T-shirt sizing: a simple way to estimate, and when to switch to points

T-shirt sizing is a relative estimation technique in which a team sorts backlog items into sizes such as XS, S, M, L and XL instead of giving them numbers. It is quick, easy to explain to anyone and ideal for rough, early estimates: sizing a roadmap, a set of epics or a backlog nobody has looked at yet.

Its weakness is that sizes don’t add up. You can’t plan a sprint or forecast a date from “three mediums and a large” without deciding what those sizes are worth, so teams that need forecasts either put agreed numbers behind the sizes or move to story points. This guide covers how to run a session, how to map sizes to numbers, where sizes beat points and the signs that it’s time to switch.

What is T-shirt sizing?

Like story points, T-shirt sizes are relative. A medium isn’t a number of days; it’s “about as big as the other mediums we know”. The size still reflects effort, which takes in how much work there is, how tricky it is and how much is unknown.

The labels are the point. Nobody mistakes “L” for a promise of hours, so the conversation stays on comparison rather than calendars. That makes sizes especially good with mixed groups: product managers, designers and stakeholders can join in without learning a numeric scale first. It also makes them a gentle first step for teams new to relative estimation. Mike Cohn’s view is that starting with T-shirt sizes is fine if a team finds it easier, as long as numbers follow.

How to run a T-shirt sizing session

1. Pick reference items. Choose a finished item the team remembers for S, M and L. Everyone should be able to say why each one is that size.
2. Agree what XL means. A common rule is “too big to finish in one sprint, so it must be split before it is planned”. Decide whether you need XS and XXL at all.
3. Introduce each item briefly. The product owner gives the goal and the acceptance criteria in a minute or two. Questions are welcome; design debates are not.
4. Size at the same moment. Everyone chooses a size privately, then all reveal together. This stops the first confident voice anchoring everyone else.
5. Discuss only the outliers. If the votes are all M or L, take the larger and move on. If they range from S to XL, ask both ends what they know.
6. Record the size and any assumptions. Note anything that would change the size, and flag every XL for splitting.

Tip Give each item about two minutes. If the team can’t settle on a size in that time, the item usually needs more refinement rather than more debate, so mark it with a question for the product owner and move on.

For a very large backlog you can skip voting item by item. Put the sizes across the top of a wall or online board, let the team silently place items under them, then discuss only the items that keep moving. That variation is close to affinity estimation, and it can size a big backlog in one sitting.

A worked example: sizing a returns feature

An online retailer wants customers to return items without phoning support. In a 30-minute session the team sizes the first ideas:

In half an hour the product owner learns that most of the feature is small or medium work, that one idea needs breaking down and that the carrier integration is the main risk. For a roadmap conversation, that is exactly the level of detail needed.

Mapping T-shirt sizes to numbers

As soon as anyone wants to add sizes up, you need numbers behind them. Mike Cohn’s advice is to put underlying numbers on the sizes, for example Medium = 5, and then gradually shift to using the numbers directly. The exact values matter less than agreeing them. A common Fibonacci-shaped mapping, which keeps the widening gaps that make relative estimation work, puts M at 3:

Three rules keep a mapping honest:

• Agree it before you size, not after. Otherwise people re-argue the sizes to get the numbers they want.
• Keep it to your team. Your M reflects your codebase and your Definition of Done. Another team’s M is a different garment.
• Use a separate mapping for epics if you need one. At roadmap level, an L epic might be worth 40 points or more. Just don’t mix epic-level and story-level sizes in the same velocity.

If you already have a backlog sized in T-shirts, you don’t need to start again. Apply the mapping, then check the items heading into the next sprint or two in a quick planning poker round, because a size given months ago may no longer fit what the team now knows.

On planningpoker.org SprintPoker includes a T-shirt deck (XS to XXL) and a T-shirt deck mapped to numbers (XS=1, S=2, M=3, L=5, XL=8, XXL=13), which suits a team moving gradually from sizes to points.

Using T-shirt sizes for roadmaps and big backlogs

T-shirt sizing is at its best when the question is “roughly how big is this?” rather than “what fits in the next sprint?” That makes it a natural fit for roadmap planning, where you might be weighing 20 or 30 ideas, most of which will change before anyone builds them.

A few ways to make roadmap sizing useful:

• Size against value. A simple grid of size against expected value quickly shows the cheap, valuable ideas worth doing first, and the expensive ones that need a stronger case.
• Translate to time only through history. If your last few L initiatives each took one to two months, say so as a range. Never decree that “an L is six weeks”.
• Re-size as you learn. A roadmap size is a starting point. When an item comes up for backlog refinement, split it and estimate the pieces properly.

Back to the returns example: “show whether an order can still be returned” is small and should save support time straight away, so it goes first. The automatic refund is an XL that depends on carrier events the team can’t yet rely on, so it waits, and the roadmap says so honestly instead of pretending it has a date.

Common T-shirt sizing pitfalls

• Adding sizes up. Mike Cohn puts it best: “You cannot tell a boss you’ll be done in 3 mediums, 4 larges, and 2 petites.” Without agreed numbers, a sum of sizes means nothing.
• Different tape measures. Cohn’s second warning is that people picture sizes differently: one person thinks an XL is 50% bigger than an L, another thinks 25%. Reference items and a written mapping fix this.
• Everything is a medium. When most items land on M, the scale has stopped telling items apart. Check your references, compare the mediums with each other and split the category if you need to.
• XL as a parking space. If XLs pile up unsplit, the backlog looks sized but isn’t. Treat XL as an instruction to break the item down.
• Sizing without the people doing the work. Sizes from managers alone are opinions about effort, not estimates of it.

When to switch from T-shirt sizes to story points

Sizes and points aren’t rivals. Many teams use both: sizes for the roadmap and points for the next few sprints. The two compare like this:

Some signs suggest it is time to bring numbers into refinement:

• You need sprint-level plans or release forecasts, and people keep asking how many mediums fit in a sprint.
• Debates about “a big M or a small L” are taking longer than the work they describe.
• The team is comfortable comparing items and wants to track velocity.
• Stakeholders are already converting sizes to numbers in spreadsheets, so you may as well agree the numbers together.

The smoothest switch is gradual. Map your sizes to numbers for a few sprints, then start voting with the numbers directly, keeping your old reference items as anchors. Our guides to story points and Fibonacci story points cover what changes, and our comparison of agile estimation techniques shows where sizes fit alongside other methods.