How to Score Consistently Against Defined Criteria
A score should make a clear claim: this evidence meets this standard. This playbook helps leaders keep the rating bar steady, explain each score with specific evidence, and handle borderline cases without improvising.
Developing
Start here. Build the foundation.- 1
Before you commit a score, match the evidence to the written level it meets. Read the criteria and ask which wording the record actually supports. The score is ready when you can point to the level, not just say the number feels right.
- 2
Attach specific observed evidence to every score. Name the moments from the record that justify the rating so the person can understand it and another observer can inspect it. A score with no evidence is a guess wearing a number.
Proficient
Build consistency and rhythm.- 3
Check whether the person or your state is moving the bar. Ask whether the same evidence would earn the same score from someone you like, someone you find difficult, and someone you are scoring at the end of a long day. Adjust back to the criteria.
- 4
Use one stated rule for borderline cases. For example, score to the lower level unless a defined threshold of evidence supports the higher one. The rule is working when similar edge cases receive the same call even weeks apart.
Mastered
Operate at the highest level.- 5
Create a scoring guide with worked examples for each level. Use real evidence from past observations, stripped of sensitive details, and show why each example earns its rating. Share it with other observers and revise it when new edge cases expose ambiguity.
Common Pitfalls
Avoid the common failure modes.- Scoring from overall impression first, then searching for evidence to justify the number. Start with evidence, not the answer.
- Letting how much you like someone change the height of the bar. Consistency matters most when the relationship makes consistency hard.
- Splitting the difference on every borderline case. If every hard call lands in the middle, your scores stop distinguishing real differences.