Count Stories Instead of Points

If you use an iteration-based agile approach, you will need some idea of how much work the team can commit to for an iteration. Some teams count story points and estimate the number of points they can deliver in an iteration.

Here’s a wild and crazy idea: instead of spending time estimating larger work, consider breaking apart those large stories into small, one-day stories. Here’s why: your customers buy and use features or feature sets. Your team releases features or feature sets. Why not estimate by feature?

One of the problems with large stories is that a team creates tasks to break apart the story. Too often, the team says, “We need a front end of this many points, a back end of this many points, and testing of that many points.”

That kind of estimation is task-based estimation. Agile allows us to deliver features, which allows us to perform deliverable-based estimation, a different kind of estimation.

When we estimate deliverables, we often realize that what we are thinking about—the deliverable—is too large. We have a one-, two- or even a three-week deliverable. In Predicting the Unpredictable [Rot15], I said that the estimate can be off by the unit of estimation. That is, if you have a one-week estimate, you might be off by one week. That’s because you are not estimating the deliverable; you’re estimating work that gets you to the deliverable.

Story points can help a team learn what it can do in a certain amount of time. And I have seen teams try to convert story points to hours or days, and then fill in the tasks. Instead of deliverable-based estimation (the story), the team uses task-based estimation, interim work that might or might not be useful.

Instead of counting story points, count features or stories. Workshop the stories to be one-day stories, or decide to pair, swarm, or mob on large stories so the team maintains a high throughput and can count its stories. If your stories are larger than one day, use your cycle time for a feature to see how large the stories are.