Estimates do more harm than good

We ditched estimates completely at tutti.ch in 2019, and I haven’t used them ever since. Instead, we did “right-sizing”.

Most of the industry is doing some form of agile now, yet only 7% of software projects are on time and on budget.

So why do we obsess over perfectly predicting what we can do in 2 weeks?

Estimating complexity, not time, does not make it better.
Even after attending all the agile trainings, the notion of fitting tasks ordered by complexity into fixed-time cycles still does not make sense to me.

Fibonacci has good intentions, but leads to weird outcomes:
“How come this is an 8? Well, it’s bigger than a 5 and smaller than a 13”. Yuk.

Planning poker feels silly because it is silly.
It’s like a corporate advertisement from the 90s: “Here’s how grown-ups have fun at the office.” 📺

So, what should you do instead?

You fit work to time instead of estimating time for work.

Instead of asking “how long does it take?” you should ask “can we ship this in X days?”. If the answer is “no”, you break it down.

X is a number that you agree on with your team. I recommend starting with 3 days. At tutti.ch our limit was 1 day, which was too short in retrospect.

“We all know that estimates will be wrong, but the point of our estimates is not to be accurate but to have the right conversations and reach a common understanding.”
Maarten Dalmijn

Plus, if all tasks are similar in size, it makes planning easy.

I’m thankful to people like Maarten for normalizing this conversation. It’s about time we questioned these sacred cows.

If you want to read more on this topic, I recommend his article on Roman Estimation