Why Story Points Fail: Thought Experiment

Scenario: A Scrum Master is facilitating a team-building event. One of the activities is to teach the team how to estimate. The Scrum Master tasks each individual on the team to think of relative story point estimates for two items,

“How long does it take people to solve a Rubik’s Cube relative to how long it takes them to solve a 100-piece jigsaw puzzle?”

These puzzles are decent candidates for estimating; they cannot be broken down, have well-defined requirements, are familiar to most people, and are a solved problem. But I suspect that, for many people, it still isn’t enough to estimate.

The scrum master shares the historical data:

The team analyzes and summarizes:

• The historical averages are similar
• The fastest people can solve jigsaw puzzles about 18 times faster than the average
• The fastest people can solve Rubics Cubes about 3450 times faster than the average
• Only 6% of people can solve a Robic’s Cube.

Can they be estimated now?

The Problem:

Honest estimates are represented by a probability distribution. Honest representation of an estimate has values that a single number cannot represent: