When teams start using planning poker, one of the first decisions they face is which deck to use. Fibonacci numbers (1, 2, 3, 5, 8, 13, 21…) and T-shirt sizes (XS, S, M, L, XL) are the two most popular options, and teams often pick one based on what they've seen elsewhere without understanding why. That's worth fixing — the right scale can meaningfully improve estimation accuracy and team buy-in.
The Fibonacci Sequence: Why It Works for Estimation
The Fibonacci sequence isn't arbitrary. Each number is roughly the sum of the two before it, which means the gaps between values grow as the numbers increase. Going from 1 to 2 is a 100% increase. Going from 13 to 21 is about 60%. This mirrors something true about software estimation: the larger and more complex a story, the harder it is to distinguish meaningfully between "this will take two weeks" and "this will take three weeks."
Fibonacci forces the team to make coarser distinctions at the high end of complexity — which is exactly right, because precision in large estimates is an illusion anyway. Trying to distinguish between a 17-point story and a 19-point story is wasted effort. Distinguishing between a 13-point story and a 21-point story represents a meaningful difference in scope or risk.
Fibonacci works best for:
- Technical teams comfortable with numerical comparison
- Teams that track velocity and want meaningful historical data
- Backlogs with stories that vary widely in complexity
T-Shirt Sizing: Intuition Over Precision
T-shirt sizes — XS, S, M, L, XL, and sometimes XXL — replace numbers with categories that feel more intuitive. Most people have a natural sense of what "small" versus "large" means in the context of work, even without a formal definition. This makes T-shirt sizing particularly accessible for mixed audiences.
The trade-off is granularity. With five or six categories instead of a continuous scale, T-shirt sizing is coarser. You can't say "this story is a 5 but that one is an 8 — let's pair on the 8." You can say "these are both Medium." The relative distinctions are fewer.
T-shirt sizing works best for:
- Teams with non-technical stakeholders in estimation sessions
- Early-stage backlog refinement where precision isn't yet warranted
- Teams new to agile who are still building estimation intuition
- Higher-level capacity planning across quarters rather than sprints
The Practical Differences in a Session
In a planning poker session, the format difference shows up in outlier discussions. With Fibonacci, a spread of 3, 5, 8, 13 tells you clearly that someone sees significant complexity others are missing — and the discussion that follows is about what specifically they're accounting for. With T-shirt sizes, a spread of S, S, M, L gives you similar information but slightly less granularity about where the disagreement lies.
For teams using ScrumTool's planning poker, both Fibonacci and T-shirt decks are available out of the box. You can also define a custom deck — some teams use modified Fibonacci sequences (0, 1, 2, 3, 5, 8, 13, 20, 40, 100) or pure powers of 2. Pick what resonates, stay consistent, and your velocity data will become meaningful over time.
Combining Both: The Two-Pass Approach
Some teams use T-shirt sizing for initial backlog triage and Fibonacci for sprint planning. In the first pass, stories get bucketed into rough size categories to determine which ones belong in the next few sprints. In the second pass, the shortlisted stories get proper Fibonacci estimates during the planning poker session.
This works well for large backlogs where the cost of Fibonacci-estimating every story is prohibitive. It keeps the estimation ceremony focused on work that's actually sprint-ready.
The One Thing That Matters More Than the Scale
Whether you use Fibonacci or T-shirt sizes, the most important thing is the conversation that happens when estimates diverge. The number on the card isn't the point — the discussion that surfaces different assumptions, risks, and requirements is. The scale is just the mechanism that makes disagreement visible.
Pick one, stick with it for at least six sprints so your velocity data is comparable, and invest in the outlier discussions. That's where estimation gets better. ScrumTool makes the session mechanics seamless so your team can focus on the discussion.