A dashboard label is not ground truth. To assess qualification quality, define the fields being reviewed, draw a traceable sample and compare each recorded value with the call evidence and the business rule that applied at that time.
The method below describes a measurement process. It reports no HeyRik customer performance or benchmark.
Define the unit and reference
- 1Choose a record-level field, such as callback preference or requested product.
- 2Write what counts as supported, contradicted, missing or not reviewable.
- 3Link each sampled record to a call or source reference and the rule version used.
- 4Have a trained reviewer apply the rubric; resolve disagreements before reporting.
- 5Keep a correction log and retest a later sample after changes.
Report more than one percentage
Show the sample size, date range, inclusion rule, reviewer agreement, missing-evidence share, correction rate and the categories of error. A single “accuracy” percentage can hide a small sample, a difficult language subset or a high-impact error. Report confidence limits when the sample design supports them; do not extrapolate a convenience sample to all calls.
Separate system output, reviewer judgement and downstream business outcome. A correctly captured preference does not prove that a human called back or that the lead became a customer.
Worked review log
- Record ID and evidence link.
- Expected rule and rule version.
- Observed value and evidence excerpt or timestamp.
- Reviewer decision plus reason.
- Correction owner, due date and verified closure.
- Whether the issue changes caller handling or reporting.
A little more before you begin.
Can a reviewer’s judgement be treated as objective truth?
Not automatically. Define the rubric, train reviewers, measure disagreement and document how disputes are resolved.
Start here
Design a reviewable qualification scorecard
Build a voice agent, upload your list or connect your leads, and run your first campaign on HeyRik — free to start, no credit card required.