How do you coach reps to ask better questions using call data in 2026?

Two coaching models: the scorecard model versus the moment model

Every manager who sets out to coach questioning with call data ends up choosing between two philosophies, and most teams pick badly because nobody names the choice out loud. The first is the scorecard model: you define a small set of countable question metrics — total questions asked, open-to-closed ratio, whether a cost-of-inaction question appeared, whether a next step was confirmed — and you score every reviewed call the same way, every week, for every rep. The second is the moment model: you skip the counting entirely and hunt for two or three high-leverage moments per call where the rep had an opening and took the wrong turn, then coach those moments deeply.

The scorecard model's advantage is that it scales and it is objective. A rep can argue with "your discovery felt shallow." A rep cannot argue with "you asked four questions on that call, three of them yes/no." That objectivity is what makes call data worth using in the first place — it converts a subjective craft debate into a measurement conversation. It also travels: a scorecard can be applied by a frontline manager, an enablement lead, a peer reviewer, or increasingly by the conversation-intelligence platform itself, and the numbers mean roughly the same thing in every hand. When a RevOps team wants to report coaching coverage upward, the scorecard is the only version that produces a defensible number.

Its weakness is equally real. Counting rewards volume, and volume is easy to game. A rep told to hit ten questions per discovery call will hit ten questions, and three of them will be throwaway confirmations — "so you're the decision maker, right?" — that inflate the count while teaching nobody anything. The scorecard is also blind to sequencing. Ten strong questions asked in a random order are worth far less than six asked in a chain where each one builds on the last answer. And it cannot see whether the rep actually listened to the answer, which is the entire point.

The moment model inverts those trade-offs. It catches the thing the count misses: the buyer said "renewals are a headache" at minute seven and the rep answered with a feature. That single moment, replayed, is worth more coaching value than a full page of metrics, because it is specific, emotionally memorable, and directly re-runnable in a role-play. Reps remember moments. They do not remember ratios. The moment model also handles sequencing and listening natively, since you are watching the actual conversational turn.

Its weakness is that it does not scale and it does not measure. It depends heavily on the manager's own questioning skill — a manager who was never a strong discovery rep will pick the wrong moments. It produces no trend line, so you cannot tell whether the rep improved or whether you simply got better at finding flaws. And it is slow: finding two good moments means listening to or skimming a real call, which is fifteen to forty minutes of manager time per rep per week.

The practical answer for most teams is a hybrid with a clear division of labor: the scorecard diagnoses, the moment coaches. Let the platform count. Let the numbers tell you which rep to spend time on and which behavior is drifting. Then, once you are in the 1:1, drop the numbers after the first ninety seconds and spend the rest of the session on one moment. The scorecard earns you the right to have the conversation; the moment is the conversation. Teams that only do the first half produce reps who hit question quotas and still lose deals. Teams that only do the second half produce great individual sessions and no visible trend, which is why their coaching program gets defunded in the next budget cycle.

There is a third option worth naming because it keeps showing up in RevOps roadmaps: fully automated scoring with no manager in the loop, where the conversation-intelligence tool flags calls, generates a question-quality summary, and emails it to the rep. Treat this as an input, not a coaching model. Automated summaries are genuinely useful for triage and for giving the rep a self-service mirror between sessions. They are not a substitute for a human asking "what did you hear when you listened back?" Behavior change in questioning is mostly a confidence problem, and confidence does not respond to an automated email.

How to decide which model fits your team

The choice is not a matter of taste. It falls out of four variables you can actually measure about your own org: manager span of control, rep tenure mix, call volume per rep, and whether your discovery calls are structurally allowed to be discovery calls at all.

Start with span of control. A manager with five reps has roughly enough hours to run one deep moment-based session per rep per week — call it forty-five minutes of prep and conversation each, which is a manageable four hours weekly. A manager with twelve reps does not, and pretending otherwise produces the worst outcome: shallow sessions that feel like surveillance. At span above eight, the scorecard has to carry more weight, and you move to a rotation where each rep gets a deep moment session every second or third week and an automated scorecard summary in between.

