## What units should you estimate in?

We’ve been discussing units in our team. For several years we’ve been estimating in the classic planning poker Fibonacci units – 1, 2, 3, 5, 8, 13, 20, 30, 50… It was starting to feel a bit silly. These are estimates, yet numbers like 8 and 13 imply a level of precision which isn’t there and isn’t intended. We’re also seeing people breaking down a larger story and worrying if the smaller stories add up to 11 or 14 when the original was a 13.

I was talking with a colleague about this, and we started off thinking about changing the numbers. If we pick round numbers like 1, 2, 5, 10, 20, 50 that no longer looks so precise. (I’ve seen someone describe these values as currency estimating.)

We talked about this a bit. That gets round the precision problem, but doesn’t address people trying to make the numbers match, and relating them to real units (ideal days) rather than concentrating on the relative sizes.

So instead we’ve started using t-shirt sizing for our estimates. Most things in an iteration should be small or medium. If it’s large it will probably take the whole team most of the iteration. XL needs breaking down. Huge we haven’t a clue.

Since doing this we’ve noticed the discussions have changed. We used to concentrate on the numbers. Now it is much more about “is this story the same as that one? Larger? Smaller?” Which is back to what it should be.