T-shirt sizing for agile estimation
T-shirt sizing is a way of estimating work by putting each item into a clothing size instead of giving it a number. A typical deck has XS, S, M, L, XL and XXL, plus “?” for “no idea yet” and ☕ for a break. Like story points, the sizes are relative: an M is bigger than an S and smaller than an L, and that’s all it claims. Nobody pretends an L is 23 hours of work. Because the scale is so coarse and so familiar, agile teams reach for it most when they need a rough picture of a lot of work quickly.
The scale
A common set has six sizes:
- XS: trivial. Fix a typo in the onboarding email, change a default setting.
- S: small and well understood. Add a “copy link” button to a report page.
- M: an ordinary story with a few moving parts. Let users export their invoices to CSV.
- L: a sizeable piece with some unknowns. Add Apple Pay to an existing checkout.
- XL: probably too big for one sprint. Move user accounts to single sign-on.
- XXL: an epic, or a sign that nobody knows yet what the work involves. Launch the product in a second country.
Some teams drop XXL and stop at XL, others add XXS. What matters is that everyone reads the sizes the same way, so write down one or two real examples for each size, preferably things your team has already built. Examples like the ones above are only a starting point; your own backlog will give you better ones.
It’s hard to say exactly where the technique started, and it isn’t tied to any particular framework. It spread through agile teams as a friendlier alternative to numbers and turns up in many guides to agile estimation under slightly different names.
When T-shirt sizes beat story points
Story points on a Fibonacci scale are good for sprint-level work, where the team needs numbers it can add up into a velocity. T-shirt sizes are better in a few other situations.
Early backlog and roadmap sizing
When a product manager arrives with forty ideas for next year, nobody has the detail to tell a 5 from an 8. The useful question is which ideas are small, which are big and which are enormous. Sizes answer exactly that and don’t suggest more precision than exists.
Epics and initiatives
An epic that will be split into a dozen stories later doesn’t need a point value today. Calling it an XL tells the roadmap discussion what it needs to know, and the detailed estimates can wait until the stories are written.
Non-technical stakeholders
A number on a ticket invites conversion into days, and a “13” can end up in a slide as a deadline. “Large” is harder to misread. Sales, support and leadership can join a sizing session and follow it without learning what a story point is.
Quick triage of many items
Sizes are fast. With only six options there’s less agonizing per item, and a team can get through a long list in one meeting. That makes the technique handy for backlog clean-ups, for comparing options before a planning session, and for spotting items that are clearly too big to take on as they are.
How to run a session
The mechanics are the same as in planning poker, just with a different deck.
- Agree on the reference items. Before you start, pick an existing S and an existing L from your own history. Everyone should be able to say “about like that one”.
- Read out the item. The person who knows it best (often the product owner) describes it in a minute or two. Keep the discussion short; this is sizing, not design.
- Vote at the same time. Everyone picks a size privately, then all cards are revealed together. Voting in secret stops the loudest or most senior person from anchoring everyone else.
- Talk about the extremes. If the votes are S and XL, ask the two people at the ends what they’re thinking. Often one of them knows something the other doesn’t: an old integration, a legal requirement, a component that can be reused.
- Revote or settle. After the discussion, vote once more. If the team still disagrees by one size, take the larger one or the median and move on. Arguing between M and L for ten minutes defeats the purpose.
- Keep a timebox. For a roadmap session, two or three minutes per item is plenty. Anything the team can’t size in that time probably needs a spike or a conversation with someone outside the room.
Mapping sizes to points or time
Sooner or later someone asks what an M actually means. Some teams publish a mapping, for example:
| Size | Story points | Rough duration |
|---|---|---|
| XS | 1 | a few hours |
| S | 2–3 | a day or two |
| M | 5 | up to a week |
| L | 8 | one to two weeks |
| XL | 13 | a sprint or more |
| XXL | split it | unknown |
This is an illustration, not a standard. Each team’s mapping should come from its own completed work, and it will be different for a mobile team and a data platform team.
Be careful with mappings for three reasons. First, the moment an M officially equals five points, people start estimating in points while calling it sizes, and the benefit of a coarse scale is gone. Second, sizes hide a huge spread: two items marked L can differ by a factor of two, and adding up forty rough sizes produces a total that looks far more precise than it is. Third, durations in a table get quoted back as promises. If you need a forecast, give a range and say what it’s based on.
A reasonable compromise is to use sizes for the roadmap and switch to points once an item is refined into stories for a sprint. The size gives the early picture; the points and the team’s velocity give the forecast.
Common mistakes
- Treating sizes as hours. “M means three days” turns relative sizing back into time estimates, with all the usual problems.
- No shared examples. Without reference items, one person’s M is another person’s L and the votes mean little.
- Sizing items that nobody understands. If the honest answer is “no idea”, say so and schedule a spike instead of voting XXL by default.
- Adding sizes up as if they were numbers. Ten Ms aren’t a meaningful total unless you’ve agreed on a mapping and accept its error margin.
- Keeping XL and XXL items as they are. Big sizes are a signal to split the work, not a final answer.
- Letting one person size everything. The value comes from different people spotting different risks, so everyone who’ll touch the work should vote.
T-shirt sizing in SprintPoker
SprintPoker rooms have a deck switcher with two options: Fibonacci and T-shirt sizes. Anyone in the room can switch it. Switching clears the current votes, so pick the deck before the round starts rather than halfway through.
The T-shirt deck has the cards XS, S, M, L, XL and XXL, plus “?” and ☕. On a keyboard, keys 1 to 6 vote XS through XXL.
After the reveal, the room shows the median size, whether the team reached consensus, and how the votes are distributed. There’s no average, because an average of sizes doesn’t mean anything. When the votes split evenly between two sizes, the median is shown as a range, for example “M–L”, which is usually a good cue for a short discussion.
Like the rest of SprintPoker, it’s free and needs no sign-up: create a room, share the link, and start sizing.
FAQ
Is T-shirt sizing better than story points?
Neither is better in general. T-shirt sizes suit early, rough sizing of many items and epics; story points suit sprint planning where you need a velocity. Many teams use both at different stages.
How many sizes should we use?
Six (XS to XXL) is a common set. Fewer sizes make sessions faster but lump more work together; more sizes start to feel like points. Pick one set and keep it stable.
Can we calculate velocity from T-shirt sizes?
Not directly, because sizes can’t be added up. You’d need a mapping to numbers, and at that point you’re effectively using points. For velocity, estimate refined stories in points.
What should we do with an XXL?
Split it. An XXL means the item is an epic or the team doesn’t understand it yet. Break it into smaller pieces or run a spike, then size the pieces.