"Doubled" Sequence
The Doubled deck (1, 2, 4, 8, 16, 32, 64) uses powers of two — a scale that feels natural to engineers who think in binary. Each step is exactly double the previous, which means there are no fuzzy midpoints. In Fibonacci, you can debate whether something is a 5 or an 8 (a 60% difference). In Doubled, the equivalent gap is 4 versus 8 (100% different) — a more meaningful threshold. This makes Doubled well-suited for infrastructure and platform work where complexity genuinely doubles: adding one server is not a little harder than adding zero, it is categorically different work once you include networking, IAM, monitoring, and runbooks. The large end of the scale (32, 64) handles genuinely large engineering efforts without reaching for a "100" protest card.
Cards: 1, 2, 4, 8, 16, 32, 64.
How to use "Doubled" Sequence in planning poker
Calibrate the scale with a known reference at 4: "The last time we added Redis caching to a service was about a 4 — it took two days and had a few networking surprises." Estimate relative to that anchor. When votes split between two adjacent values (e.g. 4 and 8), treat that as a 2× disagreement — significant enough to warrant discussion, not averaging. A split on 16 vs 32 is a signal to consider splitting the story before committing. The deck works best for sprint-level estimation when the team has established shared references for each value.
Example tasks and point mappings
1: Fix a misconfigured alert threshold. 2: Add a health check endpoint to an existing service. 4: Add Redis caching layer to one service with monitoring. 8: Migrate a service to a new database schema with zero-downtime strategy. 16: Extract a module from the monolith into its own service. 32: Build a new internal developer platform with CI/CD, secrets management, and observability. 64: Rewrite a legacy system — this is probably a quarter-long effort, consider splitting.
When to consider a different deck
Mixed teams with non-engineers: the powers-of-two scale is not intuitive to product managers or designers. Feature-level user stories where the 5/8/13 Fibonacci range is more useful. New teams that need a simpler scale to get started.
Related reading