We are all used to just implement the well-known “best practices”, not even thinking much about the reasons they were created. Especially in Agile. We like all those rituals, as they make us fill as we are going the right direction. Just follow the rituals and you’ll be ok.PMs/Scrum Masters usually hate all these clever questions from the sarcastic developers, as the only thing they can answer is – “It’s obvious, stupid!”.

You can’t even explain, why 🙂 I want to write several posts about the most frequent questions managers can’t answer, so you are prepared and next time your team mentions it – you brains can shine in full managerial glory. Also I will use levels of the answers complexity, so you can always adjust them to your current audience.

### Question: Why Fibonacci?

When we estimate user stories during the planning poker or any other estimation activity – we usually use some sequences. The most popular are powers of two – 2,4,8,16,32 and Fibonacci sequence – 1,2,3,5,8,13,21. We are all used to the powers of two in our everyday IT work, but what about Fibonacci? Why we use it? Is it better than other sequences?

### Level 1

**Answer:** “Because it is cool, bro!”

**When to use:** When yo are too cool to answer stupid questions. They know you know the answer. But you have no time to spend on such a thing! Also it can work when having beer with your team in a pub.

**Benefit: +10 to manager’s charisma**

### Level 2:

**Answer: **To be honest, any exponential sequence would work. The further we go from the 1 when we estimate the user story – the more uncertainty is included in our estimation. So with exponential sequence the larger numbers present this increasing level of uncertainty, also reminding you to break down the stories. Because what does estimation 64 means? I think, nothing except from the fact that we know almost nothing about how to implement this story.

Also the increasing gaps between the estimations helps the team to understand the same fact.

To understand that better just imagine that you have the user stories estimated 1,3,5,8,21. What is the difference between 3 and 5? And what is the difference between 3 and 8, 8 and 21? Our brains see the perceptive difference better between 5 and 8 that between 5 and 6.

Exponential sequence just force us to make a choice between the less/more uncertain user stories.

**When to use:** During retrospective, when you would like to look smart.

**Benefit:** +10 to managerial smart ass

### Level 3:

**Answer: **You may be surprised, but there is also some math behind all of that. So you can provide the strongly math backgrounded answer.

The answer of Level 2 seems to be very logical, it even has the reference to the “how you brains work”, but there were no proven statements or axioms. The best math prove of the progressive estimation success was described in the post by Alex Yakyma here. I will try to explain that in more simple words.

He uses the same terms – level of uncertainty, but applies the Information theory to it.

Pre-set data:

Let’s assume we have the user story **U** and we assume(our estimation) it is not bigger than number of points **L**. But in reality it can appear to be from 0 to **L**. Let’s assume we have the estimation technique which allows us to estimate U with the boundaries +- which is presented by **P**.

Let’s assume we have 2 experiments – **B**, where we find the accurate estimation of user story **U**. Experiment **A** is applying the estimation technique we have and reducing the uncertainty.

Information which is enclosed in experiment **A** regarding experiment **B** can be expressed as follows: **I(A,B) = H(B) – HA(B)**. Here **H(B)** is the level of uncertainty of experiment** B**, **HA(B)** – level of uncertainty of experiment **B** after the experiment **A** takes place.

Then Alex uses the Shannon’s formula (tells the maximum rate at which information can be transmitted over a communications channel of a specified bandwidth in the presence of noise) to get the information we obtain from the estimation:

**I(A,B) = log(L/P)**

Usually the log base used is 2.

Now we are ready for the magic:

*X axis represents L/P, Y axis represents information in bits. Graph shows the information we get as a result of the estimation process.*

… the information we are getting from the estimation grows much slower than the precision of estimation.

Now from the graph of logarithm function below we can see why “little estimation effort helps a lot and big estimation effort helps little”.

So you can see now how the exponential estimation scale adds the information that is valuable faster.

**When to use:** When you feel rather confident saying “logarithm” and “approximation”.

**Benefit:** +100 to team’s respect, +100 to manager’s brains size

Let me know what questions you have difficulties answering to? Any of them. Even “why my boss is an idiot?” will work) I will try to help you with that.

Good post again. I hate the fact that people start doing rituals without knowing why.

One thing though, don’t just lump PMs and Scrum Masters together like that please: “PMs/Scrum Masters usually hate all these clever questions from the sarcastic developers, as the only thing they can answer is –”.

1) PMs make the worst SMs

2) It’s my job as SM to answer these clever questions from excellent developers (they should ask these questions)

3) Any SM not able to answer this question is not worth the title. I am a SM and my answer to this question would be:

That we humans are really bad at absolutes, but better at relatives as studies have shown. So we are better at observing that a seed is smaller than an orange. That is why we use the Fibonacci sequence, as it is based on this knowledge.

Having said that, if anybody asks about the EXACT science I can refer them to your post in the future. 😉 So thank you!

> One thing though, don’t just lump PMs and Scrum Masters together like that please:

That is why I used them together, it was better to write – PMs who wear the hat of SM))) When the companies I worked in started implementing Scrum they were a bit confused what to do with all these project managers) So mostly they became scrum masters. But I am for the team members to become the scrum master – I am moving towards that)

Great answer to the question, by the way!

Ah I see. Carry on. 🙂 And, thanks!

I arrived at the Fibonacci sequence ‘experimentally’ (I guessed.) I accepted that uncertainty increases with story size, tried doubling and found that to be an overestimate and backed off a bit 🙂

Level 3 is a fascinating idea. I’m reading ‘The Information’ by James Gleich (the ‘Chaos’ guy.) I wanted to write a book about information but he got there first and made my job much harder :-). He talks about Shannon quite a lot, though I think ‘Information Theory’ is really ‘Data Theory’. Shannon believed that ‘data’ is a measure of unexpectedness, or uncertainty, so the idea that his formula might fit seems perfectly valid.

Thank you very much for this insight.

Seems like a case of someone bored with standard exponential sequences, coming with Fibonacci and looking for mathematical rule to emphasize it’s better. I’d like to see real life examples to show it works better than the already classic exponential ways. At this point in time I see it as a additional option, but somewhere down the line the numbers will be off the scale with Fibonacci s well.

And why include the Scrum Master acting as PM or PM acting as the Scrum Master discussion here? Seems like noise making it harder to bring the real message across as the real reason to work on ways to improve estimates has nothing to do with this discussion. If you wish to address the PM / Scrum Master topic, then this won’t help a bit either.

What are the standard exponential sequences? Did you read Shannon’s Information theory? As sequences get bigger, our ability to guess accurately becomes increasingly uncertain, in accordance with the Fibonacci sequence. A squared rule, for example, would over-estimate uncertainty, on average. It doesn’t seem very likely that the ratio between the radius and the circumference of a circle would be a number as weird As pi, but it is.