Relative Estimation at near zero-cost
We’ve been through the full spectrum of estimation, as I’m sure a few of you have. From Gantt charts demanded from above to zero-estimation derived from below.
We’ve come to a conclusion*: Fine-grained estimation is ultimately wasteful. Tend towards zero-estimation.
Meet the The Estimation Matrix.
- A cross-functional team
- A week
- A prioritised set of work, of roughly equal size (i.e. 2-3 days per `work item`)
- A limited work in progress (of say, 3 work items at once)
- Draw up a Matrix on a small, portable whiteboard (say 5ft x 4ft).
- X Axis: Monday, Tuesday, Wednesday, Thursday, Friday
- Y Axis: Number of `lanes` that your Work In Progress limit(s) allow
- Get the team to `best-guess` by sticking cards on this Matrix. For cards that take longer than 1 day, draw a line to when the team think they’ll finish.
- Take a photo of this board, and email it around: It’s your set of work and importantly when the team thinks they’ll have it done by.
- Put the board up in the work area for all of the team to see.
This is another simple but highly effective tactic. As with all estimation exercises, it helps the team understand the work at hand.
The team can visually score off cards and see the week progress. This is substantially less abstract than a burn-up, burn-down or cumulative flow graph. It’s simple, the team just get it.
Leave the matrix up on a 5ft x 4ft whiteboard. This lets the team see how they are progressing. The same benefit as a burn up/down chart but for 5% of the overhead.
A major benefit of planning poker is everyone has to contribute. Everyone has to pluck out a card. People can’t just sit back and follow the alpha-types. You need to balance this against the time it takes: we’ve sat through multi-hour planning poker sessions that add no value.
Substantially less empirical: This is certainly true. But as we’ll show you in a later article, you can actually get substantially more empirical by measuring more effectively.
The power of this approach comes when combining with cumulative flow and empirical evidence. With near-evenly sized work items and measuring cumulative flow – you’ll be able to predict based on evidence for near-zero estimation cost. But more on that later….
Proof is in the pudding
We’ve been using estimation matrices as a handy 5minute `best guess` for a good while now. Although we Limit Work in Progress and we believe commitment is dead we’ve found estimation matrices a good proportional investment to the estimation problem. They really help visualise the old prioritisation problem of: “OK, we can’t fit this in but you’re asking for more? What’s going to give?”
Here’s a few of our estimation matrices over the last couple of months.
*no conclusions are permanent. Given the same set of inputs, we’d arrive at the same result. If our inputs change, we might well end up at a different result.