Published October 27, 2026.

Call coaching fails when the sample is chosen by whoever is available and remembers a bad call. The result is a meeting about one dramatic recording, a defensive dispatcher, and no change to the script. A defensible sample is selected before the review, covers routine work as well as failures, and produces a small number of changes that can be tested.

Define the review question first

Decide what the review is for. A new-client launch may need to test disposition accuracy. A spike in callbacks may need to test read-back and closure. A complaint may need to test a specific handoff. Write the question at the top of the review sheet so the sample and the rubric match it.

Do not combine a compliance audit, a coaching session, and a client performance review into one meeting. They need different samples and different audiences. Separating them makes each one more honest.

Build a stratified sample

Select calls across shift, dispatcher, client, channel, and outcome. Include routine calls that met the standard, calls that ended in a transfer or callback, and calls with an adverse event such as a missed page, a wrong disposition, or a complaint. Include unresolved calls, not only closed ones.

For a weekly review, ten to fifteen calls is usually enough to find a pattern. If the desk handles multiple clients, sample at least two calls per client or use a rotating client schedule. Record the selection rule and the random seed if you use one, so the sample can be reproduced.

Score with a small rubric

Use five dimensions: greeting and identity, issue capture, disposition accuracy, promise and close, and documentation. Score each as met, partially met, or not met, with a required note for any partial or missed score. Add one client-specific dimension when the account has a critical rule, such as a safety referral or a required approval.

Keep the rubric stable for a quarter. Changing the dimensions every week makes trend comparison meaningless. If a dimension is consistently irrelevant, retire it at the quarterly review rather than revising it mid-cycle.

Separate the person from the system

Before giving feedback, ask whether the defect was caused by the dispatcher, the script, the form, the phone system, or the client's rule. A missing gate code may be a form problem. A wrong disposition may be an ambiguous script. A missed callback may be a staffing gap.

Only the defects caused by the individual belong in individual coaching. System defects belong in a change log with an owner and a test date. This distinction keeps coaching from becoming a substitute for fixing the workflow.

Give feedback with the recording, not a transcript alone

Play the relevant segment, not the entire call. Ask the dispatcher what they heard and what they would do differently before giving your view. Then connect the observation to the rubric and to the caller's outcome. End with one specific behavior to practice on the next shift.

Avoid stacking five criticisms in one session. Pick the two that matter most for the review question. Write them down and check them in the next sample.

Track whether coaching changes anything

Compare the next sample's scores on the same dimensions. If the behavior did not change, the coaching may have been too vague, the script may be preventing it, or the dispatcher may need a different kind of practice. Try a role-play or a side-by-side call before repeating the same feedback.

Do not use call volume or average handle time as a proxy for quality. A dispatcher who closes calls quickly by skipping the read-back will look efficient and produce callbacks.

Report to the client with context

Share the review question, the sample size, the selection method, and the aggregate results. Do not send individual scores unless the contract requires them. Include the system changes made and the date they will be re-tested. A client who sees only a list of defects will assume the desk is not managing quality; a client who sees the method and the follow-up can judge the process.

Protect privacy in the review

Use the minimum recording segment needed. Mask payment card numbers and health information. Do not play recordings in an open room or share files in a group chat. Store coaching notes with restricted access and retain them per the desk's policy.

Implementation checklist

  1. Write the review question before selecting calls.
  2. Sample across shifts, clients, outcomes, and adverse events.
  3. Score with a stable five-dimension rubric.
  4. Classify each defect as person, script, form, system, or client rule.
  5. Coach with a short recording segment and one practice behavior.
  6. Re-sample to see whether the behavior changed.
  7. Report method and aggregate results to the client.
  8. Protect privacy in storage and playback.