The Fibonacci scale in estimation

What is wrong with an even scale

Give a team a scale from one to ten, and it will start arguing whether an item is a seven or an eight. That argument has no answer: the difference between seven and eight is smaller than the precision you can get for a large item at all.

An even scale promises accuracy that does not exist. The bigger the work, the more unknowns it has, so the step between values should grow too.

What the gaps give you

Fibonacci numbers grow faster and faster: 1, 2, 3, 5, 8, 13, 21. Between two and three the difference is one, between eight and thirteen it is five. Neighbouring cards always differ enough for the choice to matter.

The scale also shows what the discussion is really about. “Five or eight” actually asks whether the item has another chunk of work in it. “Seven or eight” asks nothing.

Finally, it is faster. There are few options, so choosing takes seconds. This matters: if everyone takes a long time to pick a card, rounds eat up the whole meeting.

What the deck looks like

The classic Mike Cohn deck is 0, 1, 2, 3, 5, 8, 13, 20, 40, 100, plus “?” and a coffee cup. The numbers after 13 are not strictly Fibonacci. They were rounded, because at those sizes there is no precision anyway.

Many teams cut the deck at eight or thirteen, which makes sense. A “40” card only tells you the item needs splitting. Zero is kept for work done in passing or work that takes a couple of minutes.

• Zero: almost nothing to do.
• 1, 2, 3: small and medium items whose size is clear right away.
• 5, 8: large items, usually made of two or three parts.
• 13 and above: the item needs splitting.
• A special card for “I don’t know” or “sizing this is pointless”. It is better than a random number.

What to do with an item bigger than eight

Split it. Split by outcome: which part can be finished and shown on its own. Splitting by layer does not work: backend and frontend cannot ship separately, they are one item.

If an item cannot be split at all, it is probably not understood yet. Then it needs clarifying first. Half an hour of discussion can save a sprint.

Other scales

Fibonacci is not the only option. All working scales follow the same idea: the step grows with the size.

• T-shirt sizes (XS, S, M, L, XL). For teams that want no numbers and do not track velocity.
• Powers of two (1, 2, 4, 8, 16). The step grows even faster than with Fibonacci.
• Only 1, 2, 3. For teams that split everything large in advance.
• Counting items instead of points. All items are roughly the same size, and you count how many are done.

The “I don’t know” card

The classic deck has “?” for this, meaning “I don’t have enough information”. If that card is on the table, the round did not work: answer the questions first.

There is another case, when the size of an item does not matter. It will be done now anyway, and arguing whether it is a one or a three takes longer than the item itself. Our deck has the “poher” card for this. It says out loud what teams already say to each other.