Numbers invite arithmetic. Somebody multiplies, somebody divides, and within a month the estimates are being treated as a schedule. T-shirt sizes resist that — you cannot add up a column of Mediums.
What t-shirt sizing is
T-shirt sizing replaces the numeric deck with clothing sizes: XS, S, M, L, XL, sometimes XXL. Everything else about the method is unchanged. Each person picks a size privately, everyone reveals at once, outliers explain, the team converges.
Sizes are relative in exactly the way points are — an L is bigger than an M, and roughly the way a jumper is bigger than a t-shirt rather than by any defined ratio. What is removed is the false precision, and with it the temptation to do sums on the results.
| Size | Typical meaning | Rough point equivalent |
|---|---|---|
| XS | Trivial. Minutes to an hour. | 1 |
| S | Small and familiar. | 2–3 |
| M | A normal piece of work. | 5 |
| L | Large. Probably splits. | 8–13 |
| XL | Too big to plan as one item. | 21+ |
Treat that last column as a translation aid for teams migrating, not a conversion rate. Publishing a fixed mapping recreates the arithmetic problem you switched to avoid.
When sizes beat points
Roadmap and quarterly planning
Work that is months out is not understood well enough to carry a number, and a number implies it is. Sizing a quarter's candidate initiatives as S/M/L is honest about the resolution available and still good enough to decide what fits.
Teams new to estimation
Points require a shared numeric anchor, and establishing one takes a few sprints. Sizes work immediately because everyone already has an intuition for what "large" means. Many teams start with sizes and move to points once they have history.
Mixed audiences
When stakeholders outside the team see the output, "8 points" invites the question "how many days is that?". "Large" invites "what would make it smaller?" — a far more useful conversation.
Speed
Five options instead of nine, and no debate about adjacent values. Teams sizing a large backlog for the first time often get through it in half the time.
You cannot compute velocity from t-shirt sizes without converting them to numbers first, which reintroduces exactly the arithmetic you were avoiding. If you need sprint-level forecasting, points are the better tool.
When points beat sizes
- You need velocity. Forecasting requires something summable. Sizes are not.
- Your backlog clusters. If nearly everything is an M, the scale has stopped discriminating and a numeric scale will give you more resolution where you need it.
- You want to track estimate accuracy. Comparing estimates against outcomes over time is much easier with numbers.
The hybrid most teams land on
A common and sensible arrangement uses both, at different altitudes:
- Size epics and roadmap items in t-shirts. Fast, coarse, honest about how little is known.
- Split anything L or XL before it gets near a sprint.
- Point the resulting stories during refinement, when the team knows enough for a number to mean something.
The sizes drive "what should we take on this quarter"; the points drive "what fits in this sprint". Neither is asked to do the other's job.
Running a t-shirt session
Identical to a numeric session, with one addition worth the time: agree a reference item for at least M and L before you start. "Medium is like the notifications work we did in March" is a more durable anchor than anyone's private sense of medium, and it keeps the scale from drifting the way unanchored sizes reliably do.
Keep the hidden vote and the simultaneous reveal. The anchoring problem is not caused by numbers — it is caused by hearing someone else's answer first, and "this one's a large" anchors a room just as effectively as "this one's an eight".
GiveItPoints supports t-shirt scales alongside numeric ones, so a team can run sizes for roadmap sessions and points for sprint planning without changing tools. See custom voting scales.
Common questions
How many sizes should we use?
Five (XS to XL) is the sweet spot. Adding XXL and XXXL rebuilds a numeric scale with worse labels; if you need that much range, use points.
Can we convert sizes to points for reporting?
You can, and teams do, but be aware it quietly turns sizes into points. If reporting needs numbers regularly, estimate in numbers.
What if everything comes out Medium?
Either the backlog is genuinely uniform — in which case count stories and skip estimation — or the scale is not discriminating, and it is time to move to points.