Kurt's Blog

November 1, 2012

The uncertainty implied by planning poker cards

Filed under: agile, scrum, Software Development, xp — Tags: , — Kurt Häusler @ 5:24 pm

After hearing about some experiments in using planning poker cards to come up with a estimate range rather than a single estimate, it occurred to me that planning poker cards are categories rather than accurate measurements, and so have ranges built in. I was curious about what sort of uncertainty was implied by each card. So here is a table that summarizes my findings:

Card From To Midpoint Uncertainty

0

0

0.5

0.25

100%

1

0.5

1.5

1

50%

2

1.5

2.5

2

25%

3

2.5

4

3.25

23%

5

4

6.5

5.25

24%

8

6.5

10.5

8.5

24%

13

10.5

16.5

13.5

22%

20

16.5

30

23.25

29%

40

30

70

50

40%

100

70

100

85

18%

 
Some decks are different with different values, but this is what I have here at the moment.

Effectively, when we play a 5, we are saying “I think it is about a 5.25±24%.

Playing a 0 expresses the most uncertainty with the estimation. 0.25±100%

Also the 1 has 50% uncertainty! You would expect the 0 and 1 cards to have the least uncertainty. Or at least you might assume that when expressed as a percentage the uncertainty should be roughly the same over all cards.

Cards 2 to 20 all have uncertainties in the 22-29% range, which seems appropriate. 40% uncertainty for the 40 card makes sense too I guess.

The 100 card is a special case. I am just assuming that 100 is the upper bound here. You could assume ∞ if you wanted, but I don’t know how to calculate the uncertainty in that case.

I guess it means we have a range built in. There is no sense playing two rounds for a separate lower and upper bound. Or perhaps taking the lowest card in each round, and the highest card in each round and adding them up.

If we are estimating a larger project we can use the midpoints and uncertainties to calculate a range if required, using normal rules for calculating uncertainties.

If we have for example 5 stories, a 1, a 2, a 3, a 20 and a 40. That adds up to 66. The lower bounds add up to 51 and the upper bounds add up to 108.

When you add them up though, the 66 is not relevant. The values on the cards were only names for categories. There is no category 66. The estimation total is actually something more like 79.5±36%

You have to be careful with planning poker cards. The number on the front is basically an arbitrary label for a category of values that fall within a range, and that label doesn’t correspond to the midpoint.

If you are adding them you should really be adding the midpoints, not the category labels, and preserving the uncertainties.

Advertisements

2 Comments »

  1. Nice piece of analysis, I hadn’t looked at it that way before

    Comment by allan kelly — November 11, 2012 @ 8:16 pm

  2. Inetersting. However the story point is usually arrived at by a team and is a consensual number. So this may skew the percentages given above. That is if in the final round the team members show cards with 8 and 13, and settle on 8 (rather than 13) then the percentage may be calculated as (13-8)/8 * 100 , ie 62.5!

    So are you barking up the wrong tree? or am I simply barking?

    cheers

    Comment by srinivas C — February 10, 2013 @ 4:43 pm


RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Create a free website or blog at WordPress.com.

%d bloggers like this: