How do you coach reps using Gong or Chorus call recordings?

What call-recording coaching actually is, and why RevOps cares

Conversation intelligence is two products wearing one login. The first is a *search index over every sales conversation your team has ever had* — transcribed, speaker-separated, keyword-tagged, timestamped, and joined to CRM opportunity records. The second is a *coaching workspace* — clips, comments, scorecards, playlists, and shared libraries. Most managers buy the first and never use the second, which is why so many Gong and Chorus deployments look busy and produce nothing. The recordings pile up. Nobody watches them. Renewal comes around and someone in finance asks what the platform actually changed.

The distinction that unlocks the tool is the difference between aggregate signal and specific moment. Aggregate signal is anything computed across dozens or hundreds of calls: talk-to-listen ratio, longest monologue, question count, patience (how long a rep waits after asking), tracker firing rates, topic duration. Aggregate signal answers *what skill should I coach?* It is statistically honest because it washes out the one weird call where the buyer talked for forty minutes about their boat. Specific moment is a 45-to-90-second segment where a rep did something you can name. Specific moment answers *how do I coach this person, today, in a way they will accept?* It is emotionally honest because the rep cannot argue with the tape.

Managers who conflate the two get one of two failure modes. They coach aggregates — "your talk ratio is 68%, work on that" — which is unactionable, because nobody knows how to *do* a percentage. Or they coach moments without aggregates — "you talked over the CFO on the Acme call" — which reps correctly dismiss as cherry-picking. The method is to fuse them: the aggregate proves the pattern is real and not personal, the clip proves what the pattern looks like in the rep's own voice, and the drill converts it into a behavior they can execute on Tuesday.

RevOps sits underneath all of this whether the title exists on your org chart or not. Somebody has to decide which trackers get built and who maintains them. Somebody has to ensure the recording platform's opportunity mapping matches the CRM's, or every "win rate by behavior" chart silently lies. Somebody has to define what counts as a discovery call versus a demo versus a check-in, because the benchmark bands differ enormously by call type and a single blended number is worse than no number. Somebody has to decide retention, consent geography, and who can see whose calls. When conversation intelligence fails, it usually fails on plumbing that a sales manager was never staffed to own — which is exactly the seam where RevOps earns its keep.

There's a second reason to care that shows up on the finance side. Ramp time is one of the few sales metrics you can genuinely compress with tooling, and call libraries compress it more than almost anything else. A new hire who can search "how did we handle the security-review objection" and watch six real examples from the six people who handled it best is learning from the compressed experience of the whole team instead of from whoever happens to sit next to them. That is a materially different onboarding than a slide deck and a shadow week. It is also measurable: track calls-to-first-qualified-opportunity and calls-to-first-close for cohorts before and after you build a real library, and you have a defensible number for the renewal conversation.

Adjacent teams get pulled in fast once the habit exists. Product marketing mines competitor-mention clips to figure out which battlecard claims survive contact with a real buyer. Product management listens to feature-request moments in the buyer's own words rather than in a rep's paraphrase. Customer success runs the identical loop on renewal and QBR calls, where the coaching targets are different — expansion signals, risk language, executive presence — but the machinery is the same. Support and implementation teams that record calls find the same value in escalation handling. The coaching method described here generalizes cleanly to all of them; only the trackers and the benchmark bands change.

The step-by-step coaching loop, from tracker setup to Friday check

Start with instrumentation before you start with people. If your trackers don't reflect your actual playbook, every downstream number is noise, and reps will feel graded by a rubric nobody explained to them.

Step one — define six to ten trackers, no more. A tracker is a phrase-matching rule that fires when a call contains certain language. Build them for the moves your playbook actually requires: securing next steps, naming a decision process, multi-threading ("who else needs to weigh in"), pricing discussion, competitor mentions, and whatever methodology language you've standardized on. Ten is a soft ceiling. Every tracker beyond that dilutes attention and adds maintenance debt, because language drifts and each rule needs periodic tuning. Write each tracker with at least eight to fifteen phrase variants and then validate it by hand: pull twenty calls the tracker fired on and check how many are true positives. Under about 70% precision and reps will stop trusting the dashboard, which poisons everything built on top of it.

