Scene: planning poker. Fourteen people. Fibonacci cards in everyone’s hand. One, two, three, five, eight, thirteen, twenty-one.
Story points only move up. Estimate a 3, it takes a week, it’s a 5 next time. Estimate a 5, it takes two weeks, it’s an 8. Estimate an 8, it takes a month — we should have known it was a 13. Nobody has ever, in the history of agile, re-estimated a ticket downward after the fact. Check your Jira.
Given this physics, the math is simple. The optimal vote is the number nobody will use against you later. Too low, you over-promised. Too high, you sandbagged. In the middle, where the conversation moves on and the chart stays green —
Always say 5.
Why Fibonacci, actually
Pause here. Ask your team why Fibonacci. You’ll get a guess dressed up as an answer, or the honest version: “I don’t know, that’s just how it’s done.”
There is, in fact, a real reason. Nobody in the room remembers it.
The reason is Weber’s Law, a piece of psychophysics from 1834. It states that the human brain perceives differences in percentages, not absolute amounts. Hand someone a 1 kg weight and a 2 kg weight — they’ll instantly say which is heavier. Hand them 20 kg and 21 kg — same one-kilo difference — and they cannot. The first gap is 100%. The second is 5%. The brain doesn’t do subtraction. It does ratios.
This is why Fibonacci. The sequence — 1, 2, 3, 5, 8, 13, 21 — has a quiet mathematical property: after the first two numbers, each is roughly 60% larger than the one before. The ratio converges to the golden ratio, 1.618. That 60% jump lands just above the threshold at which the human brain reliably perceives a difference. Vote between a 5 and an 8 and you can feel the gap. Vote between a 5 and a 6 — only 20% — and you can’t. Two engineers will swap answers and neither will know why.
Mike Cohn — the agile consultant who popularised planning poker — formalised all of this in an essay everyone cites and nobody reads. The scale is nineteenth-century psychophysics, dressed in twelfth-century numerology, sold as a twenty-first-century best practice.
It’s also supposed to capture three things at once. Complexity — how tangled the problem is. Uncertainty — how much you don’t know. Effort — how much work is in front of you. C-U-E. The Fibonacci gaps even encode uncertainty into the scale itself: bigger numbers, bigger gaps. The math is honest about its own blindness.
On paper, this is genuinely elegant. Weber. Golden ratio. C-U-E. Built-in humility about precision. If you described it without a team attached, you’d think: of course we should estimate this way.
Why none of it works
Weber’s Law only functions when the stimulus is consistent. A 1 kg weight has to weigh 1 kg every time. Otherwise the 60% jump means nothing — you’re comparing apples to estimated-apples.
This is where every team breaks the system. The senior engineer’s 5 includes infrastructure risk and review cycles. The junior’s 5 is “two days of typing.” The platform team’s 5 is the product team’s 13. Nobody is wrong. They are using the same number to measure five different things. Weber cannot help you. Weber needs a kilo. You have brought an undocumented API — different behaviour on every call.
And Cohn knows this. Read his own essay closely. He admits the original Fibonacci broke for him at 21 — “implied a precision we couldn’t support.” So he modified it. 21 became 20. Then he added 40 and 100. The inventor of the scale admitted the scale breaks.
The teams kept using it anyway. The math left the room. The cards stayed. That is the exact moment a tool becomes a ritual.
I wrote about ritual in Always Serve the Process — the moment a tool becomes a ceremony, the ceremony becomes the point and the tool becomes invisible. Fibonacci is now a ceremony. Story points are now a ceremony. Performed correctly. Signifying compliance.
This is why your team cannot answer why Fibonacci. Rituals don’t have why. Rituals have because we do.
So what actually works
Two things. Only two.
Clear definition. Write down what one point means on your team. One sentence. If you can’t write it, you don’t have a unit — you have a vibe. Weber doesn’t work on vibes.
Consistency. Use the same definition every sprint. The unit can be Fibonacci. It can be T-shirts. It can be undocumented APIs. Mechanically identical. The system isn’t the number — it’s the agreement, held steady.
That is the entire technology. Everything else — the cards, the apps, the certified coach, the velocity chart — is either decoration over those two things, or distraction from them.
Five things you can do this week
You don’t need to overthrow anything. You can fix estimation inside whatever ceremony is already running.
Write down what “one point” means. A single sentence, pinned in the channel. Calibrate everyone against it.
Anchor against a known ticket. Pick a closed ticket everyone remembers. Call it a 3. Every future estimate is “bigger or smaller than that one.” Concrete beats abstract.
Track delivered vs. estimated. Not just velocity. Estimated 40, delivered 22 — that’s the number. Velocity hides it. Hit-rate exposes it.
Ban re-estimation after the fact. A 3 that took a week stays a 3 in the record. Otherwise you’re editing the past until the chart looks nice.
Run one sprint with two buckets. “This sprint” / “Not this sprint.” No numbers. If nothing breaks, the numbers were decoration.
None require permission. None require a new framework. All drag estimation back toward the only two things that ever mattered: a clear definition, and consistency.
Until then
Fibonacci has a real reason. Weber’s Law is real. The golden ratio is real. None of it works on a team that hasn’t agreed what a 5 measures.
Until your team writes it down, the number on the card is a finger-quote. The finger-quote goes on the slide. The slide goes up. Somewhere, a director with a budget makes a real decision based on a number that started life as air.
In that environment, the safest answer is the one in the middle. The one nobody re-estimates.
Always say 5.
Until the team writes down what 5 actually is — at which point, you can stop saying 5 and start saying anything.
By order of thebeanengineer.com — always say 5.
Be an engineer. 🚀