Estimation of Features
By “estimation of features”, we are talking about estimating effort and/or time—not value.
A feature estimate is a guess - nothing more. For digitally-defined systems, you don’t know how long a non-trivial feature will take. An off-the-cuff guess is as accurate as a detailed careful estimate. This is because one digital system is so different from another: one never designs the same thing twice.
Each feature is a new experience. Constructing a digital feature is not a repeatable process like painting a wall or manual piecework. It is a creative process, and there are unknowns lurking. A huge part of digital development is fixing the design to address unexpected issues as they become known. It is literally like someone asking you how long it will take you to traverse a maze that you have never been through before.
Why estimate features? The only possible reason to estimate a feature is so that,
Someone who depends on the feature can decide if they are likely to have it in time.
A manager can “roll up” the estimates and project forward, to predict milestone slip or cost overrun, which might enable them to start a conversation about that.
However, feature estimates can be misused, by:
Treating an estimate as a commitment.
Using estimates to “pack” an iteration up to “capacity”, and treat that packed amount as a commitment for the iteration.
Feature estimates do not “roll up” into capability estimates.
We feel that estimates are very useful for capabilities, even though they are likely to be extremely incorrect. They are somewhat useful for features, even though they are also likely to be extremely incorrect. Their usefulness is based on 1 and 2 above. The dangers of misuse need to be mitigated through behavior and culture.
One thing that is certain is that feature estimates do not “roll up” into capability estimates. They are all guesses: you can’t add them. If you try to add them up, there are usually missing pieces that are not accounted for. Intuitive guesses by an experienced person account for those missing pieces; rollups don’t.
Related Topics
Defining a Lean Experiment
Managing Spend Rate, Not Cost
Quick Links
Identify the optimal sequence of capabilities to demonstrate or release
Identify key intersection points, critical paths, and integration strategies
Decompose each capability into a set of features to be concurrently developed
Start test marketing them as MVPs are produced, and feed results back to adjust visions