Konubinix' opinionated web of thoughts

Tragedy of Guestimations in Softwares


When we create some piece of software, the new work is about adapting it to new use cases or creating another piece of software dedicated to something else.

We almost never create twice the same thing.

I decided to believe that guestimates only work for work you already did. Based on past experience, you can infer how much time it will make.

But, because you almost never had a similar past experience, estimation are almost always pretty useless.

On the other hand, they are needed for strategic decision and commitment making. Most clients want to have a promise about the work done and when it will be done.

There are attempt to palliate this lack of estimation power, like agile methods, that need the client to be always synced with the work done. I think those are elegant ways to substitute estimation making by continuous collaborative learning and production with the stakeholder (tragédie de la définition d’un méthode).

Yet, most client (that I know of) and most people I work with don’t get that point and would rather have useless estimation (while still believing in their predictive power1 ) than changing their commitment habit for a domain that needs it.

I believe that the best I can do it acknowledge this and decide not to feel bad about it (dichotomy of control).

Notes linking here