Quality and team

Fair Performance Reviews for Phone Agents: A Practical Guide

How to review phone agents fairly: choose measures they can influence, use samples of real calls, calibrate reviewers and turn results into coaching.

Call editorial team · Published · 5 min read

A performance review on a phone team has a built-in difficulty: the work is mostly invisible to the person doing the reviewing. A supervisor cannot sit in on every call, so they rely on numbers and samples. Both can mislead. A number can reflect the queue an agent was given rather than their skill, and a sample can reflect the three worst calls in a month. A fair review is one that deals with those weaknesses on purpose.

A supervisor and an agent reviewing results together on a laptop
Illustrative image

This guide sets out a way of running reviews that agents can trust and managers can defend. It builds on Quality Management for Calls and A Call Coaching Framework; read those if you need the basics of scorecards and coaching.

Measure what the agent controls

Start by separating things the agent controls from things they do not. An agent does not control how many calls arrive, how complicated the callers' problems are, or whether the knowledge base has the answer. They do control how they greet, listen, explain, record the call and follow the process.

Controllable and uncontrollable measures

Mostly in the agent's handsMostly outside it
Call qualityGreeting, listening, accuracy, tone, closingQuality of the scripts and tools provided
ProcessNotes, outcome recorded, correct transfersWhether the process is clear and the system works
PaceHandling time within the range of the workMix of simple and complex calls in the queue
AvailabilityKeeping to the rota and agreed breaksRota design and unplanned queue pressure
OutcomesOffering next stepsCustomer mood on arrival, product problems

Use numbers such as handling time and outcomes as prompts for questions, not as verdicts. A very short average call may mean efficiency or rushing; a very long one may mean thoroughness or confusion. Listen before concluding.

Use samples of real calls

Call samples give a review its substance. The principle is to review enough calls, chosen without bias, that the sample says something true about the agent. Pick a mix across days and times, include both short and long calls and avoid selecting only complaints. If a call is flagged for a reason, treat it as a special case, not as the typical sample.

A scorecard keeps reviewers consistent. List the behaviours that matter, such as opening the call properly, understanding the need, giving correct information, handling objections and closing with next steps. Give each item a weight and mark a small number as critical, such as handling personal data correctly. See QA Sampling for Call Reviews for how many calls to review.

A review cycle that works

  1. CollectGather the period's numbers and a mixed sample of calls.
  2. ScoreMark calls on the scorecard, noting evidence from the call.
  3. ComparePlace results beside the team range, not just against a target.
  4. DiscussShare results first, listen to the agent's view and play a call together.
  5. Agree actionsOne or two specific changes with a follow-up date.
  6. Follow upCheck the actions at the next review, and recognise progress.

Make reviewers consistent

Two reviewers scoring the same call will not always agree. If one reviewer is generous and another strict, agents are being judged partly by who happened to listen. Calibration sessions fix this: reviewers score the same call independently, compare results and discuss every difference until they agree what each item means. See Calibration Sessions for QA Teams.

Running the review conversation

The conversation matters as much as the data. Begin with what went well, using specific examples from real calls. Share the scores before the meeting where possible, so the agent can prepare. Ask open questions about what was difficult before you give your view; often the agent already knows the weak spots and can name the obstacle behind them, such as a missing answer in the knowledge base.

Watch for bias

Turn results into coaching

A review that ends with a score and nothing else wastes the effort. The useful output is a short plan: the one or two behaviours to change, an example of what good looks like, practice or support, and a date to look again. Keep the number of goals small; five goals means none get done. Link back to wider issues too. If many agents miss the same item, the problem is likely in training, scripts or knowledge, not in the individuals.

Before each review

  • Period and sample size agreed and applied equally
  • Calls chosen without bias
  • Scores shared before the meeting
  • One good call and one improvement call ready to play
  • Agent's workload and queue mix considered
  • Actions limited to one or two, with a date

Adapting reviews to different situations

One scorecard does not suit everyone equally. A new starter needs frequent, gentle feedback focused on the basics, with more calls reviewed in the first weeks. An experienced agent needs fewer reviews but more challenging conversations about development. A hypothetical remote agent may need extra attention to how connection quality, a shared home and tools affect calls, which are not performance problems.

Handle part-time and shift-pattern differences with care. A person who only works the busiest evening shift has a harder job than someone on the quiet morning shift, so compare like with like. Where a role mixes inbound and outbound, or different languages, score against the standard for each. Keep the principle constant: the same criteria, applied the same way, with context noted.

Questions to ask before judging a result

  • What queue and shift pattern did this person work?
  • Is the sample large enough to say anything?
  • Was the knowledge or tooling available to do the job well?
  • Is the result a pattern or a single bad week?
  • What did the agent say about it?

Running reviews in Call

Call has a Review and coaching page under AI agents. Admins and managers work through a review queue and score a call item by item, choosing Pass, Fail or N/A for each scorecard item and adding a comment the agent can read. Items marked critical set the whole score to zero if they fail. Further down the page you can see agent scores for the last 30 days, the scorecard items the team fails most often, calibration for calls scored by more than one reviewer, and coaching plans with an objective, action, coach and due date. Agents see their own evaluations and coaching sessions.

The Reports page also has an Agents tab with scorecards covering calls, connect rate, talk time, average call length and quality score over 30 days, plus the spread of call lengths. Read these as questions to ask, not conclusions. Call length data and quality scores together show more than either alone.

Frequently asked questions

How often should agents be reviewed?

Many teams review informally each week or two and run a fuller review each month or quarter. The right rhythm depends on team size and how fast the work changes.

How many calls should I review per agent?

Enough to be representative, chosen across days and call types. See your sampling plan; the exact number depends on volume and risk.

Should agents see their scorecards?

Yes. Showing the criteria and the scored calls lets agents prepare, challenge unfair scores and know what is expected.

Can handling time be a fair measure?

Only with care. Handling time depends heavily on the mix of calls, so use it to prompt questions rather than rank people.

What if two reviewers disagree on a call?

Discuss the difference in a calibration session and agree what each scorecard item means. Disagreement is useful information about unclear criteria.

What should a review end with?

One or two specific actions, an example of good practice, any support needed and a date to follow up.

Score calls and run coaching sessions in Call

In the demo console you can try incoming calls in a simulation.

Related guides