Card-Based Estimation

Purpose

Card-Based Estimation combines expert opinion, analogy, and structured discussion to produce reliable relative estimates. By revealing cards simultaneously the technique prevents anchoring bias and ensures every voice is heard.

At a glance

How to run

1. Introduce the scale (5 min) — Agree on the number sequence (Fibonacci: 1, 2, 3, 5, 8, 13, 21, or T-shirt mapped to points). Remind the team the numbers represent relative effort, not hours.
2. Present the item (3 min per item) — Product Owner reads the user story and acceptance criteria. Team asks clarifying questions.
3. Private estimation (1 min) — Everyone selects a card face-down.
4. Simultaneous reveal — All cards flip at once.
5. Discuss outliers (3 min) — The highest and lowest estimators explain their reasoning. The goal is shared understanding, not averaging.
6. Re-estimate if needed — Repeat the round until the team converges (usually within 2 rounds).
7. Record and move on — Write the agreed estimate on the item and proceed to the next.

Materials needed

• Estimation card decks (one per person) or a digital tool (Scrumpoker Online, Parabol)
• Product Backlog items with acceptance criteria
• Reference story (a previously estimated item everyone understands)
• Timer

Pitfalls & common mistakes

• Anchoring — If someone calls out a number before the reveal, it biases the group. Always flip simultaneously.
• Debating too long — Cap discussion at 5 minutes per item; if no convergence, take the higher estimate or defer.
• Estimating in hours — The scale is relative; mixing in clock time defeats the purpose.
• Skipping the discussion — The estimate itself is less valuable than the conversation it triggers.

Inclusion & accessibility

• Use large-print cards or a digital tool with screen-reader support.
• Give everyone equal time to speak during outlier discussion.
• For remote teams, use a tool that hides votes until all are in.

Variations

• Thumb voting — Quick fist-of-five for low-complexity items to save time.
• Silent grouping — Place items on a relative scale without discussion first, then discuss only disagreements.
• Bucket estimation — For large backlogs (50+ items), sort cards into buckets rather than estimating one by one.

When NOT to use this

• If the team has fewer than 3 people — just discuss and agree.
• If the backlog is too large (50+ items) — use Affinity Estimation or T-shirt Sizing instead.
• If the team already has strong throughput data and does not need story points.

Source & further reading

• Cohn, M. (2005). *Agile Estimating and Planning*. Prentice Hall.
• Grenning, J. (2002). "Planning Poker." renaissancesoftware.net