Fibonacci vs T-Shirt Sizing: Choosing a Planning Poker Deck

Every planning poker session starts with a choice most teams make once and never revisit: which deck of cards to use. Fibonacci vs T-shirt sizing is the classic debate, with a plain 1–10 scale as the third option. The deck you choose shapes the conversation. It sets how precise people try to be, how fast you reach agreement and what the number means at sprint planning.

This guide explains what each deck is good at, where each one fails, and how to pick the right one for the work in front of you.

Why the deck matters more than it looks

Planning poker works because everyone picks a card in private and all cards are revealed at once. That stops the loudest or most senior voice from anchoring the room. The cards set the language of that vote:

• Gaps between values decide how much false precision people can express.
• Numbers vs labels decide whether you can calculate velocity and averages.
• Familiarity decides how quickly new people and stakeholders understand the result.

Pick the wrong deck and you'll either argue over a 6 vs a 7, or end up with sizes you can't plan a sprint with.

The Fibonacci deck (1, 2, 3, 5, 8, 13, 21)

The Fibonacci sequence is the default for story point estimation in Scrum teams. Each value is the sum of the two before it, so the gaps get wider as the numbers grow.

Why it works: we're good at comparing small things and bad at comparing big ones. Telling a 2 from a 3 is easy; telling a 14 from a 15 is guesswork. Fibonacci builds that uncertainty into the cards. Once a story is bigger than an 8, you only have 13 or 21 to choose from. Nobody can pretend to know the difference between 16 and 17.

Best for:

• Sprint-level backlog items your team will commit to
• Teams that track velocity in story points
• Mature teams with a shared baseline ("a 3 is about the size of the login bug we did last month")

Watch out for:

• Treating points as hours. A 5 is not "5 hours"; it's "bigger than a 3, smaller than an 8".
• Inflation. If 21s keep showing up, those stories are too big to estimate. Split them instead of scoring them.

T-shirt sizing (XS, S, M, L, XL, XXL)

T-shirt sizes swap numbers for labels everyone already understands. No one needs a lesson to know an L is bigger than an M.

Why it works: taking the numbers away takes away the urge to do maths. Discussion stays on relative size and risk, not on whether something is "really" a 5. That makes it ideal for early-stage work where precision would be fake anyway.

Best for:

• Roadmap and epic-level sizing, months ahead of delivery
• Backlog refinement when stories are still rough
• Sessions that include product, design or business stakeholders
• New teams that haven't agreed what "one point" means yet

Watch out for:

• No velocity. You can't add up an M and two Ls. If you need sprint capacity, map sizes to points later (for example XS=1, S=2, M=3, L=5, XL=8). Agree that mapping once, as a team.
• No average. A good tool shows you the spread of votes rather than inventing a number. In Votenir, the average only counts numeric cards, so a T-shirt round shows you who voted what instead.

The sequential deck (1–10)

A linear scale where every step is the same size. It's the most intuitive deck for people who've never played planning poker.

Why it works: there's nothing to explain. It suits work where items really are similar in size and you want finer steps between them.

Best for:

• Small, well-understood tasks (bug triage, support tickets, ops work)
• Scoring something other than effort, such as risk, complexity or confidence
• Workshops and one-off sessions with people outside the delivery team

Watch out for:

• False precision. The difference between a 7 and an 8 is rarely real, and long debates over one step are a sign you need a coarser scale.
• Clustering in the middle. People avoid the edges, so lots of 4s, 5s and 6s tell you less than you'd hope.

Side-by-side comparison

The two cards every deck needs: ? and ☕

Whatever scale you pick, keep two special cards:

• ? (I don't know). It's a signal, not a cop-out. If several people play it, the story needs more detail before anyone can estimate it. Stop and clarify rather than forcing a number.
• ☕ (coffee). It means "I need a break" or "I can't usefully estimate right now". Long estimation sessions wear people down, and tired estimates are bad estimates.

Neither card counts towards the average, and both are part of Votenir planning poker. We cover them in depth in The ? and coffee cards in planning poker.

How to choose: a quick decision guide

1. Is the work committed to in the next sprint or two? Use Fibonacci.
2. Is it an epic, a roadmap item or still a rough idea? Use T-shirt sizes.
3. Are the items small and similar, or are you scoring risk or confidence instead of effort? Use Sequential 1–10.
4. Are non-engineers voting? Lean towards T-shirt sizes.
5. Do you need velocity from this session? Avoid T-shirt sizes, or agree a points mapping first.

Many teams use more than one deck: T-shirt sizes in quarterly planning, Fibonacci in sprint planning. That's fine, as long as each session uses one deck from start to finish.

Best practices whatever deck you use

• Agree a reference story. Pick one finished story everyone remembers and give it a size. Every new estimate is "bigger or smaller than that". See Story points explained.
• Vote in private, reveal together. Hidden votes are the whole point. If people can see each other's cards, you're back to anchoring.
• Discuss the outliers, not the average. When one person plays a 3 and another plays a 13, ask both why. One of them knows something the other doesn't. More in What to do when estimates disagree.
• Re-vote once, then decide. If the second round still disagrees, take the larger value or split the story. Don't loop for twenty minutes.
• Keep a record. Saved rounds let you compare estimates with what really happened at your next retrospective.

Frequently asked questions