The Fibonacci scale in planning poker: why it works
The numbers 1, 2, 3, 5, 8, 13, 21 appear in nature, architecture, and stock charts. They also appear on planning poker cards — and not by accident. The Fibonacci sequence's growing gaps match the way human perception of complexity works, making it unusually well-suited for agile story point estimation.
The math behind the Fibonacci sequence
Each number in the Fibonacci sequence is the sum of the two preceding numbers: 1, 1, 2, 3, 5, 8, 13, 21, 34... The resulting sequence grows exponentially rather than linearly.
In planning poker, this exponential growth is not a mathematical formality — it is a deliberate design choice. As work grows larger, uncertainty grows faster than linearly. A story that is twice as large to describe is not twice as predictable; it tends to be much less predictable.
The Fibonacci gaps encode this relationship. The difference between a 1 and a 2 is meaningful and distinguishable. The difference between a 13 and a 21 is not — both represent "large, high-uncertainty work." The scale's structure forces teams to think in categories rather than false precision.
The psychology behind Fibonacci estimation
Weber's Law in psychophysics states that the minimum perceivable difference between two stimuli is proportional to the magnitude of the stimuli. In plain terms: you can reliably distinguish 1 kg from 2 kg, but cannot reliably distinguish 100 kg from 101 kg.
The same principle applies to estimation. A team can reliably distinguish between a 1-point and a 2-point story — one is trivially small, the other is clearly a bit more work. But a team cannot reliably distinguish between a 7-point and an 8-point story. The linear scale creates false distinctions that the Fibonacci scale wisely avoids.
The Fibonacci sequence's ratio between consecutive numbers (approximately 1.618, the golden ratio) is close to the Weber fraction for complex cognitive judgments, making it more neurologically natural for estimation than arbitrary scales.
Fibonacci variants and alternative scales
The standard Fibonacci scale is not the only option. Variations exist to serve specific team needs.
Modified Fibonacci (0, 0.5, 1, 2, 3, 5, 8, 13, 20, 40, 100)
Adds sub-1 values for tiny tasks, rounds the large numbers (20 instead of 21) for simplicity, and includes a 100 for very large items. Used by teams that frequently estimate both micro-tasks and large epics in the same session.
Powers of 2 (1, 2, 4, 8, 16, 32)
A mathematically clean progression preferred by some engineering teams. Works well in practice but lacks the psychological research backing of the Fibonacci sequence. The gaps grow faster, which drives story splitting more aggressively.
T-shirt sizes (XS, S, M, L, XL, XXL)
Qualitative rather than numerical. Best for early-stage discovery or roadmap planning when stories are too large and ambiguous for point estimates. Convert to numbers during refinement.
Why linear scales fail
Linear scales (1–10 or 1–5) create the illusion of precision without the reality. Teams cluster around the middle, argue about the difference between 6 and 7, and produce velocity data with high variance. The Fibonacci sequence's non-linearity is a feature, not an inconvenience.
Using the Fibonacci scale effectively
The scale only works well with the right process around it. A few practices make the difference between useful Fibonacci estimation and a number-picking exercise.
- Set a reference story: before the first session, identify one completed story that everyone agrees is a clear 3 or 5. Use it as an anchor for all new estimates. "Is this bigger or smaller than the 3?" is much easier than estimating in a vacuum.
- Treat 13+ as a split signal: stories that land consistently at 13 or above are probably too large for a single sprint. Use the high estimate as a trigger for story splitting, not as an accepted commitment.
- Use the question mark card: when a team member selects "?" or "I don't know", pause the vote and address their uncertainty before continuing. Estimating through acknowledged ignorance produces meaningless numbers.
- Don't force consensus: if votes diverge after two rounds, note the blocker and move on. A story that the team cannot estimate in two rounds is not ready to estimate — it needs more refinement.
Fibonacci planning poker FAQ
Why does planning poker use Fibonacci instead of other sequences?
Fibonacci is the most well-researched non-linear progression for human estimation tasks. It was popularised in agile contexts by Mike Cohn's planning poker framework and has been validated by many teams over two decades of use.
Should we use the exact Fibonacci sequence or a modified version?
Start with standard Fibonacci (1, 2, 3, 5, 8, 13, 21). Switch to a modified version only if your team consistently encounters stories that don't fit well — for example, if you frequently estimate very small tasks or very large epics.
What does it mean when the whole team votes 21?
A unanimous high vote usually means the story is not ready to estimate. The team sees it as complex, risky, or unclear. Treat it as a refinement signal: add clarity tasks to the backlog and re-estimate after more information is available.
Can we change our scale mid-project?
Yes, but calibrate carefully. A scale change makes previous velocity data incomparable to new data. Run a calibration session using reference stories to establish how the new scale relates to your historical velocity.
Try planning poker with the Fibonacci scale — free, no signup required