Data Tables in the Predictions Platform

CrossEngage's Customer Prediction Platform uses 3 types of Data:

  • Customer Data

  • Transaction Data

  • Activity Data (Optional)

Customer Data

The Customer Table contains details of your Customers. This table has two mandatory fields: customer_id (string) and customer_since (date). You can also add additional data present in your Customer records, which can help improve accuracy of your Models.

The customer_id field is the primary key of this table. Each data record must have a Unique value for this field.

Transaction Data

The Transaction Table contains details of Transactions made by Customers, such as Purchases and Returns. The Transaction table has 7 Mandatory fields, as this data is required for understanding the long-term Value of every Customer for better predictions.

The primary key for this table is the combination of 4 fields: customer_id, invoice_id, transaction_timestamp and item_number. In other words, for every transaction made by a customer at a given time, each record should contain a Unique item.

Activity Data

To improve model and forecasting quality, you can upload additional tables that contain specific activities and interactions with your customers. These tables must each contain at least a unique customer ID, a timestamp, and an activity type. Additionally, an activity_id can be passed that links the activity to a specific action. However, this data is not required for an initial well working model and is therefore optional.

You can upload one or more of these tables:

  • Online activities (customer in login area on your website / click in email)

  • Inbound activities (customer calls call center, letters and emails from customer)

  • Outbound activities (print mailings, catalogs, calls from a call center)

  • Payment activities (customer pays an invoice, receives a dunning level, is referred to collections)

In order to make use of these activities in the modelling process, it is important that they are stored in the so-called event format (just like transactions). This means that each row in an activity table describes exactly one 'event' at a time with a customer.

Last updated