algorithm is trained over time with artificial intelligence (AI). Stores where consumers typically buy every six months and stores where customers typically buy every week need to be treated differently. The AI model takes this timing into consideration when it assigns a score.

Two stores will not be assigned scores in the same way. Each score is relative to within your store and the way your customers shop. Because different stores have varying repeat purchasing rates, two stores can’t be assigned Scores in the same way. A retail brand may have a 40% repeat purchase rate, but a travel brand may have a 5% repeat purchase rate — their customers’ purchasing habits are different, and the model that takes that variation into account and scores accordingly.

Accessing segments plus

To access your Segments tool, go to Customers > Segments Plus and scroll to locate the Repeat Purchase Prediction filter. See more about getting started with Segments Plus.

What the scores mean

The AI engine studies all of your customers’ purchase behavior, and it assigns each customer a prediction in the form of a Score of 0-4. The number is based on the customer’s purchase behavior in relation to the purchase history of all the other shoppers in your store.

Score = 4: These are the customers most likely to make a purchase in the next 90 days
(compared to all the other customers in your store.)

Score = 3: They’re relatively likely to make another purchase within the next 90 days
(relative to the other customers.

Score = 2: They’re less likely to make another purchase within the next 90 days.

Score = 1: These consumers are very unlikely to make another purchase within the next 90 days.

Score = 0: These are consumers who haven't made a purchase within the last 12 months, and therefore are very unlikely to make an additional purchase within the next 90 days.

Score = Insufficient Data: This will display if the store does not have enough data history to train an accurate model.

FAQ