Next, tenure mix. New reps in their first ninety days benefit enormously from the scorecard because they genuinely do not know what the bar looks like; the count gives them a target and the open-versus-closed split gives them a vocabulary. Reps past a year usually know the theory and are failing on nerve or on buyer knowledge, and for them the count is noise — you need the moment. If your team is mostly tenured and you are still running counting sessions, you are coaching a problem they do not have.

Then, call volume. A rep running twelve discovery calls a week produces a real sample; their numbers mean something and a bad week is visibly a bad week. A rep running three enterprise discovery calls a month produces a sample so small that any ratio is statistical noise, and coaching them off a count is close to superstition. Low-volume, high-value motions are moment-model territory almost by default, and the compensating control is that you review a larger share of the calls — sometimes all of them.

Finally, and most importantly, is the call structurally allowed to be discovery? If your process puts a demo on call one, or your meetings are booked at twenty minutes, no questioning coaching of either kind will work. You are looking at a system problem wearing a skill problem's clothes. Fix the meeting structure first; the rep's question count is a symptom, not the disease.

One more decision input that teams skip: who owns the scorecard definition. If each manager writes their own, you get five incompatible standards and no cross-team trend. This is where RevOps earns its keep — the question scorecard should be defined once, centrally, versioned, and shipped to every manager the same way a stage definition or a forecast category is. Managers own the coaching conversation; RevOps owns the measurement. Blur that line and the numbers stop meaning anything within a quarter.

The numbers behind each model

Be honest about what the published research does and does not tell you, because inflated precision here destroys manager credibility fast. Conversation-intelligence vendors publish call-analysis research showing that stronger discovery calls tend to contain more questions than weaker ones, and that the strongest calls skew toward open-ended, problem-centered questions rather than closed confirmations. Gong's published call research is the most widely cited example, and it puts strong discovery calls in a band of roughly eleven to fourteen questions. Treat that band as a guardrail, not a quota — it is derived from a particular vendor's customer base, skewed toward mid-market software sales, and it describes correlation, not a mechanism you can install.

Here is how to think about the cost side, which nobody publishes and which you can compute yourself. A moment-based session costs a manager roughly twenty to thirty minutes of prep — skimming or scrubbing one call to find two usable moments — plus a thirty to forty-five minute 1:1. At a span of six reps that is five to seven hours a week of manager time, or somewhere near fifteen percent of the working week. That is a defensible spend if it moves conversion. It is an indefensible one if you never measure whether it did, which is exactly why the pure moment model tends to get quietly abandoned around month four.

A scorecard-only program costs almost nothing in manager time once the scoring is automated — call it five minutes per rep per week to read the summary and send a note. Its cost shows up elsewhere: platform licensing, the RevOps time to define and maintain the scorecard, and the very real risk of behavior gaming that produces better numbers and identical win rates. Budget the scorecard's maintenance honestly. A question scorecard that nobody revisits for a year will encode last year's buying process and will actively mislead reps.

For targets, resist the urge to set a single number for the whole team. Set a direction and a floor instead. A workable floor for a full-length discovery call is that the majority of questions asked should be open, that at least one question should attempt to quantify the cost of the status quo, and that at least one should surface who else must sign off. Those three are binary and observable on any transcript, they are hard to game with filler, and they map directly to the parts of a business case that stall deals later. Notice that all three are quality checks, not counts — which is the point.

On the measurement side, separate leading from lagging indicators and give each an honest time horizon. Leading indicators — questions per discovery call, open-to-closed ratio, presence of a cost-of-inaction question, next-step confirmation rate, and talk-to-listen ratio — should move within two to four weeks, because questioning is genuinely one of the most coachable behaviors on a sales team. If a rep's reviewed calls show no shift after a month of weekly sessions, your diagnosis was wrong; go back and re-route the symptom. Lagging indicators — discovery-to-demo conversion, stage-two-to-three progression, overall win rate, and average deal size — need a full quarter minimum, and in enterprise motions with six to nine month cycles they need two or three quarters before the signal separates from noise.

