Product recommendation
The Product recommendation model discovers relationships between products through a technique called collaborative filtering.
How does it work?
This model generates a ranked list of products to recommend to the user based on their past purchase behavior/product interest, which you can deploy for a wide variety of marketing and analytical use cases. Each product has a score associated with each other product – the higher the score, the stronger the relationship between the two products. You can then use the recommended product and accompanying score in segmentation or personalization for a wide variety of marketing use cases.
The HISTORICAL_WINDOW determines what transactions are included to calculate the relationships between the products. To include a transaction in the HISTORICAL_WINDOW, the purchasing customer must have made two or more purchases within the last 365 days.
Requirements
- Past purchase behavior/product interest
- Data Input Schemas for Transaction/conversion data
Results
This model delivers back a long form table that contains a row for each product pairing, with a score that represents the strength of the product pairing. Each product is has a score with every other product.
Schema
| MG Field Name | Type | Range | Description |
|---|---|---|---|
| RecipientId | VARCHAR | Unique customer identifier. | |
| MLRecPurchased | VARCHAR | Product the customer purchased. | |
| MLRecProduct | VARCHAR | Product recommended for the customer. | |
| MLRecScore | FLOAT | 1-100 normalized output of scoring | Metric that indicates the strength of the relationship between the two products – the higher the score, the more likely the customer is to purchase it. |
| MLRecLow | 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. |
| MLRecMed | BOOL | 0 or 1 | |
| MLRecHigh | BOOL | 0 or 1 |