Step two — set baselines by call type, not in aggregate. Segment discovery, demo, and negotiation calls separately. On discovery, a rep talking 45–55% of the time is a reasonable target band; Gong's published research has consistently associated lower rep talk ratio on discovery with better outcomes. On demos, rep talk naturally runs much higher and coaching to 50% would be actively harmful. Capture each rep's 30-day baseline per call type before you coach anything, because without a before-number you cannot prove a change later and you will end up arguing about vibes.

Step three — diagnose before you open a single recording. Decide what you're testing. A rep missing quota with a 68% discovery talk ratio has a different problem than a rep at 40% whose next-steps tracker fires on 30% of calls. The recording is evidence, not the diagnosis.

Step four — pull two or three clips, not full calls. Search by tracker plus rep plus date range, jump to the timestamp, trim 45 to 90 seconds around the moment. Budget 20 minutes of your prep time per rep per week. If prep takes an hour you'll skip it by week three.

Step five — run the 1:1 on self-discovery. Fifteen minutes, three blocks. Two minutes on the aggregate ("across your last ten discovery calls, you're at 68%; the band is 45–55%"). Eight minutes on clips — play one, stop, ask "what do you notice?", then say nothing. Silence is the coaching. Most reps name the miss before you do, and a gap they name themselves is one they'll actually work on. Five minutes to co-create one behavior for the week, stated concretely enough to count: "I'll ask 'who else needs to weigh in before this moves forward' on every discovery call."

Step six — drill it before it goes live. Replay the exact 60 seconds they fumbled, pause, and run the moment live: "I'm the CFO. I just said send me a proposal. Go." Three reps until it's automatic. Watching is not practicing.

Step seven — close the loop Friday. Pull the same tracker for the same rep and look at the firing rate together. This step is the one everyone skips and it is the one that makes it coaching instead of a conversation. Without a check, you've had a nice chat about a video.

Costs, timelines, and what a realistic rollout looks like

Conversation intelligence is priced per recorded seat per year, usually with a platform fee layered on top, and both Gong and Chorus quote rather than publish. Plan on a meaningful per-seat annual number and a minimum seat count — this is not a tool you buy for three people. Ask specifically about the platform fee, whether it scales with seats, and whether integrations you need (CRM sync depth, dialer, video conferencing, engagement platform) sit inside the base or behind an add-on. The other lever is who counts as a recorded seat: managers who never take calls but need to review them are sometimes priced differently, and expanding to CS or support later is a separate conversation. Get multi-year pricing in writing if you expect headcount growth, because per-seat costs on renewal after a big hiring year are where budgets break.

Timeline runs longer than the sales cycle implies. Technical setup is genuinely fast — connect the calendar, conferencing, dialer, and CRM and you'll have recordings flowing within days. What takes time is everything after. Budget roughly two weeks for tracker design and validation, another two to four weeks to accumulate enough call volume for stable per-rep baselines, and then the real work: changing manager behavior. A useful rollout shape is 30/60/90.

Days 1–30: instrument and normalize. Trackers built and hand-validated. Recording consent and disclosure handled properly — announce it to the team, explain what you'll use it for, show them the dashboard they can see about themselves. Baselines captured per call type. Coach exactly one thing team-wide, in a group setting, so everyone learns the format without anyone feeling singled out. Do not run individual scorecards this month; you'll get the tone wrong before people trust the tool.

Days 31–60: individual coaching. Every rep gets a weekly 15-to-30-minute session on one behavior with two or three clips and a written commitment. Managers keep a simple log: rep, behavior, clip links, commitment, Friday result. Six to eight direct reports is about the ceiling for this cadence; past that, coaching quality degrades or the manager's calendar collapses.

Days 61–90: shift ownership. Reps self-select a best call and a worst call weekly and self-score before the 1:1. Team huddles run peer review on one anonymized clip. The manager's job moves from finding the gaps to validating the ones reps found. This is where the time cost drops and the program becomes sustainable.

