Fibonacci Series = Accurate Agile Estimates?

Episode 123 - 20 Jun 2018

Download your FREE Agile Estimating Cheat Sheet

Does using the Fibonacci Series for estimating improve the accuracy of those estimates? That's the question we're going be looking at today

Estimating is hard

If you've been following along with this series on agile estimating you'll have got the message by now that estimating is hard.

But some estimates are much easier then others. We've talked previously about estimates of absolute size versus estimates of relative size.

"How heavy is this?" is a hard question. "Which of these is heavier?" is an easy question. Why is it an easy question? Well, it's because there's a large difference you assume in the weight of these two.

When I say difference is that the absolute difference or the relative difference?

Which of these two coins is heavier? Which of these two bridges is heavier? I hope that answers the question. It's not the absolute difference that's important, it's the relative difference.

The Fibonacci Series

Hold on a second. Wasn't this supposed to be an episode on the Fibonacci Series?

I think it's time we rolled it in. Actually, let's build it from scratch.

  • The first two numbers are zero and one.
    • To get the third we add the first two together: Zero plus one is one.

We carry on adding pairs of numbers:

  • 1 + 1 = 2
  • 1 + 2 = 3
  • 2 + 3 = 5

and so on:

  • 8
  • 13
  • 24
  • 34
  • 55
  • 89

It's in the spaces

What's interesting in this series is the gaps between the numbers. Not the absolute gaps: the relative gaps.

  • 0, 1 - The relative gap between these two is, oh, that's infinite. Yeah. That ones a little bit large. Let's move on.
  • 1, 1 - the relative gap is zero
  • 1, 2 - The relative gap between these two is 100%. Okay.
  • 2, 3 - 50%.
  • 3,5 - 66.6666667%
  • 5, 13 61 and a bit percent.
  • 13, 21 almost 62%.
  • 21, 34 - 61.8%.
  • 34, 55 -61.8%

After some craziness at the beginning of the series the relative gap between he member of the series settles down to around 61 and a bit percent.

Let me see if I can demonstrate that to you a little more visually. Here are the first few. Zero, one, one, two, three, five, ah, yeah now we're up to space. I'm going to zoom out around about 60%. There's the eight. Zoom out another 60% there's the 13. Zoom out again, 21. Zoom out once more, 34. Zoom out again, 55. Zoom out one last time, 89.

I hope you can see that although the bars are getting skinnier as we zoom further and further out the relative size between this one and this one stays pretty well constant.

Fibonacci for estimating

The reason that this scale works so well for estimating is that it encourages us to stay in the realm of easy estimates. It encourages us to stay with relative estimates. To say in slightly different terms if we are estimating two things and their sizes, their relative sizes are not sufficiently different then we consider that they both have the same size, which brings us right back to the question that we started with today.

"Does the Fibonacci Series lead to more accurate estimates?"

I think the answer has to be no. If anything what it does is protect us from attempting to make accurate estimates. It keeps us in a realm of making rough or broad estimates.