Guide for managers

Call center coaching: a manager's guide

How to structure coaching on a live floor — which model fits which agent, how often to run sessions, what to actually score, and the metrics that show whether any of it worked. Written from how Silent Echoes™ customers run their floors.

See the real-time coaching tool Escalation guide

1. Pick a coaching model per situation

Most floors use one model for everyone and wonder why it works for half the team. Match the model to the gap.

GROW

Recurring 1:1s where the agent already knows the gap

Goal, Reality, Options, Will. The agent names the target, you both describe what actually happened on the call, list the ways forward, and close on one committed action. Best for motivation problems rather than skill gaps.

Call calibration

Keeping QA scores consistent across supervisors

Supervisors score the same call independently, then compare. Score variance above roughly one point on a five-point rubric means the rubric is ambiguous, not that the agent is inconsistent.

Side-by-side / live cueing

New hires and habits that only show up mid-call

Feedback lands while the behavior is happening, so the correction attaches to the moment instead of a memory of it. This is the model real-time tone coaching automates.

Peer benchmarking

Plateaued tenured agents

Compare an agent's delivery pattern to the top quartile on the same queue, not to an abstract ideal. Shows a concrete, reachable target.

2. Set the cadence

Group	Cadence	Why
Weeks 1–4 (new hire)	Two 20-minute sessions a week, one call reviewed each	Habits set fast. Correcting pace and openers early is cheaper than retraining at month three.
Ramped agents	One 30-minute session every two weeks, plus live cues daily	Enough contact to keep the rubric front of mind without pulling productive agents off the phones.
Top performers	Monthly, focused on one stretch skill	Over-coaching a strong agent reads as distrust. Give them one advanced target and their calls as calibration material.
Whole floor	Weekly 15-minute calibration huddle	Aligns supervisors on scoring and surfaces the objection of the week while it's still current.

3. Score delivery, not just script

Transcript-only rubrics miss the part customers actually react to. These six dimensions cover both.

Pace: Words per minute against the agent's own baseline, not a fixed number. Drift above baseline is the earliest reliable signal that a call is going badly.
Warmth / affect: Flat delivery on a compliant script still loses the sale. Prosody, not wording, carries most of this.
Talk-to-listen ratio: Long uninterrupted monologues correlate with disengagement. Track the longest single stretch, not just the average.
Objection handling: Was the objection acknowledged before it was answered? Recovery quality matters more than whether an objection appeared.
Disclosure compliance: For regulated queues, whether required disclosures landed before benefits were discussed. Binary, and non-negotiable.
Recovery: After a bad moment, how long until delivery returns to baseline. Fast recovery is a coachable, measurable skill.

4. Run the session in five steps

Open with the agent's own read: ask what they'd change about the call before you say anything.
Play or reference one specific 30-second moment — not the whole call, not a general impression.
Name the behavior, not the person: 'the pace climbed after the price objection', not 'you get nervous'.
Agree on exactly one change to practice before the next session. One, not three.
Write the commitment down where both of you can see it, and open the next session with it.

5. Add live cues between sessions

Scheduled sessions can only cover a handful of calls. Everything else goes uncoached unless something is watching delivery in the moment. That is what Silent Echoes™ does: a small on-screen dolphin reads pace, warmth and steadiness live and cues the agent — slow down, warm up, land the close — without a supervisor listening in. Nothing is recorded; only derived coaching metrics are kept.

Read the real-time coaching overview for the state machine and cue design, the escalation guide for severity tiers and hand-off scripts, and the compliance checklist before you put it on a regulated queue.

6. Measure whether it worked

Time-to-proficiency: Days until a new hire hits the team's median conversion or CSAT. The clearest ROI signal for a coaching change.
QA score variance between supervisors: Falling variance means your rubric is finally objective.
Percentage of call time in a steady delivery state: A leading indicator — it moves weeks before conversion does.
Repeat-defect rate: How often the same coached behavior reappears. Rising repeats mean the session format isn't landing.
Attrition at 90 days: Under-coached new hires leave. Coaching cadence shows up here before it shows up in revenue.

Five mistakes to avoid

Coaching the score instead of the call — agents learn to game the rubric.
Reviewing whole calls. Attention collapses after about two minutes of playback.
Batching a month of feedback into one session. The behavior is long gone.
Sampling only bad calls. Agents need a labeled example of their own good delivery.
Changing three things at once, so nothing can be attributed to anything.

Coach every call, not the three you sampled

Pay-as-you-go starts at $29/mo plus $0.25 per call minute, so you can trial it on one desk before rolling it to the floor.

See pricing For agency floors