Hacker News Reader

I got pretty decent at software estimation at my last job.

I would spend an entire day "pre-programming" everything in my head, estimating the length of each little chunk, adding them up, then multiply by ~2.

It worked for me. But I still would never trust the estimates.

View on HN

Instead of 2: Multiply by π for new kind of projects. Multiply by φ for known projects.

Another angle.

'How long did a project this size take last time?'.

As an aside, years ago I worked at a company that did thorough (and inaccurate) bottom-up schedules. I got dinged for not using quarter hour accuracy in the various task estimates.

Reply on news.ycombinator.com