Expect first behavioral movement — a tracker firing rate climbing, a talk ratio drifting into band — in four to eight weeks of consistent weekly coaching. Expect pipeline and win-rate effects to lag by roughly a sales cycle plus a quarter, which for most B2B teams means you're looking at two quarters before an outcome number is trustworthy. Anyone promising win-rate lift in six weeks is either selling something or measuring a coincidence.

The honest cost that never makes the business case is manager time. Twenty minutes of prep plus a 15-to-30-minute session plus a five-minute Friday check is roughly 45 minutes per rep per week. At seven reps that's five-plus hours weekly — a real chunk of a frontline manager's capacity. If you don't explicitly protect that time and take something else off their plate, the program dies quietly around week five and everyone blames the software.

Where teams get it wrong

Binge-watching whole calls. A manager who sits through three 45-minute recordings has spent two hours and can coach maybe one thing per call, badly. Search, trackers, and timestamps exist so you can jump straight to the twelve moments that matter across thirty calls. If your review workflow starts with pressing play at 0:00, you're using a search engine as a DVR.

Coaching five things at once. The rep fixes zero. One behavior per week, stated so concretely you can count it. Vague commitments — "be more consultative" — cannot be checked on Friday, which means they cannot be coached.

Weaponizing the scorecard. The instant reps believe the tool is a surveillance dashboard, they start performing for the recording: reciting tracker phrases mechanically, saying "what's our next step" as a checkbox rather than to actually secure one. Your numbers improve, your calls get worse, and you've built a metric that lies. Frame it as their film room, share dashboards openly, let them self-review first, and never surface a clip in a group setting without asking the rep beforehand.

Chasing tracker firing rate as the goal. Trackers are a proxy for a behavior, not the behavior. Pair every behavior metric with an outcome metric — next-steps tracker rate alongside stage-to-stage conversion, competitor tracker alongside competitive win rate. If the proxy moves and the outcome doesn't, your diagnosis was wrong and more of the same coaching will make it worse.

Trusting AI-flagged moments without listening. Both platforms surface key moments and auto-summaries, and both are genuinely useful for cutting review time. But keyword matching misses tone. "Our pricing is competitive" flagged as pricing handling could be a rep who's defensive, curious, or genuinely collaborative — three different coaching points from one flag. Use AI highlights to narrow the haystack, then listen to the 90 seconds around the flag before you coach it.

Rescuing the rep. Telling them the answer is faster and teaches nothing. The gap a rep finds themselves is the one that changes behavior. Ask; don't tell — even when the silence is uncomfortable, and it will be.

Coaching the deal instead of the skill. Saving today's deal feels productive and is sometimes necessary, but it's deal strategy, not coaching. Run those separately, and in the coaching session always ask "what's the repeatable move here?"

No follow-through. Coaching without a check on the same behavior a week later is a conversation about a video. Close the loop or don't hold the meeting.

Uniform treatment. A ramping SDR needs clip-heavy, high-frequency, foundational coaching. A top performer needs a stretch challenge and mostly to be left alone. Applying the same cadence to both wastes your time on one and insults the other.

Ignoring the data quality layer. If calls aren't mapping to the right opportunities, or half the team's calls aren't recording because of a dialer misconfiguration, every aggregate is wrong. Audit coverage monthly: what percentage of expected calls actually got recorded and correctly attributed? Under 80% and you're coaching a biased sample.

Choosing what to coach, and when to stop coaching calls

Not every performance problem is a call-skill problem, and conversation intelligence is at its most valuable when it tells you to stop looking at calls.

The first fork is volume. If a rep isn't having enough conversations, nothing about call quality matters yet. That's an activity, territory, or pipeline-generation problem. Coaching their objection handling while they run four discovery calls a month is fixing the paint on a car with no engine.

The second fork is skill versus will. Recordings separate these better than any other instrument. If the behavior never appears — the multi-thread question simply isn't asked — that's typically skill or knowledge, and it responds to clips plus drills. If the behavior appears when the rep is being observed and vanishes otherwise, that's will or habit, and drills won't touch it; that's a conversation about priorities, confidence, or motivation.

