jump to navigation

Planning Poker and Blink February 8, 2007

Posted by James Webster in : development , trackback

Planning poker is a technique that I have used a few times on agile projects. I’m not sure who it is originally attributed to under this name but it apparently has its roots in the Rand Corporation’s wideband delphi estimation technique.

The idea is that an agile team estimates the effort required for a group of stories by following this process:

  1. Create a deck of around 5 estimating cards for each ‘player’ using standard index cards. Write the same series of numbers on each deck, most teams use a fibonacci sequence; 1, 2, 3, 5, 8.
  2. A story is discussed until all players are convinced they have enough information to estimate the effort required.
  3. Each player places face-down on the table the card from their deck that represents their feel of how much effort is in the story. Playing card 1 would indicate the player thinks the story is pretty simple, card 8 would suggest that it is much more complex.
  4. If the estimates are uniform, return to step 2 and the next story.
  5. If not, discuss the story further to find out the reasons for divergence and re-estimate, returning to step 3 until consensus is achieved.

It looks like Mountain Goat Software have set up www.planningpoker.com to facilitate this process for distributed teams, or even teams sitting in the same room with each team member playing on their own laptop. I’m not so sure that I would be comfortable entering my story details into a website for my latest top secret agile plan to take over the world, but it would be great to see something like this integrated into Buildix.

Recently I have been reading Malcolm Gladwell’s Blink (or rather listening to Gladwell himself read the book). For a while my position on the planning poker process is that step 5 is potentially damaging to the overall quality of the estimates and thanks to Blink I now have some support for my argument. The quick discussion and subsequent rapid estimation of steps 3 and 4 are, I believe, representative of the thin-slicing technique that is so effective in letting our subconscious minds tap into our experience to come up with surprisingly accurate results. However step 5 moves us towards thick slicing, or overloading ourselves with additional information/discussion that does not fruitfully contribute to our ability to cut through to the core of the issue. I have also seen that this step sometimes turns into a power game, albeit usually unintentionally, as those that have estimated significantly higher or lower than the main consensus are persuaded to bring their estimate back in line.

I feel that simply coming up with the average of all estimates given at step 4, which to be fair might be a number like 3.52 so some sane rounding should be applied, would result in more accurate consensus estimates without prolonging the estimation process. Perhaps www.planningpoker.com should allow a ‘blind’ option to be selected on each game, to permit people to make their estimate whilst hiding it from everyone with only the average estimate being revealed?

Comments»

no comments yet - be the first?