Why estimation decks use the Fibonacci sequence

Ask a team whether a task is a 20 or a 21 and you'll get shrugs. Ask whether it's an 8 or a 13 and you'll get an argument worth having. That difference is the whole reason estimation decks look the way they do.

If you've ever wondered why your planning poker cards jump from 13 to 20 instead of politely counting upward, it's not tradition and it's not maths worship. The missing numbers are doing work. They force you to make a call instead of splitting hairs.

The gaps are the point

A plain number line treats all estimation as equally precise. It says: you can tell a 14 from a 15, so go ahead. But nobody can. Once work gets past a certain size, the honest error bars swallow the differences between neighbouring numbers. The gap between 20 and 21 is noise. The gap between 13 and 20 is a real question: is this thing merely big, or is it big enough that we're mostly guessing?

Fibonacci-style decks encode that honestly. Small numbers sit close together (1, 2, 3) because small work is where you actually can be precise. You've done a hundred one-point tasks; you know what a 1 looks like. As the numbers grow, the gaps widen, because your certainty shrinks at roughly the same rate. By the time you're weighing 8 against 13, the deck is asking a question you can genuinely answer: which of these two buckets does this belong in? Not "give me a number" but "make a choice".

Forced choices surface disagreement, and disagreement is the actual product of a planning poker session. When half the room says 5 and the other half says 13, somebody knows something the others don't. Maybe there's a data migration hiding in the story. Maybe someone already built half of it last quarter and nobody else noticed. A deck with every integer available lets people hide in the middle of the range. A 5-versus-13 split can't be averaged away to an 8 with a straight face, or at least it shouldn't be. The estimate that comes out of that conversation matters less than the conversation itself.

Your brain already works this way

There's a well-worn observation from psychophysics, usually filed under Weber-Fechner, that we perceive differences relative to the size of what we're comparing. Hold a feather and add a coin, and you'll notice immediately. Hold a suitcase and add the same coin, and you won't. The absolute difference is identical; the perceived difference depends entirely on the baseline. Your sense of effort works the same way. A day of extra work on a two-day task changes everything. A day of extra work on a two-month project vanishes into the rounding. A deck whose steps grow with the numbers matches how you actually perceive size, instead of pretending your judgment has fixed resolution across the whole scale.

That's the intuition. You don't need the formula, and the formula wouldn't help you estimate anyway.

The deck most tools ship isn't really Fibonacci

Look at the deck you'll find in most tools, including the default in ScrumMastr's planning poker: 0, 1/2, 1, 2, 3, 5, 8, 13, 20, 40, 100. True Fibonacci would continue 21, 34, 55, 89. The modified deck deliberately breaks from the sequence at the top, and the reason is worth internalising.

21 looks calculated. It looks like someone did arithmetic. 20 looks like what it is: a round-number guess. Mike Cohn, who popularised this deck, rounds everything past 13 for exactly that reason. A stakeholder who sees "21" on a roadmap might believe it. A stakeholder who sees "20" understands, at some level, that they're looking at a shrug with a number attached. The deck is honest about its own accuracy, which is more than you can say for most estimates.

The bottom of the deck earns its keep too. Zero is for work that's essentially free once someone's in the code anyway: a config flip, a copy change. The half point exists so that trivial-but-not-free work doesn't inflate into a 1 and quietly distort your baseline. Some teams drop both cards. That's fine. It depends on how much tiny work actually flows through your board.

What a 40 actually means

Here's the part teams get wrong most often. 40 and 100 aren't estimates. They're verdicts.

When a team votes 40, they're not saying "this will take five times as long as an 8". They're saying "this is too big to estimate, and probably too big to pull into a sprint". The right response is almost never to write 40 in the tracker and move on. It's one of two things:

• Split it. Most 40s are three or four 8s and 13s wearing a trench coat. Find the seams: separate the API from the UI, the happy path from the edge cases, the read path from the write path.
• Spike it. If you can't split it because nobody understands it well enough to find the seams, that is your answer. The next piece of work is a timeboxed spike to go find out. Estimate the spike, not the mystery.

A 100 is the same message, louder. Treat it as a flare, not a number. And if 40s regularly land in your sprints unchanged, the deck is telling you something about your refinement process. Listen to it.

The question mark and the coffee cup

Two cards in the deck have no number on them, and neither is a joke.

The question mark means "I can't estimate this", which is different from "I estimate this is big". Someone who plays it might be missing context the rest of the room has. Or they might be the only person who's noticed that the story description doesn't match what the codebase actually does. Either way, don't talk them out of it. Ask what they'd need to know, then go get it.

The coffee card means "I need a break". If you're tempted to remove it as unserious, count how many of your sessions run past the ninety-minute mark with estimates getting visibly lazier each round. Tired estimates are bad estimates. The cheapest fix in all of agile is a ten-minute break, and having a card for it means the most junior person in the room can call for one without saying a word.

When Fibonacci isn't the right deck

For sprint-level work, it usually is. But not always, and knowing the trade-offs beats defaulting on autopilot.

T-shirt sizes shine exactly where numbers cause trouble: quarterly planning, epic sizing, anything a spreadsheet-minded stakeholder might try to sum. You can't add an M to an L, and that's the feature. The moment roadmap estimates are numeric, someone will divide them by velocity and produce a delivery date with false confidence baked in. Sizes resist that. Use them for anything above story level and translate to points only when the work is close enough to see clearly.

Powers of 2 follow the same philosophy as Fibonacci with even fewer options in the middle. Some teams find that removing the 3 and the 5 removes their two most argued-over cards. Others find the jump from 2 to 4 too coarse for where their stories actually cluster. It depends on your distribution. If most of your work lands between 2 and 5, Fibonacci gives you resolution right where you need it. If your team burns twenty minutes per story on 3 versus 5 and the answer never changes what gets planned, powers of 2 will buy that time back.

The deck to avoid is the plain 1-to-10 line. It looks friendly and neutral, and it quietly invites everything the other decks are built to prevent. Is this a 6 or a 7? Nobody knows, but the deck implies someone should, so the debate happens anyway. Then the sixes get averaged with the sevens, the velocity chart grows a decimal place, and people start treating story points as a measurement instrument rather than a sorting tool. That confusion runs deep enough that it deserves its own article; see what story points actually measure for why points were never meant to survive that kind of arithmetic.

Pick one deck and let it get boring

Whatever you choose, the deck matters far less than sticking with it. Estimates only mean anything relative to your own history. An 8 is "like those other 8s we did", nothing more, and if you change decks you reset that history to zero. So resist the urge to tinker. Pick modified Fibonacci unless you have a specific reason not to, agree on two or three reference stories the whole team remembers, and then leave the setup alone for at least a couple of quarters.

And when the deck forces a choice that feels uncomfortable, when a story feels like it's "really an 11", that discomfort is the tool working as designed. Round up. Uncertainty has never once made a piece of work smaller.