How do you run a Gong coaching session step-by-step in 2027?

The outcome you should expect

A Gong coaching session done well produces exactly one thing: a changed behavior that shows up in a later recorded call. That is the only outcome worth measuring. Everything else — the rapport, the note-taking, the shared feeling that the conversation was useful — is process. If you cannot point to a call three weeks later and say "that is the thing we worked on, and it is different now," the session did not land.

This framing matters because most conversation-intelligence coaching drifts toward call review rather than coaching. Call review is a manager narrating a recording while a rep listens. It generates agreement and no change. Coaching is the rep doing the analytical work out loud while the manager constrains the scope to one behavior and forces a commitment with a verification date attached. The difference is structural, not motivational, and it is why the step-by-step below front-loads rep self-assessment before any manager opinion enters the room.

The realistic per-session outcome for a rep who has been through four or five sessions on the same behavior is a measurable shift in a Gong-visible metric: talk ratio moving out of a bad band, longest-monologue dropping, question count rising, a specific discovery topic getting mentioned in calls where it previously never appeared. Gong surfaces these as call-level stats and as trackers — keyword and phrase trackers you configure yourself — so the behavior you pick should ideally be one the platform can count without you re-listening to everything. If the behavior is not countable, you will be verifying by ear, which is fine but does not scale past a team of about five.

Expect the first two or three sessions with any rep to produce no visible metric movement. The rep is still learning what the session is, still defending rather than analyzing, and still treating the clip as a performance review artifact. Sessions four through eight are where change typically shows. Managers who abandon the cadence at week three because "it isn't working" are quitting before the mechanism engages. Set the expectation with the rep and with your own leadership that the read on effectiveness comes at the 60-day mark, not the 14-day mark.

The second-order outcome, and the one RevOps usually cares about more, is a coaching signal that flows into forecast and enablement. When every session logs one behavior, one clip link, and one commitment, you get a structured record of what your team is actually bad at. Aggregate that over a quarter and you have an enablement roadmap built from evidence rather than from the loudest AE's complaint. This is why the logging step is not optional busywork — it is the entire reason the program produces organizational value rather than just individual value.

What you should not expect is that Gong does the coaching. The platform finds the calls, timestamps the moments, counts the patterns, and stores the scorecards. The judgment about which behavior matters for this rep at this stage of their development is a human decision, and it is the highest-leverage thing a frontline manager does. Treat the AI summaries and automatic topic detection as a search index over the call library, not as a verdict on rep quality.

What drives that outcome

Three inputs determine whether a Gong coaching session changes anything: call selection, scope discipline, and verification. Get all three right and the format almost runs itself. Miss any one and you get a pleasant meeting with no downstream effect.

Call selection. The default failure is coaching the worst call. Worst calls are usually contaminated by circumstance — a prospect who was hostile, a demo where the product broke, a discovery where the rep walked in with a bad lead. The rep will correctly attribute the outcome to circumstance and learn nothing. Better selection heuristics: pick a call that is *representative*, meaning it looks like the twenty calls before it; or pick a call that is *close*, meaning the rep almost did the thing right and the gap is small enough to be crossable. In Gong, build a saved filter that does this mechanically — call type equals discovery, duration over 20 minutes, deal still open, within the last 10 days — and pull from the filter rather than from memory. Memory selects the calls that annoyed you, which correlates poorly with the calls worth coaching.

Scope discipline. A 30-minute session can hold exactly one behavior. Managers routinely try to hold three, and reps leave with zero, because three simultaneous behavior changes under live-call pressure is not something adults can execute. The discipline is to notice the other two things, write them in your own notes, and say nothing about them. They are your backlog for sessions two and three. This is the single hardest habit for new managers because the other two problems are visible and saying nothing feels negligent.

Verification. A commitment without a verification date is a wish. The session ends with a specific next call identified — not "next week," but "your Thursday call with the Riverside account" — and a stated re-listen. The rep knows you will listen. That knowledge does more work than the coaching conversation did.

