Evaluate results of churn prediction

Published December 06, 2021
Modules
Difficulty
Selected Clients
eobuwie logo

You can create a dashboard to monitor churn predictions.

The scenario in this example is a continuation of Predict churn, but with some tweaks it can be used for other predictions.

Reducing churn
Reducing churn

Prerequisites


Process


In this use case, you will go through the following steps:

  1. Create metric to calculate the conversion of customers who were at different risks of churn.
  2. Create dashboard.

Create metric


In this stage, you create two metrics to calculate the conversion of customers who were at different risks of churn.

  1. Create new metric.
  2. As the metric type, select Formula metric.
  3. As the first node, add a product.buy event:
    1. Set the aggregator to Count.
    2. Set Occurrence type to All.
    3. Add a contact filter with the following event conditions:
      • Action: snr.prediction.score
      • modelId equal to the model ID of your prediction.
        Tip: The model ID can be copied from the Three-dot icon menu in the Prediction list.
      • score_label equal to Low
      Contact filter for product.buy event
      Contact filter for product.buy event
  4. As the second node, add a snr.prediction.score event:
    1. Set the aggregator to Count.
    2. Set Occurrence type to All.
    3. Set the following conditions:
      • modelId equal to the model ID of your prediction.
      • score_label equal to Low
  5. Set the mathematical operator between the nodes to division.
  6. Encase the nodes in brackets.
  7. Multiply the result by 100.
    Completed metric formula
    Completed metric formula
  8. Save the metric.
    Result: This metric shows the percentage of customers who were at low risk of churn and made a purchase.
  9. Repeat steps 1-8 to create a second metric, but replace the value of score_label with High.
    This will create a metric showing the percentage of customers who were at high risk of churn and made a purchase.

Create dashboard


In this stage, you create a dashboard that counts customers at risk of churn and shows the results of the metrics created in the previous stage. You can use this information to assess the effectiveness of churn prediction.

  1. Go to Analytics icon Analytics > Dashboards > Add dashboard.
  2. Add a segment to the dashboard:
    1. In the menu above the canvas, click Segment icon.
    2. Click the widget that appears.
    3. In the panel on right, from the Segmentation drop-down list, select Create new.
      Result: The segmentation editor opens.
    4. Add the segment with customers whose probability of churn is high:
      1. Click Choose filter and select the snr.prediction.score event.
      2. Add the following condition: modelId equals the ID of the model of the prediction you want to analyze.
        Tip: The model ID can be copied from the Three-dot icon menu in the Prediction list.
      3. Add the following condition: score_label parameter equals high.
      4. Enter a meaningful name for the segment.
    5. Add a new segment by clicking the plus icon next to the current segment.
    6. Repeat step d for the new segment, but change the condition to score_label parameter equals low.
    7. Click Save & Exit.
      Result: You are returned to the dashboard creator.
  3. Add the metrics to the dashboard:
    1. Click Metric icon.
    2. Click the widget that appears.
    3. From the Metric drop-down list, select one of the metrics that you created previously.
    4. On the Style tab, change Data format to %.
    5. Repeat steps 3a-3d to add the other metric.
  4. Click Save dashboard.
    Result: The dashboard becomes available for viewing and sharing. You can change the date range when viewing the results.

Read more


😕

We are sorry to hear that

Thank you for helping improve out documentation. If you need help or have any questions, please consider contacting support.

😉

Awesome!

Thank you for helping improve out documentation. If you need help or have any questions, please consider contacting support.

Close modal icon Placeholder alt for modal to satisfy link checker