What does predictive and descriptive metrics mean?

We can divide all metrics into two broad categories.

  • Descriptive metrics. We use descriptive metrics to understand the current state of the community. Typical examples of metrics that describe the state are the number of monthly engaged users, page views, site satisfaction, etc.

  • Predictive metrics. We use predictive metrics to understand what the future of the community might look like. Predictive metrics also help when we cannot run an A/B test, so you can model several what-if scenarios with different parameters of the system and compare the outcomes. When one works on predictions they need not only to be able to predict something but also be able to define what a positive answer for a metric would look like. Any forecast is a predictive metric.

Different tasks require different metrics. Most of the tasks on the individual contributor level require metrics that describe the current state of the community, whereas tasks at the executives level require predictions. The reason is that executives use metrics to plan for the future when individual contributors use metrics to direct their current actions.


This is a fragment of a draft of the book “Lessons Learned While Working On Stack Overflow”. Read the full book on kindle or the paperback version.