A VP Sales I work with sent me a Slack message last quarter. "Our coaching program is a joke. I sit with each AE for an hour every Friday, we listen to one call, I tell them what to do better, and the next week the exact same thing happens on the next call. I am the bottleneck and nothing is changing."
He is not unusual. He runs a 14-rep team at a Series B SaaS company. The coaching he was doing was real work, eight hours a week of his time, and the team's win rate had moved by less than two points in a year. Reps liked him. The numbers did not care.
This is the gap I see in almost every B2B sales org under 200 reps. Coaching exists as a calendar invite. It does not exist as a system that changes what a rep does on the next call. After running coaching infrastructure for a dozen RevOps engagements, I have a strong opinion on why it fails and what to do instead.
What sales coaching is actually trying to do
Coaching is not training. Training is "here is how MEDDIC works." Coaching is "you ran a discovery call yesterday, you skipped the pain question, you lost the deal three weeks later, let us look at the recording together and figure out how you handle that next time."
Training is one-to-many and front-loaded. Coaching is one-to-one and continuous. They use different muscles and most sales leaders confuse them.
The job of a coaching system is to do three things, in this order. One, surface the specific behaviors that move pipeline. Two, get those behaviors in front of the rep with a real example from their own work. Three, follow up to see if the behavior actually changed on the next call. Nothing else matters. If your coaching does not do those three things, it is theater.
CSO Insights ran a study in 2022 that put dynamic coaching at a 27.6% win rate lift vs random coaching. Take the number with salt because the sample is sponsored, but the direction is right. The lift comes from coaching the specific moments that decide deals, not from the sheer hours of coaching time.
If your coaching does not change what the rep does on their next call, it is not coaching. It is a status meeting.
Every coaching session should produce one specific behavior change and a way to check it happened. If you cannot name both at the end of the hour, you wasted the hour.
Why most B2B sales coaching fails
I have audited coaching programs at 20+ B2B companies. The same five failure patterns keep showing up.
The manager listens to one call a week and calls that coaching. One call is not a sample. The rep ran 30 calls that week. The one your manager picked is probably the one the rep felt good about. Coaching from a curated sample misses the actual problem.
Feedback is vague and unverifiable. "Build more rapport." "Ask better discovery questions." "Be more consultative." These are not coaching points. They are wall posters. A rep cannot do anything different on Monday with "be more consultative" in their notebook.
There is no follow-through loop. The manager gives feedback on Friday. Nobody checks on Tuesday whether the feedback actually showed up on the next discovery call. The rep does not change because there is no accountability.
The manager is doing 100% of the work. One VP I worked with was reviewing 14 reps personally. He had 11 hours of one-on-ones a week and zero time to think about pipeline. Self-review and peer review beat manager-only review for the same time investment.
There is no data behind the call selection. The manager listens to whatever call was on the rep's calendar yesterday. A real coaching system pulls calls based on signals: deals stuck at stage three, talk-time ratios that look wrong, deals that lost after a strong discovery. The data picks the call, not the calendar.
If you read four of those five and recognized your org, you are not alone. You are normal. Normal is broken.
The coaching system that actually works
This is the framework I install when I do this for a client. It runs on top of HubSpot or Salesforce with conversation intelligence feeding the signal layer. Tools I have used in production: Gong, Chorus by ZoomInfo, Fathom, Clari Copilot, Avoma. Pick one and stick with it.
The signal layer
You cannot coach what you cannot see. The signal layer is conversation intelligence plus a few HubSpot or Salesforce reports that flag calls worth reviewing.
Concrete signals I configure:
- Discovery calls where the rep spoke more than 65% of the time (text-to-speech ratio is the cleanest leading indicator I have found of a weak discovery)
- Demo calls where no follow-up meeting was booked
- Deals where the contact stopped replying after a specific call (something on that call killed it, find out what)
- Pricing calls where the rep gave the number before getting commitment on the criteria
- Multithreaded calls where the rep only spoke to one person despite three being on the line
These are not generic. They map to specific behaviors. A 65% talk ratio tells you the rep did not ask enough questions. That is a coachable moment. "Build rapport" is not.
