Oct. 1, 2022

Ep. 3: Why software estimation sucks

Ep. 3: Why software estimation sucks

For today’s episode, Asim Razzaq takes on the topic of software estimation and how engineers can improve the quality of estimates through calibration and confidence intervals. Most engineers in companies are overestimating, and in some cases, underestimating, their software estimates. With the help of Doug Hubbard’s research, Asim can share how software engineers can train themselves for a more accurate estimation rate.

[01:19] Software Estimation

[02:02] Ways to Train Estimation

[02:45] Challenges Often Faced on Estimation

[05:37] Asim’s Experience

[09:39] Training From Doug Hubbard’s Book

[10:45] Solutions Based on Training


Missing the Mark

There are many ways to miss the points when it comes to software estimation. This varies depending on the person’s confidence level and mindset towards the estimation and creates their own bias. Some cases occur based on underestimation, overestimation, and similarities-based results. Thankfully, each error has a training method that applies to help better hit the right mark.


Practice Makes Perfect

Luckily for software engineers, estimation can be trained and be improved overtime. With the right tools, you can conquer the constant overestimation and underestimation results. Asim based his techniques on Doug Hubbard’s teaching materials. Here, he managed to showcase an estimation that was a lot more accurate than initial estimation on the population of Italy.


Got questions or suggestions for future episodes? Contact me.


Additional Resource Materials:

Doug Hubbard’s Website

The Credence Calibration Game (Andrew Critch)


--- Send in a voice message: https://anchor.fm/always-an-engineer/message