Fibonacci for anything your team will track velocity on, T-shirt sizes for early, rough roadmap sizing. The difference is really about how much precision the estimate needs to defend. Try planning poker cards, free, no login, either scale built in.
What each scale actually is
Fibonacci-based story points use 0, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89 and a “?” for “not enough information yet.” T-shirt sizing swaps numbers for XS, S, M, L, XL, XXL, plus its own “?”. Both are relative scales: a 5 or an M isn’t a fixed unit of time, it’s a size relative to the other items your team has already estimated, which is the entire point of estimating this way instead of guessing hours directly. Worth knowing if you’ve used a different deck before: many planning-poker tools round the large end of the sequence for readability, 20, 40, 100 instead of 21, 34, 89. Ours keeps the true Fibonacci numbers throughout, so if a figure looks unfamiliar next to a deck you’ve used elsewhere, that’s why.
Why the gaps widen instead of staying even
A linear scale, 1, 2, 3, 4, 5, invites false precision: is this item really a 7 and not a 6? Nobody can defend that distinction with a straight face for anything beyond a trivial task. The Fibonacci sequence’s gaps grow as the numbers get bigger specifically to route around that argument. Small items get fine-grained options, 1, 2, 3, because small items are genuinely easy to size precisely. Big items jump straight from 13 to 21 to 34, because past a certain size, the honest answer is “this is big and uncertain,” not “this is precisely 19.” The widening gap is the scale doing its job: it makes over-precise arguments about large, uncertain work structurally harder to have, not because a committee decided teams should stop arguing, but because the scale itself only offers a coarse choice once items get large. T-shirt sizing takes the same idea further, there’s no numeric gap to even discuss, just a size.
Fibonacci: the standard for tracked velocity
If your team measures velocity, points completed per sprint, and uses that number to forecast future work, use Fibonacci. Velocity math depends on a consistent unit across sprints, and a numeric scale with fixed values gives you something to sum and average over time. This is also where the “not hours” distinction matters most: a point total that quietly gets treated as an hour estimate breaks the model, because points are calibrated against your own team’s past items, not against a clock.
T-shirt sizes: for early roadmap sizing
Before a backlog is broken into real tickets, when a product lead just needs a rough sense of “is this a quarter of work or a week,” T-shirt sizes are the better fit. No one has to defend a Fibonacci number they’ll be held to later, an M just means “medium, roughly like other things we’ve called medium,” and that looseness is a feature at this stage, not a gap. Trying to run Fibonacci poker on a roadmap full of half-defined epics usually produces false precision on items nobody understands yet.
Custom decks, for teams that use neither
Type your own deck, one value per line, and it’s remembered on your device next time. Some teams use a linear 1-2-3-4-5 anyway despite the argument above, others use risk labels instead of size. Both built-in presets just cover the two most common cases, not every case.
Converting between the scales
Sometimes a stakeholder wants a number after a team estimated in T-shirt sizes, for a roadmap slide or a burn-down chart. A rough mapping: XS to 1, S to 2 or 3, M to 5, L to 8 or 13, XL to 21 or higher. Use this for reporting only, never for planning. Converting a T-shirt size back into a number the moment someone wants to do arithmetic on it reintroduces exactly the false precision the team chose T-shirt sizes to avoid in the first place, and it isn’t a standard either side of the industry agrees on, treat it as a rough translation, not a lookup table with a right answer.
Running it without another SaaS login
The honest section, stated plainly: planning poker cards has no rooms, no invite links, no accounts, and no real-time sync between devices watching different screens. It works two ways instead. On a video call, everyone opens the page on their own device, picks a card, and holds it up to the camera when the facilitator calls for reveal, exactly like a physical deck passed around a table. In a shared room, one device gets passed from person to person, each person votes in secret and the vote hides itself until everyone’s had a turn, then the whole team reveals together. Send the link to the team and that’s the entire setup, there’s nothing to create and nothing to sign into first.
Reading the reveal: average, spread, and outliers
The reveal shows the average alongside the lowest and highest vote, and both of those extremes are visually highlighted rather than left for someone to spot manually. That’s a deliberate design choice: the average alone hides the disagreement that actually matters. Two votes of 5 and 8 average to 6.5, information-free on its own, but knowing the actual spread was 5-to-8 tells the room there’s a real gap worth a short conversation before moving on.
Are story points hours?
No, and treating them that way is the single most common way teams undermine their own estimation process. A point is a relative size against your own team’s history of similarly-sized past items, not a promise about clock time. Two different teams can honestly estimate the same piece of work at different point values and both be right, because each team’s scale is calibrated against its own baseline, not a universal unit. The moment a manager starts dividing points by hours to build a schedule, the incentive flips from “estimate honestly” to “estimate defensively,” and the whole exercise stops producing useful information.
What the outliers actually tell you
When the highlighted low and high votes are close together, the team already agrees, more or less, and the average is a fine number to move on with. When they’re far apart, that gap is nearly always more useful than the number itself: someone who voted low may know a shortcut, or may be unaware of a complication someone who voted high already ran into. Standard practice is to ask the low and high voters to explain their reasoning before anyone else speaks, then vote again on the same item once the room has heard both sides, rather than settling for the average by default.
The “?” card
A “?” is a vote like any other option on the deck, and it means exactly what it looks like: not enough information to size this item yet. One “?” among otherwise clustered numeric votes is usually just one person being honest. Several “?” votes on the same item is a signal worth acting on differently, the item probably needs more detail or a smaller scope before the team estimates it again, rather than needing another round of the same vote.
Running a session, start to finish
Open the deck picker and choose Fibonacci, T-shirt sizes, or a custom deck typed in beforehand. Send the link to the team, everyone opens it on their own device, no room code and no sign-in step for anyone to get stuck on. Read out the first item and let each person pick a card in secret, on a call that means holding it up when the facilitator calls for reveal, in a shared room it means passing one device around and voting in turn before revealing together. Check the spread before moving on, as above. Round history stays in the browser the session ran in, so nothing needs re-entering for the next item.
FAQ
Should we use Fibonacci or T-shirt sizes?
Fibonacci for anything your team will track velocity on, the widening gaps force a real decision about uncertainty as items get bigger. T-shirt sizes for early, rough roadmap sizing where no one wants to defend a specific number yet.
Why does planning poker use the Fibonacci sequence?
Because the gaps between values grow as the numbers get bigger, 1 to 2 to 3 to 5 to 8, which matches how hard estimation actually gets. A small item is easy to size precisely, a big one is rarely worth arguing over 12 versus 13, so the scale skips straight to a bigger jump.
Are story points the same as hours?
No. A story point is a relative size, not a time estimate, two teams estimating the same item can honestly land on different numbers depending on their own baseline. Treating points as hours defeats the purpose of using a relative scale at all.
What does the “?” card mean?
I don’t have enough information to estimate this yet. It’s a vote like any other, and a “?” from more than one person is usually a sign the item needs more detail before estimating continues.
What do you do when two estimates are far apart?
Talk to whoever voted lowest and whoever voted highest before anyone else speaks, they usually know something the rest of the room doesn’t. Our own reveal highlights exactly those two votes for this reason, then the team re-votes once the gap is understood.
Ready to run a session? Try planning poker cards, free, no login, Fibonacci or T-shirt built in. Timing the discussion between votes? The online timer covers that, or sketch the item on the online whiteboard while the room talks it through. Looking for something else first? Browse free online tools for the rest of the lineup.