The third fork is skill versus system. Sometimes the calls are good. The rep asks the right questions, secures next steps, handles the competitor cleanly, and the deals still die. When aggregate call quality looks healthy and outcomes don't follow, stop coaching delivery and go upstream: lead fit, qualification criteria, an SDR-to-AE handoff that loses context, pricing that's genuinely uncompetitive in that segment, or a product gap. Call coaching cannot fix a broken funnel, and continuing to coach into a system problem is how you lose a good rep who correctly senses they're being blamed for something outside their control.

The fourth fork is individual versus team. Look at the distribution, not the average. If 70% of the team sits outside the discovery talk-ratio band, that's a playbook or enablement problem and it belongs in one 30-minute group session, not seven separate 1:1s. Reserve individual coaching for outliers on both ends. The middle of the distribution improves substantially from group work plus self-review, and treating it that way is what makes the cadence survivable past eight direct reports.

There's also a segment dimension worth naming. A transactional SMB motion with 20-minute calls and a two-week cycle wants coaching on speed, disqualification, and next-step discipline — high frequency, short clips, fast feedback. An enterprise motion with 60-minute multi-stakeholder calls wants coaching on executive presence, multi-threading, and business-case articulation — lower frequency, longer clips, more deal context. Same tool, materially different trackers and benchmark bands.

Adjacent uses that make the same recordings pay for themselves twice

Once the library exists, the coaching use case is the smallest thing it does.

Onboarding libraries. Curate 15 to 25 clips into a structured playlist: three great discovery openings, three security-objection handles, three pricing conversations, three competitive displacements, three closes. A new hire watching real calls from your actual top performers ramps on your actual motion rather than a generic training module. This is the highest-leverage artifact a sales enablement function can build and it costs a few hours of curation.

Message testing for product marketing. Search competitor mentions across a quarter and read what buyers actually say versus what the battlecard claims they say. Search a feature name and hear the words buyers use to describe the problem it solves. This is qualitative research you're already paying for and mostly throwing away.

Voice-of-customer for product. A feature request relayed through a rep's summary has been through two rounds of translation. The 90-second clip of the buyer explaining why they need it, in their words, with their frustration audible, moves a roadmap conversation in a way a Jira ticket never will.

Forecast and deal inspection. Both platforms tie conversations to opportunities, which lets you ask uncomfortable questions with evidence: has anyone from our side spoken to someone in finance on this six-figure deal we're calling commit? Have we had a single conversation with an economic buyer? Deals with a single contact and no executive conversation are risk, and the recording data proves it rather than asserting it.

Customer success and renewals. Run the same loop on QBRs and renewal calls with different trackers — expansion language, risk signals, competitive re-evaluation, executive engagement. The coaching mechanics are identical; only the playbook changes.

Cross-functional enablement. Support escalation calls, implementation kickoffs, and partner conversations all benefit from the same clip-plus-drill loop. Teams outside sales are usually coached far less systematically, so the marginal gain from applying this method there is often larger than the gain inside sales.

The RevOps discipline that makes all of this work is unglamorous: keep call-to-opportunity mapping clean, keep trackers tuned as language drifts, audit recording coverage monthly, set retention and consent policy deliberately across the geographies you sell into, and report on a small stable set of metrics rather than a new dashboard every quarter. Do that and the platform compounds. Skip it and you have an expensive archive nobody opens.

Related questions

How long should a coaching session be?

Fifteen to thirty minutes weekly, per rep. Two minutes on aggregate data, eight to twenty on clips and self-discovery, five on the commitment. Longer sessions cover more ground and change less behavior — the constraint is what a rep can execute next week, not what you can say.

Do reps need to know their calls are reviewed?

Always. Disclose recording, explain what you use it for, show reps the dashboards about themselves, and let them self-review before you review them. Hidden review destroys trust permanently, and consent requirements vary by jurisdiction, so treat it as a compliance question too.

