Why Rounding Up Some Estimates Is Good

Last series Story Point Estimates Are Best Thought of As Ranges, the approach of using numbers in estimation as ranges was discussed. For example, in the modified Fibonacci sequence (1, 2, 3, 5, 8, 13, 20), the number 13 is intended to represent all estimates from 9 to 13. This approach was compared to thinking of estimates as buckets of water.

Some may be concerned that rounding up estimates could lead to inflated or padded schedules. However, this concern can be addressed by understanding why this is not typically the case.

Consider a product backlog item initially estimated at 10 points. Using the modified Fibonacci sequence and the rounding up approach, this estimate would be rounded to 13 points. If the item proves too large for a sprint and is split into smaller stories, for example, into three stories estimated at 5, 5, and 2 points, the total comes to 12 points—one less than the 13-point estimate, but two more than the original 10-point estimate.

This approach, rounding up to the next larger estimate, enhances the likelihood of accurately predicting the delivery date for the project. If the estimate had been rounded down to the next lower estimate (8), the project would be four points behind schedule. If the original estimate (10) had been used, the project would fall two points short. By rounding up, the chances of on-time delivery are improved.

While it is possible that the initial 10-point story could have turned out to be only 8 points, it might also have ended up being 14 or 15 points, or another figure entirely. As work is examined in greater detail, it often expands in scope. In the example, what initially appeared to be a 10-point story later revealed itself to be closer to 12 points, which is a common occurrence.

For teams that have been practicing agile for some time without using such "bucket" estimation, this scenario may be familiar. For instance, a sprint might show excellent progress according to the burndown chart, but some backlog items were split during the sprint, and their estimates increased. As a result, the burndown chart may not appear as favorable as expected.

Rounding up, as described, is not an attempt to pad estimates but rather a reflection of the inherent uncertainty in estimation. Adopting an estimating approach that accounts for this uncertainty can significantly contribute to success in agile practices.