When you run a digital customer touchpoint such as a website or mobile app, you'll want to detect and then act upon a change in customer behavior. In the past, you would probably log into your dashboard periodically to eyeball precipitous changes by inspection. As an approach, the periodic, manual check scales only to a point.
You could, instead, use Treasure Data's Workflow and a bit of machine learning capabilities. You can build a robust and tunable system that detects changes in customer engagement automatically at scale, saving time, and improving agility for analytics and marketing teams.
In this article, we assume the following:
- The workflow is designed to detect anomalies on the previous day (N-1) based on historical data up to day N-2. As such, it's recommended that you run this workflow toward the beginning of the day.
- In the example, anomalies are reported as a message on Slack via HTTP webhook. The workflow can be configured to use other notification mechanisms such as email.
How It Works
The workflow works as follows