The loop matters more than any single session. A rep who gets one session and no follow-up reverts to baseline within two weeks, because the old behavior is the one that fires under pressure and nothing has replaced it in muscle memory. The re-listen closes the loop, and the fact that you retire a behavior only after seeing it in a live call is what keeps the program honest.

The pre-work is the leverage point. Fifteen minutes of manager preparation converts a 30-minute session from browsing into surgery. In practice that pre-work is: open the call, jump to the two or three moments Gong has already flagged (topic transitions, long monologues, pricing mentions, competitor mentions), scrub around them, and place a timestamped comment on the one you will play. Managers who skip pre-work end up scrubbing the timeline live, which burns eight minutes of a 30-minute session and signals to the rep that the session is improvised.

Benchmarks and realistic ranges

Session length: 30 minutes. Sixty-minute coaching sessions almost always become call reviews, because an hour is long enough to wander into three behaviors and a deal update. Thirty forces the single-behavior discipline. Some teams run 45 for new hires in their first 90 days, where there is genuinely more to cover and the rep needs more role-play reps.

Cadence: weekly per rep for new hires, biweekly for tenured reps. Weekly for everyone sounds better and collapses under manager load. A manager with eight direct reports running weekly 30-minute sessions plus 15 minutes of pre-work each is committing six hours a week to coaching alone, before pipeline reviews, forecast calls, and escalations. Biweekly for tenured reps brings that to roughly three and a half hours, which survives a bad week. Pick the cadence you can hold for a quarter, not the one that sounds most committed in a leadership meeting.

Clip length: 60 to 180 seconds. Under 60 seconds and the rep cannot hear the setup that produced the moment. Over three minutes and you are playing a call, not a clip, and the rep's attention decouples. If the moment genuinely needs five minutes of context, that is a signal you have picked a structural problem — a whole-call-flow issue — rather than a discrete behavior, and structural problems need a different session format.

Talk ratio. Gong reports talk ratio per call, and the widely referenced healthy band for discovery calls sits somewhere in the 40-to-50-percent range for the seller, with demos naturally running higher. Treat these as directional rather than as a target to optimize. A rep at 70 percent on discovery has a real problem. A rep at 48 versus 52 has no problem, and coaching to the decimal teaches reps to game the metric by staying quiet rather than to ask better questions. The useful version of this metric is longest monologue: if a rep's longest uninterrupted stretch on a discovery call exceeds three or four minutes, something has gone wrong regardless of the aggregate ratio.

Behaviors per rep per quarter: three to five. At biweekly cadence that is six sessions a quarter, and behaviors typically take two to three sessions each to stick. A manager claiming twelve coached behaviors in a quarter is logging topics, not changing behavior.

Time to visible metric movement: four to eight weeks. This is the number to set expectations against. Trackers and scorecards will show noise week to week; the trend needs a month of calls under it before it means anything.

Scorecard usage. Gong scorecards let you define criteria and score a call against them, and they are genuinely useful as a shared rubric — the rep can score themselves, you score independently, and the delta between the two scores is often more diagnostic than either score alone. A rep who scores themselves 4/5 on discovery where you scored them 2/5 has a calibration problem, and calibration is a different coaching target than skill. Do not, however, publish scorecard averages as a leaderboard. The moment scores become a ranking, reps optimize for the score and the rubric stops measuring the behavior.

Session logging: under two minutes. If logging the session takes ten minutes, managers stop doing it inside a month. Keep the record to four fields — rep, behavior, clip link, verification call — in whatever system you already use. RevOps should own making this cheap, because RevOps is the function that wants the aggregate data.

Risks, edge cases, and failure modes

The surveillance read. The most common way a Gong coaching program dies is that reps experience it as monitoring. This is a reasonable read — the platform records everything and the manager arrives with timestamps — and denying it makes it worse. The counter is behavioral: coach reps who are performing well, not just reps who are struggling. If the only people who get sessions are the bottom third, the session is a disciplinary instrument and everyone knows it. Run the same cadence across the team and the frame changes.

