How Agents Should Dispute a QA Score

If an agent cannot appeal, the score is a verdict from a person they may not trust. If every agent appeals every point, analysts do nothing else. The process below keeps both sides honest.

What may be disputed

A score on a written criterion, where the agent believes the call met the line as written. Not the existence of the criterion. Not the weight. Those are a governance conversation for the scorecard owner, taken up once, not relitigated on every form.

What the appeal must contain

The criterion. The timestamp. One sentence on why the call meets the written line. “I felt I did well” is not an appeal. “At 02:14 I read the disclosure in the criterion, and it was marked missed” is an appeal.

Who decides

Not the analyst who gave the original score, alone. A second evaluator, or the QA lead, scores that line against the criterion without a discussion first. If your calibration is healthy, this is quick. If it is not, the dispute queue is telling you to fix calibration, not to hire more analysts.

The clock

Agents have a fixed window — five working days is common — and so does the reviewer. A score that can be changed in perpetuity is not a record. Overturns are visible on the original evaluation. The analyst is not punished for an overturn on an ambiguous line. The line gets rewritten.

What you publish

The number of appeals, the overturn rate, and the criteria they cluster on. A high overturn rate on one line means the criterion is unclear. A high appeal rate and a low overturn rate means agents do not yet believe the evidence. Those are different problems. Do not hide the queue to keep the dashboard calm.

This is also an employment-record issue. In South Africa, a warning that leans on QA scores needs the score to be explainable. An agent who was refused the clip will say so later. Keep the evidence with the score.

Programmes that skip this step lose the floor, and then they lose the data, because people perform for the form instead of the call. The best-practice list puts disputes next to calibration for that reason.