Purchase propensity
This classification model scores all customers in the population on their likelihood to make a purchase within a given window of time.
How does it work?​
The model predicts the probability that an individual will make a purchase in the next 30 days. If needed, we can assign a custom prediction window that fits your unique sales cycle. We transform each customer’s unique data into a custom feature set, with automated training and tuning to deliver the most performant model possible.
The model uses a historical window to determine who is eligible for scoring. For the Next Purchase model, the default historical window criteria is that a customer must have made a purchase, clicked an email, or triggered a web event within the last 90 days. After training and tuning, the model is ready to score the active population.
The model’s output is a normalized SCORE that ranks customers from 100 (most likely) to 1 (least likely) to make a purchase. The model then assigns the customer into a percentile, creating a (mostly) even distribution that ranks customers against each other. A customer with a score of 99 is considered more likely to make a purchase than a customer with a score of 75, who in turn more likely to make a purchase than a customer with a score of 25. We then make the score available to use in segmentation or personalization for a wide variety of marketing use cases.
Requirements​
Behavioral data (such as clicks, opens, sends, and push interactions) from one or more channels (email, push, in-app, and so forth) and transaction data (purchases, conversions) needs to be available in the your data warehouse.
- Data Input Schemas for Transaction/conversion data
Schema​
| MG Field Name | Type | Range | Description |
|---|---|---|---|
| RecipientId | VARCHAR | The unique customer identifier, the key to join back to customer tables. | |
| MLPurchasePropensityScore | INT | 1-100 | 1-100 normalized output of scoring. |
| MLPurchasePropensityLow | BOOL | 0 or 1 | Predefined segments. Our recommended audience definitions today break out high/medium/low, where the top quintile = high; mid three quintiles = medium; bottom quintile = low. However, we can customize this during implementation. |
| MLPurchasePropensityMedium | BOOL | 0 or 1 | |
| MLPurchasePropensityHigh | BOOL | 0 or 1 |