Does Gong versus Chorus change the coaching method?

No. Both give you trackers, talk-ratio analytics, timestamped clips, and CRM-linked calls — everything the loop needs. Chorus sits inside the ZoomInfo data stack; Gong leans harder into deal and forecast intelligence. Pick on integrations and price; the human coaching loop is identical.

Can AI summaries replace listening to calls?

They replace searching, not listening. Use AI-flagged moments and summaries to cut review time by half or more, then listen to the 90 seconds around each flag. Tone, hesitation, and intent are exactly what keyword-based flagging flattens, and they're what you're coaching.

What if a rep's calls look great but they still miss quota?

Stop coaching calls. Look upstream at lead fit, qualification criteria, and handoff quality, and downstream at pricing, competitive position, and territory. More call coaching against a system problem burns the rep's trust and your credibility without moving the number.

FAQ

How many calls should I review per rep each week?

Two or three clips of 45 to 90 seconds each, tied to one behavior — not full recordings. Three focused moments produce more behavior change than one complete 45-minute call, because the coaching is specific and the rep can act on it. Budget about 20 minutes of prep per rep. If your prep routinely runs an hour, your search and tracker setup needs work, not your discipline.

What talk-ratio should I coach toward?

On discovery calls, roughly 45–55% rep talk is a defensible target band, and Gong's published research has repeatedly linked lower rep talk on discovery to stronger outcomes. Demos and negotiations run higher by design — coaching a demo to 50% would be counterproductive. Always segment by call type before comparing anyone to a benchmark, and treat the band as a diagnostic prompt rather than a grade.

How many trackers should we build?

Six to ten, mapped to the playbook moves you actually require. Give each one eight to fifteen phrase variants and hand-validate precision on twenty calls before trusting it. More trackers means more maintenance as language drifts and more diluted attention. If reps catch the dashboard being wrong, they stop believing any of it, and you'll spend months rebuilding that trust.

How do I coach at scale past eight direct reports?

Split the work. Run one 30-minute weekly group session on the single skill the aggregate data says most of the team is missing, using one anonymized strong clip and one gap clip. Reserve individual 1:1s for the two or three outliers at each end. Push the middle toward self-scoring and peer review in huddles. This typically cuts manager coaching hours substantially while improving team-wide adoption.

When will we see results?

Behavior metrics — tracker firing rates, talk ratio moving into band — respond in four to eight weeks of consistent weekly coaching. Outcome metrics lag by roughly one sales cycle plus a quarter, so plan on two quarters before win rate or conversion tells you anything trustworthy. Track both, and if behavior moves while outcomes don't, revisit the diagnosis rather than adding more coaching.

Who should own the platform inside the company?

RevOps or enablement should own configuration — trackers, CRM mapping, coverage audits, retention and consent policy, reporting. Frontline managers own the coaching cadence. When sales managers own configuration too, trackers rot and the data quietly stops being trustworthy; when RevOps owns the coaching, it becomes a compliance exercise nobody values. Split it deliberately and revisit tracker health quarterly.

Sources

• Gong — Conversation Intelligence
• Gong Labs — Sales Talk Ratio Research
• Chorus by ZoomInfo — Conversation Intelligence
• Harvard Business Review — The Best Sales Managers Don't Chase Revenue
• RAIN Group — Sales Coaching Best Practices
• Performance Consultants — The GROW Model
• Salesforce — What Is Sales Coaching
• Winning by Design — Resources and Frameworks

Related on PULSE

• [How do you coach a rep using their own call recordings in a 1:1?](/knowledge/cg0117)
• [What is the most common mistake you see in your own demo recordings?](/knowledge/cg0948)
• [Top 10 Gong Coaching Review Prompts for Ramping Reps](/knowledge/cg0784)
• [Top 10 Gong Coaching Review Prompts for SMB Reps](/knowledge/cg0782)
• [Top 10 Gong Coaching Review Prompts for Mid-Market Reps](/knowledge/cg0781)
• [Top 10 Gong Coaching Review Prompts for New Hires](/knowledge/cg0783)