Creating Custom Estimation Scales

Creating Custom Estimation Scales

While Fibonacci and T-shirt sizes work for most teams, some situations call for a custom scale tailored to your specific context.

When Standard Scales Fall Short

• Your backlog items are all very small (1-3 points), making Fibonacci gaps irrelevant.
• Your domain has natural categories that do not map to numbers (e.g., "config change," "new endpoint," "full feature").
• The team has a strong preference for a different system.

Common Custom Scales

Powers of 2: 1, 2, 4, 8, 16, 32

Similar to Fibonacci but with uniform doubling. Each jump is exactly 2x the previous. Good for teams that think in binary "is this twice as hard?"

Linear small: 1, 2, 3, 4, 5

For teams whose work items are uniformly small. If nothing ever exceeds 5 points, a full Fibonacci scale adds unnecessary complexity.

Risk-weighted: Low, Medium, High + numeric

Low Risk  + Small Effort  = 1
Low Risk  + Medium Effort  = 3
Low Risk  + Large Effort   = 5
High Risk + Small Effort   = 5
High Risk + Medium Effort  = 8
High Risk + Large Effort   = 13

This two-dimensional approach makes risk explicit rather than silently folded into a single number.

Task-type buckets

Config/copy change    = 1
UI tweak              = 2
New API endpoint      = 5
Full feature (FE+BE)  = 8
Cross-service feature = 13

Useful when the team's work falls into predictable categories.

Design Principles

1. Keep it small -- 5-8 values maximum. More options slow decisions.
2. Ensure gaps grow -- Bigger items need bigger gaps to prevent false precision.
3. Include a "too big" value -- The highest value should signal "split this story."
4. Document it -- Write down what each value means. New team members should be able to understand without tribal knowledge.
5. Revisit quarterly -- As the team evolves, the scale may need adjustment.

Using Custom Scales in This Tool

Select the Custom scale option, then add your values one by one. The tool saves your custom scale to localStorage so it persists between sessions.