Coaching the outcome instead of the behavior. "You lost that deal" is not coachable. "You accepted the first stated timeline without testing it" is. The tell is whether the rep could have done the thing differently inside the call. Deal outcomes depend on budget, competitors, champions leaving, and a dozen things outside the call. Behaviors are inside the call. Stay inside the call.

The manager who talks for 25 of 30 minutes. This is the most frequent failure and the hardest to self-diagnose, because it feels productive. The fix is mechanical: the rep speaks first, for at least five minutes, before any manager assessment. Ask "what would you do differently" and then actually wait. The silence is uncomfortable and the discomfort is the point — reps fill it with genuine analysis if you let them.

Over-reliance on automatic detection. Gong's topic detection, AI call summaries, and automatic trackers are pattern matchers over transcripts. They are good at "was pricing mentioned" and bad at "was this a good discovery call." Trackers in particular fire on the keyword regardless of context — a tracker for a competitor name hits when the prospect dismisses that competitor and when the rep gets steamrolled by it. Always listen to the moment before coaching to a tracker hit. Managers who coach from the dashboard without opening the call produce confidently wrong feedback, and reps who get confidently wrong feedback stop bringing real problems to sessions.

Transcription and language edge cases. Accuracy varies with audio quality, accents, cross-talk, and industry jargon. Coaching a rep for "not saying X" when the transcript simply missed X is a trust-destroying error. When the evidence is a transcript absence rather than something you heard, listen before you assert.

Recording consent and jurisdiction. Call recording law varies by jurisdiction — some US states and many countries require all-party consent rather than one-party. This is a legal and compliance question, not a RevOps preference, and the answer determines which calls can even enter your coaching library. Get the policy from legal, make sure the consent mechanism is actually firing, and do not coach from a call that should not have been recorded.

Sensitive content in the library. Calls contain compensation discussions, health details, security incidents, and legal matters. Permission scoping in Gong matters, and so does manager judgment about which clips get shared into a team channel. A clip that is instructive and embarrassing is still embarrassing, and sharing it once will cost you candor for a year.

Remote and async drift. Sending a rep a clip with written comments and calling it coaching is a downgrade, not an efficiency. Async comments are excellent as reinforcement between sessions and poor as a substitute for one, because the rep cannot ask the follow-up question and you cannot hear the defensiveness that tells you the point did not land. Use async for the re-listen confirmation; keep the session live.

Small teams and the cadence trap. A manager with two reps can run weekly sessions comfortably and will, and the reps will be over-coached — new behaviors stacked before old ones set. Even with capacity, respect the two-to-three-session-per-behavior arc. Extra capacity should go into deeper pre-work and more role-play, not more behaviors.

Turnover resets the clock. Every new manager inherits reps mid-behavior with no record of what was being worked on unless the logging discipline held. This is the argument that finally convinces skeptical managers to log: it is not for RevOps, it is for their successor and for the rep who otherwise starts over.

A practical rollout plan

Rolling this out across a team is a four-to-six week sequence, and the sequencing matters more than the speed.

Week 1 — instrumentation and agreement. Confirm recording is capturing what you think it is: the right call types, the right integrations, and consent handled correctly. Build the saved filters managers will pull from, because a manager searching a raw call library will not do the pre-work. Define the scorecard rubric with the managers who will use it, not for them. Then tell the team what is happening and why, explicitly including the fact that coaching will run across all performance levels. Do this before the first session, not after someone asks why they were pulled into a call review.

Week 2 — manager calibration. Get every manager on one call and have them independently pick the one behavior they would coach. The spread will be wide the first time. Argue it out until the group converges on what "one behavior" means and what counts as coachable versus circumstantial. This single exercise does more for program quality than any documentation.

Weeks 3-4 — pilot with two managers. Full cadence, full logging, no exceptions. Collect what actually broke: sessions running long, pre-work getting skipped, logging taking too many clicks, reps showing up cold. Fix the mechanics while the blast radius is two managers.

Weeks 5-6 — full rollout and the first aggregate read. Everyone on cadence. At the end of week 6, pull the logged behaviors and look for clusters. If six reps are being coached on the same discovery gap, that is not six coaching problems — that is one enablement problem, and it belongs in a training session rather than in six separate 30-minute conversations. This is the moment the program starts paying RevOps back.

