A protocol for measuring whether two reviewers assign the same tier to the same account using a written fit model.
Research question and scope
This protocol examines account tier agreement. It reports what a defined record can show in one window and does not claim an industry rate or a causal effect.
Population and unit of analysis
The unit is one account scored by two reviewers. Keeping one unit prevents linked events from being counted as independent successes or failures.
Definitions and measures
The main measure is agreement rate and the fields that drive disagreement. Report the denominator and time window beside every count so the result is not read as a bare percentage.
Evidence and data dictionary
Collect the fit fields, evidence dates, and reviewer notes. Label missing, unavailable, and confirmed negative as separate states rather than collapsing them into a blank.
Sampling and collection
The sample is accounts scored in one planning cycle. Record the observation date and exclude items that were already open before the window begins.
Analysis and reporting
Present agreement by segment and account size. Separate seller behavior from tool behavior whenever the evidence allows that distinction.
Limitations and sources of error
Limits include legacy lists and incomplete firmographic data. State them beside the finding and avoid turning a small operational sample into a universal standard.
Review cadence and next action
The practical next step is to clarify one tier rule that causes disagreement. Re-test a later sample after the change so the review produces a decision, not just a report.