Every planning poker deck skips numbers. There is a 5 and an 8 but no 6 or 7, a 13 and a 21 but nothing between. That is not decoration — the gaps are doing the work.
The sequence itself
The Fibonacci sequence starts at 1 and 1, and each subsequent number is the sum of the two before it: 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89. Planning poker drops the duplicate 1 and usually stops somewhere around 21 or 34, because anything larger has stopped being an estimate and become a warning.
- 1
- 2
- 3
- 5
- 8
- 13
- 21
- 34
Notice what happens to the spacing. Between 1 and 2 the gap is one. Between 13 and 21 it is eight. As a proportion the jump settles at roughly 1.6× each step — the golden ratio, which is where the sequence's reputation comes from. But the mathematics is not why teams use it. The useful property is much more mundane.
Uncertainty grows with size
A team can size a half-day task quite precisely. They have done that kind of work dozens of times, they know what it touches, and there is very little room for surprise. Ask the same team about a three-week task and the honest answer is a range: somewhere between two and five weeks, depending on what they find.
A linear 1-to-10 scale ignores this. It offers the same granularity at the top as at the bottom, which quietly invites the team to claim a precision they do not have. If the options are 6 and 7, someone will argue for 7, and that argument is pure theatre — the real uncertainty on a story that size is several points wide.
Fibonacci encodes the uncertainty into the scale. Fine steps where the team knows what it is doing, coarse steps where it does not. The scale stops pretending.
The gaps mean a vote is always a real decision. There is no comfortable middle to drift toward, so "is this a 5 or an 8?" forces someone to actually say why.
It ends arguments that were never worth having
This is the effect teams notice first. On a linear scale, two people one point apart will discuss it, because a one-point difference feels like something to resolve. On a Fibonacci scale the nearest disagreement is 60% apart, which means it is either substantive or it is not worth a second round.
The result is a session that moves. Teams routinely report cutting estimation meetings in half after switching, not because the technique is faster per vote but because a whole class of pointless debate simply has nowhere to happen.
What each number is telling you
| Card | What the team is usually saying |
|---|---|
| 1 | Trivial. Understood completely, touches one thing, no meaningful risk. |
| 2–3 | Routine. We have done this shape of work before and know where it goes. |
| 5 | A real piece of work. Several parts, but nothing unfamiliar. |
| 8 | Substantial. Worth checking whether it wants splitting. |
| 13 | Large and probably not fully understood. Split it before committing. |
| 21+ | Not an estimate. This is an epic wearing a story's clothes. |
The top of the scale is best read as a flag rather than a size. A 21 is the team saying "we cannot see the bottom of this", and the correct response is to break it up or spike it, not to put it in the sprint and hope.
Modified Fibonacci and other variants
Plenty of teams do not use pure Fibonacci. The most common variant rounds the upper values for readability — 1, 2, 3, 5, 8, 13, 20, 40, 100 — on the reasonable grounds that the difference between 21 and 20 carries no information at that size. Others add ½ at the bottom for work that is barely worth tracking.
Powers of two (1, 2, 4, 8, 16) achieve a similar widening effect and some teams prefer them. So does t-shirt sizing, which drops numbers altogether. All of these work. What does not work is a linear scale with fine gradations at the top, because that is the one property the whole approach exists to avoid.
The one rule worth enforcing: pick a scale and keep it. Velocity only means anything if the unit is stable, and a team that changes scale mid-quarter has thrown away its own history. GiveItPoints ships Fibonacci by default and lets a facilitator define a custom scale per team when the default is not the right fit.
Common questions
Why is there no 6 or 7?
Because at that size the team's real uncertainty is wider than one point, so the distinction is imaginary. Removing the option removes an argument that could not have produced a better estimate.
Should we include 0?
Include it if you have work that is genuinely already done or too small to track. Watch that it does not become a dumping ground — a backlog full of zeroes is a backlog whose velocity no longer reflects the work being done.
How high should the deck go?
Most teams stop at 21 or 34. Anything above that is better handled by splitting the story. If your deck goes to 89 and people use it, that is worth a conversation about backlog refinement rather than about the scale.