Skip to main content

How to analyze prediction results in a report

Updated

Use a simple report to see how your prediction values are distributed across customers.

This example uses an Optimal send time prediction. However, you can analyze any other prediction in the same way. Prediction results are stored as customer attributes, allowing you to use them in reports, filters, segmentations, and scenarios.

Example: Optimal send time prediction

Start with a saved prediction in Analyses > Predictions, like the one in Figure 1.

Figure 1 shows the example OST prediction setup:

Figure 1: Optimal send time prediction example used for the report.

Find the prediction in the picker

When you add a row or filter to a report, open the picker and go to Predictions. Search for the prediction by name, then select it.

Figure 2 shows where to find the prediction in the picker:

If you do not see the prediction, make sure it is in the same initiative as the report, or set it to global.

Create the report

  1. Go to Analyses > Reports.

  2. Create a new report.

  3. In Metrics, add count(customer).

  4. In Drill down > Rows, open the picker, go to Predictions, and select your OST prediction.

  5. Set the time range to Lifetime.

  6. Preview and save the report.

Figure 3 shows the report setup:

Figure 3: Report setup with count(customer) as the metric and OST prediction as the row drill-down.

Read the results

The report groups customers according to their predicted send hour.

  • Each row shows one predicted hour.

  • The value in Count(Customer) shows how many customers fall into that hour.

  • Use this view to see where most customers are concentrated.

In the example in Figure 4, most customers are grouped under hour 9 because the example project contains only a small number of tracked campaign events. In this situation, the prediction falls back to the default send time.

Figure 4: Report output showing the distribution of customers across predicted send hours.

Customize the report

You can adjust the report to fit your use case.

  • Add report filters to limit the report to a specific audience.

  • Add additional metrics to compare prediction values with customer behavior.

  • Change the time range to focus on a shorter period.

For example, you can:

  • Add a customer filter for customers from one country.

  • Add a customer filter for customers in one segment.

  • Add a Sum of purchases metric.

  • Add a Sum of revenue metric.

  • Add a Count of events metric for a specific campaign or channel.

Start with the basic version of the report. Then add one change at a time and observe how the results change.

Related resources