The self-review step
This is the part most managers skip and it is where the biggest gains hide. The rep watches their own flagged moments before the coaching session. They write down what they would do differently.
Two things happen when you do this. First, the manager is no longer the only person doing review work. Fourteen reps each doing 30 minutes of self review is seven hours of coaching infrastructure that did not exist before. Second, the rep arrives at the session already half-coached. The manager is not delivering bad news. They are confirming what the rep already noticed.
I require self review notes before the coaching session. No notes, the session is cancelled. This sounds harsh. It works.
The coaching call itself
Thirty minutes. One moment. One behavior change. Written down.
Not "let us go through your pipeline." That is a pipeline review and is a separate meeting. Not "let us listen to three calls." Pick one. Not "let us talk through your quota." That is a one-on-one and is also separate.
The rep brings the moment they flagged. The manager has watched it before the call. The manager asks: "What were you trying to do there? What did the prospect actually need? What is the one thing you would do differently next time?" The behavior change goes into a coaching log that lives in the CRM on the rep's record. Not in a shared doc. Not in the manager's notebook. In the system.
The verification loop
Next week, the manager pulls a call where the same situation came up. Did the behavior show up? If yes, write it down, the loop closed. If no, why not. Maybe the behavior change was wrong. Maybe the rep forgot. Maybe the situation was different. Find out.
This is where 90% of coaching programs die. They never close the loop. The behavior change is suggested and then orphaned. With the loop in place, you build a coaching log per rep that lets you see, over a quarter, which behaviors stuck and which did not. That is the artifact you take into a performance review or a QBR.
The tools that make this work
You can run this on a notebook and a shared doc. I would not. Manual coaching infrastructure decays in six weeks. Here is the stack I install.
Pricing reality. Gong is the gold standard and runs roughly $1,600 per user per year. Chorus by ZoomInfo bundles with their data product and runs $1,200 to $1,500 per seat. Fathom is free for unlimited recording with paid tiers around $24 per user per month for the coaching features. Avoma sits in the middle at around $900 per user per year. Clari Copilot is interesting if you already use Clari for forecasting.
For a 14-rep team, Gong at $22,000 a year is real money. Fathom or Avoma at $4,000 to $13,000 is more proportional. I have seen both work. The recording tool matters less than the coaching system you wrap around it. A bad coaching program on Gong is still a bad coaching program.
Manager-led vs peer coaching vs AI coaching
The default assumption is that the manager does all the coaching. That is wrong above eight reps and it is wrong if your manager has any other job.
Peer coaching works when it has structure. Random pair-ups do not. What works: weekly call jam sessions of three to four reps, each brings one moment they want feedback on, 45 minutes, manager attends but does not run it. The reps coach each other. The manager listens for what to add to the next one-on-one.
AI coaching is real now in a way it was not 18 months ago. Tools like Second Nature, Hyperbound, and Quantified do simulated role-plays where the rep practices a specific objection or discovery flow against a generated buyer. The rep gets immediate feedback on questions asked, talk ratio, objection handling. I use these for new-hire ramp and for specific scenarios like a competitive displacement pitch. They do not replace coaching. They replace bad role plays.
The metrics that show coaching is working
If your coaching system is real, you can measure it. If you cannot measure it, it is not real.
The metric is not "hours coached." Hours coached is an input. Coach an underperformer for 20 hours and if their win rate does not move, you wasted 20 hours.
The metrics I track per rep per quarter:
- Coaching moments logged (count of specific behaviors discussed)
- Behavior change attempts (where the rep tried the new behavior on a subsequent call)
- Stuck behavior change attempts (where the new behavior held)
- Win rate change quarter over quarter
- Stage conversion rate change at the specific stage that was the coaching focus
The last one is the punchline. If you coached a rep on discovery and their stage-one to stage-two conversion did not move, the coaching did not land. Try again with a different angle.
