The modified Fibonacci sequence is used for estimation because it mirrors how humans actually perceive differences in size and effort. As tasks get bigger, your ability to distinguish between them gets worse, and the widening gaps in the sequence reflect that reality. Instead of pretending you can tell the difference between a 21-point task and a 23-point task, the sequence forces you to choose between 13 and 20, keeping estimates honest about their inherent uncertainty.
The Scale: 1, 2, 3, 5, 8, 13, 20, 40, 100
The standard Fibonacci sequence in mathematics is 1, 2, 3, 5, 8, 13, 21, 34, 55, where each number is the sum of the two before it. The modified version used in agile estimation changes the tail end to 1, 2, 3, 5, 8, 13, 20, 40, and 100. Mike Cohn, who popularized Planning Poker, introduced this modification after noticing that an estimate of “21 story points” gave stakeholders the impression the team had calibrated their estimate with surgical precision. They hadn’t. Rounding to 20 made the fuzziness of large estimates visible. Once that first deviation proved useful, the team experimented further, replacing 34 and 55 with 40 and 100.
The rounded numbers at the top of the scale are also easier to remember. When you’re working through a backlog of dozens of items, a simpler set of options keeps the focus on discussion rather than arithmetic.
Why Non-Linear Gaps Work Better Than Linear Ones
There’s a well-established principle in psychology that explains why this works. Your brain doesn’t process differences in absolute terms. It processes them as ratios. If you’re holding a 1-pound weight, adding another pound is obvious. If you’re holding a 50-pound weight, adding one more pound is nearly undetectable. The threshold for noticing a change is always a proportion of the starting amount, not a fixed number.
This applies directly to estimation. The difference between a task that takes 2 hours and one that takes 3 hours is easy to feel. The difference between a task that takes 34 hours and one that takes 37 hours is basically imperceptible. A linear scale (1, 2, 3, 4, 5, 6, 7…) ignores this and asks you to make distinctions your brain can’t reliably support. The Fibonacci-style sequence respects it. The gaps between numbers grow roughly in proportion to the numbers themselves, matching the natural resolution of human judgment.
Complexity doesn’t increase in a straight line either. A feature that’s twice as large as another often carries more than twice the uncertainty, because it involves more unknowns, more integration points, and more potential surprises. The accelerating gaps in the sequence capture this: bigger items deserve bigger buckets.
Eliminating False Precision
One of the most practical benefits is that the sequence makes it impossible to waste time on meaningless debates. If your scale included every integer from 1 to 100, a team could spend ten minutes arguing whether a task is a 14 or a 15. That argument produces no useful information. The actual uncertainty around that task might span from 10 to 20, so debating a single point is an illusion of accuracy.
With the modified Fibonacci sequence, those same team members have to choose between 13 and 20. That’s a meaningful distinction. Is this task closer to the 13-point reference item we completed last sprint, or is it noticeably larger? The wider gap forces the conversation toward the right question: what’s the rough size of this work relative to things we’ve already done?
Precise-looking numbers like 21, 34, or 55 create a related problem. They suggest the team has a level of confidence that doesn’t exist. When a project manager sees “34 story points,” it reads like a carefully measured figure. When they see “40,” the roundness signals what it really is: a rough estimate. The modified sequence communicates uncertainty through the numbers themselves.
Faster Decisions, Less Mental Fatigue
Estimation sessions often cover large backlogs. If each item requires choosing from dozens of possible values, decision fatigue sets in quickly. Simpler decks with fewer, more distinct options reduce mental overhead. Instead of agonizing over fine gradations, you’re picking from a short list where each option is clearly different from its neighbors. Teams using non-linear scales consistently estimate faster than those using linear ones, because there are fewer plausible choices for any given item and less temptation to over-analyze.
This speed matters in Planning Poker specifically. In that process, every team member independently selects an estimate, then everyone reveals simultaneously. If estimates differ, the people with the highest and lowest picks explain their reasoning, and the team re-votes. The sequence’s limited options mean disagreements tend to cluster around adjacent values (say, 5 versus 8), which focuses the discussion on a single question: is this task small-medium or medium-large? With a linear scale, disagreements scatter across many close values, and the resulting conversations are harder to resolve.
How It Compares to Other Scales
The modified Fibonacci sequence isn’t the only non-linear option. T-shirt sizes (XS, S, M, L, XL, XXL) work on a similar principle, offering a small number of clearly distinct buckets. They’re popular for rough-cut estimation of large epics, where even Fibonacci numbers suggest more precision than exists. The tradeoff is that T-shirt sizes can’t be added or averaged, which limits their usefulness in sprint planning or velocity tracking.
Powers of 2 (1, 2, 4, 8, 16, 32, 64) are another alternative. They’re non-linear, but the gaps double every step, which is more aggressive than Fibonacci. The jump from 8 to 16 skips a lot of territory that the Fibonacci sequence covers with 8 and 13. For many teams, that’s too coarse in the mid-range where most of their work lives.
Linear scales (2, 4, 6, 8, 10…) create the opposite problem. Between 20 and 100, a scale counting by fours has 20 possible values. That’s far too many choices for uncertain, large work items. Teams using linear scales tend to spend more time debating estimates and seeking details about items that are inherently fuzzy, because the scale invites a level of precision that doesn’t match the available information.
The modified Fibonacci sequence hits a middle ground: tight enough at the low end to make useful distinctions between small tasks, loose enough at the high end to acknowledge that big items are genuinely hard to size. That balance is why it became the default in most agile estimation tools and remains the most widely used scale for story point estimation.