The trap in that lagging set is attribution. Discovery-to-demo conversion moves for a dozen reasons that have nothing to do with question quality: lead source mix shifted, a competitor changed pricing, a new SDR team started booking softer meetings. Before you claim your question coaching lifted conversion, check whether the rep's *unchanged* peers moved the same direction over the same window. If everyone moved, it was the market. RevOps should be the function that insists on that check, because managers have every incentive not to run it.

Talk-to-listen ratio deserves a specific caution. It is the easiest metric to pull and the easiest to over-index on. A rep can hit a beautiful ratio by asking one question and then going silent while a buyer rambles about something irrelevant. Ratio is a smoke detector, not a diagnosis — it tells you to go listen to the call, nothing more. Use it to triage which recordings to review, never as a coaching target in itself, and never in a compensation or performance-review context, because the moment a ratio touches comp, reps optimize the ratio rather than the conversation.

Implementation and sequencing

Sequence matters more than tooling here. Teams that buy the platform first and design the coaching motion second end up with a very expensive recording archive that nobody opens.

Weeks one and two — define and baseline. Write the question scorecard before you score anything. Keep it to five or six lines, each observable from a transcript without judgment calls: question count, open-to-closed split, cost-of-inaction question asked yes/no, multi-threading question asked yes/no, next step confirmed yes/no. Publish it to the whole team, including the reps, before it is ever used. A scorecard reps have not seen is surveillance; a scorecard reps helped shape is a bar. Then pull two recent discovery calls per rep and score them, and do not coach off this first batch at all — it is a baseline, and telling a rep their baseline is bad on day one poisons the program.

Weeks three and four — model the bar, then drill. Before you critique anybody, watch a strong call together as a team. Reps need a reference recording more than they need feedback; most weak questioning comes from never having heard what good sounds like from a peer in their own segment, with their own product, against their own objections. Then run structured drills. The highest-yield ones are cheap: the transcript audit, where the rep highlights their own questions in two colors and counts them before the 1:1; question-only role-play, where the rep may ask but not pitch for three minutes; the why-ladder, drilling five layers into one stated pain; and the silence drill, where the manager stays quiet for five seconds after a strong question so the rep practices not rescuing the pause with a weak closed clarifier.

Weeks five through twelve — the weekly loop. One real call per rep per week, scored against the scorecard, coached on exactly one moment. One improvement per session, never a list. The rep self-scores before the meeting starts. This is where the GROW frame does the heavy lifting: name the goal in the rep's own numbers, let the recording be the reality, make the rep generate three to five better questions in their own words rather than accepting yours, and close on a commitment with a number and a date attached — "at least one cost-of-inaction question on each of the next three discovery calls, checked Friday off the recordings."

Day sixty and beyond — shift ownership. By roughly two months in, the rep should be self-scoring before every session and arriving with their own diagnosis. That handoff is the actual success criterion of the program, not the metric movement. A coaching motion that requires the manager forever has not built a skill; it has built a dependency.

Two implementation details are worth calling out because they are where these programs usually break. First, recording consent and rep trust. Frame recordings as the rep's own film room from day one, never as a compliance archive. The fastest way to lose the team is to review a highlight reel of one rep's worst moments in a group setting. One call, one session, one improvement — and let the rep pick the call sometimes. Second, the manager's own skill ceiling. If your frontline managers were promoted for closing rather than for coaching, they will default to coaching the deal in front of them instead of the portable skill. Run the same drill on them: have managers record and review their own coaching sessions against a coaching scorecard. It feels recursive and it works.

Adjacent surfaces are worth harvesting once the core loop is stable. The same transcript data that scores discovery questions also powers renewal and expansion conversations, where the question set is different but the coaching mechanism is identical — customer success managers who never ask about changed business priorities miss expansion the same way an AE who never asks about cost of inaction misses a business case. Support and onboarding calls carry the same signal. And upstream, the questions your best reps ask are the highest-quality raw material your marketing and product teams will ever get about what buyers actually worry about, if someone bothers to route them there. That routing is a RevOps job, and it turns a coaching program into a company-wide listening system rather than a manager's private habit.