For a manager, the dashboard shows the same metrics rolled up across the team plus one extra: coaching coverage. What percentage of reps had at least one closed coaching loop this quarter. Below 80%, the system is breaking. Above 90%, it is healthy.
One focused coaching session per week, with self-review beforehand and verification after, beats four hours of unstructured manager time with the same rep. Every audit I have done confirms this.
How to roll this out without a revolt
If you walk into Monday and tell your sales team they now have to do self review and log behavior changes in HubSpot, half of them will silently revolt and the other half will half-comply for a month and then drift back. I have made this mistake twice.
What works:
Start with one rep. Pick the rep who is most curious about getting better. Run the full loop with just them for four weeks. Tweak based on what breaks.
Bring the next three in as a pilot. These reps now have a peer who already does it. Show the coaching log and the moment-to-behavior-change pipeline. Roll out the workflow automation that creates coaching tasks.
Bring the rest in once the pilot has a number to show. Ideally a stage-conversion improvement or a win rate change. Without a number you have an opinion. With a number you have a case.
Make the manager do self review on themselves first. If the VP Sales is unwilling to listen to a recording of their own demo and find a moment, they will lose moral authority to demand it of the team. I make every leader I work with run the loop on themselves for the first two weeks.
Tie it to the comp plan eventually, not initially. Comp pressure on coaching activity creates fake behavior. Get the system to work because it works, then make the metrics visible in the QBR, then make ramp and quota expectations dependent on the closed-loop count. By the time you tie it to comp, reps understand why.
What this connects to in the rest of the GTM stack
Coaching is one piece of the RevOps tech stack. It depends on conversation intelligence, on a CRM with a coaching object, and on a manager who has time to do the work. It feeds into ramp time, forecasting accuracy, pipeline coverage, and rep retention.
The under-appreciated link is ramp time. A new AE who joins a team with a real coaching system ramps in four months instead of seven. The system is documenting what good looks like for every stage. New reps watch the coaching logs, listen to the calls that got flagged, and see the behavior changes that stuck. That is your onboarding curriculum and it builds itself if you let it. The sales playbook reps actually use write-up goes deeper on that.
Coaching also closes the loop with sales enablement. Enablement builds the asset. Coaching makes sure the rep uses it correctly. Without coaching, the battle card sits in Notion and nobody reads it. With coaching, the manager sees the rep skip the battle card in a competitive deal and that becomes the moment of the week.
Coaching program stuck in status-meeting mode?
Book a free 30-minute audit. We will look at how you select calls, how you log feedback, and the one workflow change that closes the loop. No pitch, no slide deck.
Book an audit →FAQ
How many reps can one manager coach effectively?
With the system above, 12 to 14. Without it, six. The lift comes from self review and peer pods doing half the work. A manager who personally reviews every call for every rep tops out around six and is overworked at that.
Do we need Gong or can we start with Fathom?
For teams under 20 reps, Fathom or Avoma is enough. The Gong topic detection and deal warnings are useful at scale but the coaching loop itself works on any tool that records and transcribes. Pick the cheaper one, get the system working, then upgrade if you outgrow it.
How long does it take to see a win rate change?
Six months minimum on a B2B sales cycle of 60 to 120 days. The first quarter is for the system to start producing coaching moments and behavior changes. The second is for those behavior changes to land in actual deals. Anyone promising a win rate lift in 60 days is selling you a tool, not a system.
Should the manager listen to every call?
No. The signal layer picks the calls worth reviewing. Listening to every call is theater and it is how you end up with a manager doing 11 hours of review a week. Pull two to three calls per rep per week based on triggers. Skim the rest with AI summaries.
What if a rep refuses to do self review?
Two options. One, the manager runs the loop without the self-review step for that rep for two weeks. The coaching is less effective but it still happens. Two, you have a different conversation with the rep about what they signed up for. If a rep is unwilling to look at their own work, no coaching system will fix that. Performance management is a separate problem.