Every estimation scale makes a trade. Fine grained scales invite debates about 4 versus 5. Coarse scales are fast but hard to add up. The right choice depends on what the estimates are for, how much the team knows about the work, and who reads the numbers afterwards. You can try every scale below in a free planning poker room on Nextretro.
This guide compares the four scales most agile teams use: T-shirt sizes, Fibonacci, modified Fibonacci, and powers of 2. It includes a decision table, a way to convert T-shirt sizes into points, and the mistakes that make any scale less useful. All four are available as Nextretro decks, so you can compare them with your team without setting anything up.
If you have not run a session before, how planning poker works covers the voting rules that apply whichever deck you pick.
Why estimation scales use gaps
All four scales share one idea: the gaps between values grow as the values grow. That matches how uncertainty behaves. A team can usually tell a 1 from a 2, because both are small and familiar. Telling a 20 from a 21 is false precision; at that size, the unknowns swamp any difference of one point.
If an item feels like it sits between 8 and 13, the gap is telling you the item is uncertain. Discuss what makes it uncertain rather than inventing a 10.
T-shirt sizing in agile
T-shirt sizing uses XS, S, M, L, XL, and XXL. The labels are deliberately not numbers, which changes how people think about them. Nobody adds an M to an L, and nobody mistakes a size for a number of hours.
Where T-shirt sizes work well:
- Roadmap and epic level sizing. When the work is months away and loosely defined, rough buckets are all the information honestly available.
- Large backlogs that need a first pass. Sorting dozens of items into sizes is much faster than voting on precise points for each.
- New teams. Sizes are easier to agree on before the team has a shared set of reference items.
Where they fall short:
- Velocity and forecasting. You cannot sum sizes without converting them to numbers first.
- Distinguishing work within a sprint. If most sprint items land on M, the scale has stopped telling you anything.
- Tool fields. Trackers such as Jira and Azure DevOps store estimates as numbers, so sizes need converting before they go in.
Fibonacci story points
The Fibonacci deck in Nextretro uses 0, 1, 2, 3, 5, 8, 13, 21, 34, 55, and 89. Each value is the sum of the two before it, so each step is roughly 60 percent larger than the last. That spacing is the reason teams like it: small items get fine distinctions, large items get coarse ones.
Where Fibonacci works well:
- Sprint level estimation. Items the team will build soon, with enough detail to compare against past work.
- Velocity based planning. Points add up, so the team can compare a sprint's planned total with what it has completed before.
- Spotting items to split. Many teams agree that anything above a certain value, often 13, is too large to bring into a sprint without splitting.
Where it falls short:
- The top of the deck rarely gets used well. Votes of 34, 55, or 89 usually mean "too big to estimate," and the honest response is to split the item.
- Numbers invite misuse. Points can be converted to hours or compared across teams, neither of which they were designed for.
Modified Fibonacci
Modified Fibonacci keeps the shape of the sequence at the low end and rounds the large values: 0, ½, 1, 2, 3, 5, 8, 13, 20, 40, and 100. Many teams know this as the classic planning poker deck.
Two differences matter in practice. The ½ card gives a place for trivial changes, such as a copy fix, without calling them a full point. And the rounded 20, 40, 100 at the top signal that these are not precise estimates. Nobody believes the difference between 34 and 55 is meaningful; a 40 reads honestly as "big, and we are unsure."
Choose modified Fibonacci over plain Fibonacci if your team often has very small items, or if people take the precise looking large numbers too literally.
Powers of 2
Powers of 2 uses 0, 1, 2, 4, 8, 16, 32, and 64. Each value doubles the one before it, so the gaps grow faster than Fibonacci's.
Doubling is easy to reason about: "is this twice as big as that reference item?" The trade off is fewer middle options; there is nothing between 4 and 8, so the team has to make a clearer call. Some trackers offer an exponential scale with the same doubling pattern, which this deck matches.