Ongoing governance. Once a quarter, review whether the cadence actually held. Sessions scheduled versus sessions run is the honest metric, and it will be lower than managers report. Also review whether behaviors are being retired — a log full of behaviors opened and never closed means verification is not happening, and verification is the load-bearing step. RevOps owns this instrumentation because managers will not audit their own follow-through.

What to give managers as an artifact. One page: the saved filter link, the 30-minute agenda with time boxes, the four logging fields, and three example behaviors written at the right altitude so managers can pattern-match. Do not ship a 20-page playbook. It will not be read, and the format is simple enough that it does not need one.

Related questions

How long should a Gong coaching session run?

Thirty minutes for tenured reps, up to 45 for new hires in their first 90 days. Longer sessions drift into call review and multi-behavior coaching, both of which reduce the odds any single change sticks.

Should the rep watch the call before the session?

Yes. Ask them to re-listen and come with one thing they would do differently. It converts the first five minutes from manager narration into rep self-assessment, which is where most of the learning happens.

Can Gong pick which call to coach automatically?

It can filter and flag — scorecard scores, tracker hits, talk-ratio outliers — but the judgment about which behavior matters for this rep now is human. Use the platform as a search index, not a verdict.

How do you coach without it feeling like surveillance?

Run the same cadence across all performance levels, coach behaviors rather than outcomes, and be explicit about the purpose up front. If only struggling reps get sessions, the session is discipline and reps will read it correctly.

What if the behavior does not change after three sessions?

Reassess whether the behavior is actually coachable or is a skill gap needing training, a motivation problem, or a role fit issue. Three failed sessions on the same behavior is a signal about diagnosis, not about rep effort.

FAQ

What is the single most important step in a Gong coaching session?

The rep speaking first. Before any manager assessment, the rep self-assesses for at least five minutes on what they would do differently. This step determines whether the session is coaching or narration, and skipping it is the most common reason sessions produce agreement without change.

How much pre-work does a manager need to do?

About 15 minutes per session. Open the call, jump to two or three flagged moments, pick the single behavior you will coach, and place a timestamped comment on the clip you will play. Managers who skip pre-work scrub the timeline live and burn a quarter of the session.

Should you use scorecards in every session?

Not necessarily, but they are useful when calibration is the issue. Have the rep score themselves and score independently yourself; the gap between the two scores is often more diagnostic than either. Avoid publishing scorecard averages as a leaderboard — reps will optimize the score rather than the behavior.

How do you verify a coached behavior actually changed?

Name a specific upcoming call at the end of the session, not a vague timeframe, and re-listen to that call. Retire the behavior only after seeing it present in a live recording. Without this step, reps revert to baseline within roughly two weeks.

Can this work for a fully remote or distributed team?

Yes, and Gong is arguably more valuable there since the manager cannot sit in on calls. Keep the session live over video rather than sending clips with written comments; async comments work well for reinforcement between sessions but poorly as the session itself.

What should RevOps own in this process?

The instrumentation: saved filters, scorecard rubrics, permission scoping, and a logging format that takes under two minutes. RevOps should also run the quarterly aggregate — clustering logged behaviors to find where a training session beats six individual conversations.

Sources

• https://www.gong.io/
• https://help.gong.io/
• https://www.gong.io/resources/labs/
• https://hbr.org/2016/11/the-most-effective-sales-coaching-happens-in-the-moment
• https://www.salesforce.com/blog/sales-coaching/
• https://www.saleshacker.com/
• https://www.forrester.com/blogs/category/revenue-operations/
• https://www.ftc.gov/business-guidance/privacy-security
• https://hbr.org/2015/05/how-to-give-feedback-people-can-actually-use

Related on PULSE

• How do you build a sales call scorecard that managers actually use?
• What talk ratio should a discovery call have?
• How do you set up conversation intelligence trackers without false positives?
• How often should a sales manager coach each rep?
• What does RevOps own in a sales enablement program?
• How do you measure whether sales coaching is working?