Related questions

Does this work for SDRs, or only for AEs on discovery calls?

It works for SDRs with a modified scorecard. Cold-call windows are too short for eleven questions, so score for one qualifying open question and one permission-to-continue question instead of counting volume. The moment model transfers unchanged.

Can AI-generated call summaries replace the manager's review?

They replace the counting, not the coaching. Automated summaries are excellent triage — they tell you which call to open. Behavior change in questioning is mostly a confidence problem, and confidence does not respond to an emailed score.

What if the rep asks great questions and still loses deals?

Then questioning is not the constraint. Check territory quality, pricing fit, competitive displacement, and whether the buying committee was ever fully mapped. Continuing to coach a skill the rep already has is how managers avoid harder diagnoses.

How do we keep reps from gaming the question count?

Score quality binaries alongside volume — cost-of-inaction question asked, multi-threading question asked, next step confirmed. Those are hard to fake with filler. And never let any question metric touch compensation or the performance review.

Should peers review each other's calls instead of the manager?

Peer review is a strong supplement and a weak replacement. It scales the exposure to good calls and builds a shared vocabulary, but peers rarely deliver uncomfortable feedback. Use peer review for modeling the bar, manager review for diagnosis.

FAQ

How many questions should a rep actually ask on a discovery call?

There is no universal number. Published conversation-intelligence research points to strong discovery calls clustering around eleven to fourteen questions, weighted toward open and problem-focused. Use that as a guardrail rather than a quota — quality, sequencing, and whether the rep listened to the answer matter far more than raw count, and a rep chasing a number will pad it with throwaway confirmations.

Should I share the raw call data with the rep or just give my feedback?

Share the data first and let the rep self-diagnose. When a rep counts their own open versus closed questions on their own transcript, the gap becomes their conclusion rather than your criticism, and people defend their own conclusions instead of arguing with them. Your feedback lands better in the second half of the session, after they have already named the problem.

What if a rep gets defensive about being recorded?

Reframe recordings as their film room, not surveillance, and back that up with behavior: one call per session, one improvement, never a compilation of mistakes, and let them choose the call sometimes. Watching a strong peer call before you ever review theirs also helps, because it establishes that the recordings exist to show the bar, not to catch people.

Open or closed questions — which should I push?

Push open, problem-centered questions during discovery: walk me through what happens today, what does that cost you, who else feels this. Closed questions still earn their place confirming details and gaining commitment. A discovery call dominated by closed questions almost always produces weak pain, a weak business case, and a deal that stalls in legal or procurement months later.

How long before question coaching shows up in the numbers?

Leading indicators on reviewed calls should shift in two to four weeks, because questioning is highly coachable. Conversion and win rate need a full quarter at minimum, and longer in enterprise cycles. Before crediting your coaching for a conversion lift, check whether uncoached peers moved the same way over the same window.

Who should own the question scorecard — managers or RevOps?

RevOps should own the definition, versioning, and reporting; managers own the coaching conversation. If every manager writes their own scorecard you get incompatible standards and no cross-team trend, and the program becomes unmeasurable exactly when someone asks whether it is worth the manager hours.

Sources

• Gong Labs — Discovery call research and question data
• Gong Labs — Talk-to-listen ratio and win rates
• Harvard Business Review — The Surprising Power of Questions
• RAIN Group — Sales Discovery Questions
• Performance Consultants — The GROW Model of coaching
• Winning by Design — Sales frameworks and resources
• Harvard Business Review — What Great Listeners Actually Do
• Salesforce — Sales coaching resources

Related on PULSE

• How do you coach reps to ask better discovery questions?
• How do I train reps to ask better follow-up questions?
• How can I ask a question that helps a rep structure a better follow-up sequence?
• How do you coach reps to tell better customer stories?
• How do you coach a rep to write better follow-up emails?
• What questions should I ask before hiring a fractional CRO?