Decision table: which scale fits your team
| Situation | Recommended scale | Why |
|---|---|---|
| Sizing epics or roadmap items months ahead | T-shirt sizes | Rough buckets match the uncertainty, and stakeholders read them easily |
| First pass over a large, unrefined backlog | T-shirt sizes | Fast sorting; convert or re-estimate later |
| Sprint level stories feeding velocity | Fibonacci or modified Fibonacci | Values add up and widen with uncertainty |
| Many tiny items such as copy and config changes | Modified Fibonacci | The ½ card handles trivial work |
| Debates keep stalling on adjacent values | Powers of 2 | Fewer middle options force a clearer choice |
| Tracker uses an exponential scale | Powers of 2 | Values match what the tracker allows |
| A brand new team without reference items | T-shirt sizes first, then Fibonacci | Build a shared sense of size before adding numbers |
Pick one, use it consistently for a few sprints, and review it in a retrospective before switching.
How to convert T-shirt sizes to story points
Many teams use T-shirt sizes for an early pass and story points for sprint planning. To move between the two, agree on a mapping once and write it down. A common starting point:
| T-shirt size | Story points (Fibonacci) | What it usually means |
|---|---|---|
| XS | 1 | A small, well understood change |
| S | 2 or 3 | A straightforward story with few unknowns |
| M | 5 | A typical story with some open questions |
| L | 8 | Larger work, worth checking whether it can be split |
| XL | 13 | Probably too large for one sprint as is |
| XXL | 21 or more | An epic; split it before estimating in points |
This is a convention, not a standard. The right mapping for your team depends on the reference items you already use. Calibrate it by taking two or three recently finished items, agreeing what size and what point value each had, and anchoring the table to those examples.
Two cautions when converting:
- Do not convert XL and XXL straight into sprint commitments. Those sizes say the work is too big and too uncertain to plan in detail. Split the item and estimate the pieces.
- Re-estimate when you know more. A size given months ago reflects what the team knew then. When the item reaches the top of the backlog, run a quick planning poker round on it with a numeric deck.
Common mistakes with any scale
- Treating points as hours. Points mix effort, complexity, and uncertainty. Converting them to time reintroduces the precision the scale removed.
- Comparing teams. A 5 on one team means something different from a 5 on another. Points are calibrated to a team's own reference items.
- Averaging instead of discussing. If votes are spread across the deck, the spread is the information. Ask the outliers to explain before voting again.
- Estimating the unestimable. If people keep reaching for the largest card or the ? card, the item needs splitting or refinement, not a bigger scale.
Try the scales with your team
Each Nextretro deck includes a ? card for "I do not understand this item yet" and a coffee card for "I need a break." Votes stay hidden until reveal, and the results show the average, the median, the level of agreement, and a suggested card.
A low effort experiment: estimate the same five recently completed items twice, once with T-shirt sizes and once with Fibonacci, and ask which conversation surfaced more useful information. Bring the answer to your next retrospective, and use a sprint planning board once you have settled on a scale.
If your backlog lives in a tracker, estimating Jira tickets with planning poker and the guide to Azure DevOps and Linear show how to import items by CSV and take the estimates back.
Frequently asked questions
What is T-shirt sizing in agile?
T-shirt sizing is a relative estimation technique that sorts work into sizes such as XS, S, M, L, XL, and XXL instead of numbers. It is fast and easy for non technical stakeholders to read, which makes it useful for roadmap planning and first passes over large backlogs. It does not add up, so it is less suited to velocity based planning.
Is Fibonacci better than T-shirt sizing?
Neither is better in general. Fibonacci suits sprint level stories where estimates feed velocity and planning, because the values add up. T-shirt sizes suit early, uncertain work where rough buckets are all the team honestly knows. Many teams use both: sizes for the roadmap and Fibonacci points for the next few sprints.
How do you convert T-shirt sizes to story points?
Agree on a mapping as a team and anchor it to recently completed work. A common starting point is XS as 1, S as 2 or 3, M as 5, L as 8, XL as 13, and XXL as 21 or more. Treat XL and XXL as signals to split the item rather than as sprint commitments.
What is the modified Fibonacci sequence?
Modified Fibonacci is 0, ½, 1, 2, 3, 5, 8, 13, 20, 40, and 100. It keeps the Fibonacci spacing for small values, adds a ½ card for trivial work, and rounds the large values so they read as rough rather than precise. It is a common choice for planning poker decks.
When should a team use powers of 2 for estimation?
Use powers of 2 when your tracker uses an exponential scale, or when debates between adjacent Fibonacci values keep stalling. Each value doubles the previous one, so there are fewer options in the middle and the team has to make a clearer call about relative size.