Source: Facebook

THOUGHTS ESTIMATING CREATIVE AND SOFTWARE PROJECTS : POKER

You know the whole card game thing actually works pretty well. But it’s harder to do in some settings than others. In house dev is pretty easy. But international groups, with different skill sets, it’s pretty hard.

If you read everything out there (I have) most of it’s pretty obvious behavioral psychology. In my view I want to encourage people to invest very little time up front estimating, but to get estimates from a lot of people on the team, and narrow those estimates as we come closer to execution. The reason is that the statistical analysis of each individual’s estimate of any given work item, over time, is pretty accurate. In other words, if you estimate 100 things over three months, and the end work actuals are captured as hard data, over time, we can do a better job of predicting estimates on first blush than we can with deep analysis. Just how it is. Intuition being what it is across a group of people.

What bothers me is that in house people are pretty comfortable sitting around pondering this stuff. But in my experience, high speed high production service businesses, have an extremely hard time doing this kind of thing, partly because they have less direct influence over team members who have much more autonomy, and more varied and irregular demands.

So, the way I’ve looked at the feature in oversing, after looking at a zillion alternatives, is to set up a mixture of triangulation and card game that can be done real-time or not. The PM/DM/SM or whatever picks a range of stories / features to estimate. We handle both narrative and functional approaches. (In-house I use functional not narrative because of the workflow engine’s ability to modify all experiences.) Then launches the game at a particular time, and can do it interactively, or iteratively. Personally I absolutely love the iterative approach because I freaking hate sitting there working at the rate of the slowest bandwidth at the table.

Each item is presented for triangulation against other previous estimates (if there are any) and you vote by dragging the card into the position with the appropriate points. When done, you commit that card. At some point, the PM/DM/SM calls the hand, flips the cards, and all the players are informed about the ‘hand’.

This means I can run through thirty items at my speed, and someone else at theirs, and we don’t have to do it at the same time, and we can do it as a cognitive break from our other work.

Now, I’ve set it up to ‘award the pot’ to people who are accurate estimators. So it’s a little bit of a game. The pot is just a score, and this score is part of your profile. The better your ‘winnings’ are the better you look at estimating. It’s pretty fun really.

This kind of gamification matters because in large consulting companies people don’t know much about you and these sort of metrics build an empirical reputation for you tat’s visible on your FB-like profile.

Alternatively, you can run the game interactively, with people right in the room, or over Skype etc. The point is that you can run it either way. Or even a combination of the two. (I don’t have to be there in person to play my hands – yes I know purists. But again. I dont get to tell people how to run their projects. I just get to find ways to help them run them the way that they want to. Advocacy is your job. Enablement is mine. )

Now, when you move from the backlog to the sprint you estimate, (or at least most people do) in hours, and break the story or feature into discrete measurable tasks.

Oversing records actuals in painful detail. You sort of plan your day or week on an normal calendar, and we just assume that whatever you plan is what you did unless you change it. It’s pretty easy to reconcile your billings this way, by handling exceptions to your agenda rather than trying to remember what you did. (I’m not doing it justice here.) You don’t really fill much out on your timecard. The system does all of that for you. (I know. Cool. It’s awesome. Thank Max Romanenko. It’s his doing.)

So, we have SWAGS in points, Refined estimates from the triangulation-poker-game, hourly estimates for tasks, forecast time against them, actual time recorded. And our funnel statistics are pretty solid with that information.

That said, any brilliant insights would be appreciated.

Cheers.