How Good are our estimates? Should we care?

Our team decided against estimating stories or tasks in real time (eg hours) during sprint planning. We assign relative points during backlog grooming and may change that number later or during the sprint planning as more information becomes available. I was wondering: How good are our estimates? How do we measure that? How do we improve? Do we care?

The following 20 minute (!) retro gave us some insight into our estimates over the last three sprints. I provided the team with a list of the stories we played in the three sprints and asked them to point them based on the actual effort made. The team agreed on all but three of the stories. The points for these stories were decided based on a quick discussion.

When comparing with the original estimates we were less concerned about stories shifting one point bracket. Often a story falls between one bracket or the other. And, estimate are not predictions. We may find simpler approaches or new requirements during execution. Stories that shifted 2 brackets are of more interest. This may indicate a more fundamental error in the estimate.

We found that 5 stories, about 10% of the total had big point shifts. When we looked at these individual stories we found that they all were “review” stories. In the review stories the team looked at the implementation of certain aspects of the service developed by another team. We learned that these efforts are difficult to estimate. We overestimated / underestimated a significant number of those stories. All other story estimates seemed close enough.

Another observation was that our average estimate went up by 4%. We did not do anything with that information for now. This will be more interesting when we do this again in a few months time.

We learned two things within 20 minutes

    • Review stories carry high estimation risk.
  • Our estimates are good enough. There is no need or value in investing in improvement

Easy Tool To Speed Up the Retro

We managed to do this exercise very quickly by using a google sheet I prepared in advance. Google sheets is great for collaboration, everyone can work on the same sheet in real time. Send me a message if you would like to have a copy of this sheet.

The sheet uses links to keep everything together. To set up the sheet you just need to add the stories and story points to columns A and B of the “Original Point Estimates” sheet. Keep the number small. I suggest 50 stories or less.

Before the retro I recommend you hide the x and y sheets to avoid confusion. The team can hit an X for the assigned points in the columns C-F of “Retro Estimates”. The estimate will show in column B. If two team members selected different estimate the cell will light up red.

It took the teams less than ten minutes to point these stories they recently worked on. When finished we looked at the comparison sheet. It shows the summary numbers and all stories with the difference marked in blue if the points shifted two buckets or more

Reach out to me if you want a copy if the template.

  • The images are examples and not the exact data from our test
  • Feature Image reproduced from http://agilelucero.com/scrum/who-does-in-scrum-estimate

The opinions represented in this blog are my own, and not that of my employer or the  organizations that I work with.