New teams almost always ask the same question when they see a Planning Poker deck: "Where's the 4? And the 6?" The cards run 1, 2, 3, 5, 8, 13, 21 — a Fibonacci-like sequence where each number is roughly the sum of the two before it. The missing numbers aren't an oversight. They're the whole point.
The gaps grow on purpose
Look at the spacing. Between 1 and 2 the gap is small. Between 8 and 13 it's five. Between 21 and 34 it's thirteen. As the numbers grow, the distance between adjacent options widens. That mirrors a simple truth about estimation: the bigger and less familiar a task, the less precisely anyone can size it.
The uncertainty principle of estimation
You can tell a 1-point task from a 2-point task. You cannot reliably tell a 20-hour task from a 22-hour one. A scale with even gaps pretends a precision that doesn't exist; the Fibonacci scale refuses to.
What the missing numbers save you from
A linear scale (1, 2, 3, 4, 5, 6…) invites a specific kind of waste: teams spend ten minutes arguing whether something is a 6 or a 7. That argument is pure friction, because at that size the difference between 6 and 7 is smaller than the error bars on either estimate. The Fibonacci scale removes the option. You're left with a sharper, more useful question: is this more like a 5 or more like an 8?
- Fine resolution where it matters — small tasks (1, 2, 3) are close together, because small differences are real and detectable.
- Coarse resolution where precision is fake — large tasks (13, 21, 34) are far apart, because you genuinely can't split those hairs.
- Fewer, faster decisions — the deck offers a handful of distinct sizes, not a spectrum to haggle over.
The '13 means split it' rule
Most teams adopt a soft rule: anything that lands on 13 or above is probably too big for one sprint and should be broken down before it's committed. The large cards act less like precise estimates and more like alarms. A 21 is really the team saying "we don't understand this well enough yet" — which is far more valuable than a fake number.
A quick calibration exercise
Before your next estimation session, pick three past stories your team remembers well — a small one, a medium one, and a painful big one. Assign them 2, 5 and 13. Pin those on the wall as reference points. New estimates get compared to these three anchors instead of debated in the abstract.
What about the ? and the coffee cup?
Most decks include two non-numeric cards. The question mark means "I don't have enough information to estimate this" — a legitimate, important answer that should stop the round and trigger a conversation, not a guess. The coffee cup means "I need a break." Both exist because honest signals beat forced numbers, which is the same philosophy behind the gaps themselves.
Where the deck actually comes from
The sequence itself is ancient — Leonardo of Pisa described it in 1202 — but its use in estimation is recent. James Grenning sketched Planning Poker in 2002 as a way to stop consensus meetings from dragging, and Mike Cohn popularised it in "Agile Estimating and Planning", adding the modified deck most tools ship today: 0, ½, 1, 2, 3, 5, 8, 13, 20, 40, 100. Note the top end: 21 and 34 became 20 and 40. That was deliberate. Round numbers at the large end signal "this is a rough bucket, not arithmetic" — nobody is tempted to believe a 40 is precisely 40 of anything.
There's even a perceptual argument for why ratio-based steps feel right. The Weber–Fechner observation says humans perceive differences in proportion to magnitude: you notice a 1 kg difference when lifting 2 kg, but not when lifting 50 kg. Estimation behaves the same way — the jump from 5 to 8 feels as meaningful as the jump from 2 to 3, because each is roughly a 60% increase. A linear scale ignores how our judgment actually scales; a geometric-ish one cooperates with it.
Fibonacci vs. the other decks
Fibonacci isn't the only defensible choice, and it helps to know the trade-offs before you pick a deck for your team.
- Modified Fibonacci (0, ½, 1, 2, 3, 5, 8, 13, 20, 40, 100) — the default in most tools. The ½ catches trivial tweaks; 20/40/100 are honest "too big, split it" buckets.
- Powers of two (1, 2, 4, 8, 16, 32) — even more aggressive gap growth. Some teams like that it forces a hard double-or-not decision at every step; others find the jump from 4 to 8 too brutal for mid-sized stories.
- T-shirt sizes (XS–XL) — removes numbers entirely, which kills point-arithmetic abuse but also velocity tracking. Fine for roadmap-level sizing, coarse for sprint planning.
- Linear (1–10) — looks friendly, but reintroduces the 6-vs-7 haggling the gaps were designed to eliminate. Rarely worth it.
Common mistakes with the Fibonacci deck
The deck can't protect you from everything. Three failure modes show up again and again. First, averaging the reveal: if half the team shows 3 and half shows 8, the answer is not 5.5 — it's a conversation, because those two groups are imagining different work. Second, treating the sequence as arithmetic: two 5s do not make a 10; points are buckets, not currency. Third, letting estimates expire: a 13 sized three months ago, before you refactored the module, is stale data — re-estimate briefly rather than trusting it.
If consensus lands between two cards
When the discussion converges on "bigger than 5, smaller than 8", take the 8. Rounding up costs you a little slack; rounding down costs you a sprint commitment you can't keep. The asymmetry makes the choice easy.
Getting a skeptical team to adopt it
If your team currently estimates on a linear scale — or in hours — don't argue the theory; run an experiment. Take one refinement session and estimate the same five stories twice: once your usual way, once with the Fibonacci deck and hidden votes. Then compare not the numbers but the discussions. Teams consistently notice two things: the Fibonacci round was faster, and the disagreements it surfaced were about the work rather than about the scale. That observation converts more skeptics than any blog post, including this one. Give the experiment two sprints before judging — the first session with any new deck is always a little awkward, and the value shows up once the anchors settle.
Try it with your team
The best way to feel why the gaps work is to run a round. You can start a free Planning Poker session with a Fibonacci deck in seconds, or read the broader agile estimation guide for how points and velocity fit together. Either way, the next time someone asks where the 4 went, you'll have a good answer.