Estimate with your data, not your gut

Published by on July 20, 2013.

People are bad estimators. There is a really great overview on Quora about how and why we get estimations wrong. At, we don’t estimate in days, but rather let the app tell us when we’ll be done.

Our saving grace against bad estimations is that people tend to work consistently over a long enough time frame. This is especially true across teams of people. Given a team and a series of quarters, we tend to deliver the same amount of work over the same amount of time. Sometimes I’ll be doing more work and Dave is off at a conference or two. A few months later, I become distracted with purchasing a house, so my productivity slips. These things are tumultuous in the short term, but they even out in the long term.

Using the “people are consistent in big groups over time” approach, we can use software to determine when things will likely be finished. We do this by calculating a rolling average of work that the team has finished over the past 3 weeks. We distill this into a rough estimate in the timelines view. We show you which tickets the data suggests you’ll finish at 7, 14 and 21 days out. These are powered by estimations of complexity (aka the t-shirt sizes) that each card has. The more tickets that are estimated, the better the predictions will be.

Using the data available to us, we can make predictions with some degree of confidence. In short, we’re likely to ship as much this week as we have across the past few weeks. That’s why we’ve dropped gut-based estimations in favor of data-backed prediction.