How to Practice Discovery Calls Before You’re on One

A discovery call is the first moment a rep has to run a conversation instead of recite one. Most reps get their first live discovery call as their training, which means a real prospect absorbs the learning curve. Discovery call practice moves that curve somewhere cheaper, where a bad question costs nothing.

The Actual Anatomy of a Discovery Call

New reps are usually handed a question list and told to stay curious. A question list is a long way from a call structure, and the distance between the two is where most first calls come apart.

A sales discovery call has five moving parts, and the order they happen in matters:

• The frame. The opening where the rep says what the call is for, how long it will take, and what happens at the end. Skipping the frame is why so many calls drift into a demo nobody asked for.
• Context. What the prospect’s world looks like today. Team size, current tooling, who owns the process, what changed in the last two quarters.
• Diagnosis. What is going wrong inside that context, how long it has been going wrong, and what they have already tried to fix it.
• Impact. What the problem costs, stated in the currency the prospect’s own leadership uses.
• Access and next step. Who else has to be in the room, and what specifically happens after the call ends.

Reps who struggle rarely struggle with the whole sequence. They collapse the middle. They collect context, skip diagnosis, and then assert impact instead of asking for it. The customer discovery work that belongs in the middle of the call gets replaced by a pitch wearing the costume of a question.

That collapse is predictable enough to build practice around, which is the useful part. The skill you are training is narrow: a rep who can survive the middle of the call.

Qualifying Questions and Diagnostic Questions Are Different Tools

Qualifying questions serve the seller. Does this account have budget, a timeline, a decision process, a shape that matches your ideal customer. Frameworks exist to make that systematic, and structures like BANT and MEDDPICC are worth teaching explicitly rather than leaving them to folklore. Lead qualification is a discipline on its own, and reps who are sloppy at it fill a pipeline with deals that were never real.

Diagnostic questions serve the buyer. They surface something the buyer had not put into words yet, or had only put into words as a complaint. “How many people touch a deal before it gets quoted?” is qualifying. “Walk me through the last deal that got stuck at quoting. What actually happened?” is diagnostic.

The difference shows up in what the prospect does next. A qualifying question produces an answer. A diagnostic question produces a story, and the story contains the exact language your champion will use internally when you are not in the room. Reps who only ask qualifying questions end up writing the business case themselves, in their own words, which is why so many of those business cases die in a meeting the rep never attends.

Under pressure, reps default to qualifying. Qualifying questions have right answers and feel like forward motion. Diagnosis takes longer and sometimes surfaces information that kills the deal. A rep who has never practiced sitting inside diagnosis past the point of comfort will leave it early on every call they run, and a qualification scorecard filled in after the fact will not tell you they did.

The Moment a Rep Talks Past a Buying Signal

This is the most expensive failure in discovery and the hardest to spot on a call recording unless you already know what you are listening for.

A prospect says: “Honestly, the reason I took this call is that we lost two of our better people last year, and I think how long it takes to get someone productive is part of it.” A door just opened. What the rep says in the next six seconds decides the call.

The weak version sounds supportive. “That makes sense, we hear that a lot. Let me show you how our onboarding piece works.” The rep acknowledged the signal and then walked past it to get back to the deck. The tell is structural and easy to score: an acknowledgment token, then a pivot to product. “Totally.” “Great question.” “Makes sense.” Followed by a feature.

The strong version stays in it. “What made you connect those two things?” Then, after the answer, one more. “When you say productive, what does that look like when it goes right?” The rep is now in a conversation that the prospect is driving, and every sentence the prospect says is material for the proposal.

This is active listening as an operational skill. It is measurable in a call recording, it correlates with how much of the call the rep spent talking, and talk time ratio is a reasonable proxy to watch even though it is a blunt one. A rep can talk very little and still miss every signal. A rep who talks through two thirds of a discovery call has almost certainly missed several.

How to Build Discovery Call Practice That Transfers to a Live Call

Full call run-throughs are the wrong unit. A rep who runs one end to end gets a single attempt at the hard moment buried inside forty minutes of setup. Drill the moments instead, in isolation, many times.

The moments worth their own drill

• The prospect gives a vague answer. “It’s fine, I guess.” The rep has to get specific without sounding like an interrogation.
• The prospect names a competitor unprompted, and the rep has to stay curious instead of defending.
• The prospect asks for a deck to circulate internally, which is usually a soft exit.
• A buying signal arrives disguised as a complaint about something unrelated.
• The person on the call has no authority, and the rep has to ask for access without insulting them.
• The third-level follow up, where a rep asks a question about the answer to their own follow up.

Run each one repeatedly against a persona that behaves consistently, with feedback after every attempt. Repetition works here because of how memory forms. Roediger and Karpicke showed that actively retrieving information produces stronger retention than restudying it, which is why a rep who has produced the question out loud fifteen times will produce it on a live call, and a rep who read it on a slide will not.

Why AI roleplays fit this kind of drill

This is the practical case for AI roleplays in discovery training. The bottleneck on drilling a single moment fifteen times has always been finding someone to play the prospect fifteen times. With Yoodli, reps run the same moment as often as they need on their own schedule, and managers get scored sessions instead of a scheduling problem.

Design the persona with the same care you would give the scenario. Decide what the persona knows, what it will volunteer without being asked, and how much resistance it offers. A persona that answers every question generously teaches reps nothing, because the failure mode you are training against is a prospect who gives you almost nothing until you earn more.

How Often Should Reps Run Discovery Call Practice?

Before the first live call, every new rep should have run each of the drills above until the strong response is the one that comes out under pressure. For most sales onboarding cohorts that means daily practice during ramp, in short sessions, with the persona and the scoring rubric held constant so the rep can see their own trend.

After ramp, the cadence drops but should never hit zero. Two triggers are worth building into the program:

1. A new pattern shows up on real calls. A competitor starts appearing in deals, a new objection lands, or pricing changes. Build a drill around the new moment and push it to the whole team the same week.
2. A rep’s call data slips. If talk time creeps up or follow-up questions drop off in recordings, assign the matching drill rather than a generic refresher.

Tenured reps benefit from a lighter version of the same loop. One or two drills a month, chosen from what their manager is seeing on recordings, keeps the reflexes current without turning practice into a chore. The point of the cadence is to keep the gap between the practice persona and the real prospect small enough that nothing on a live call feels new.

What to Measure

Score the practice against the same rubric your managers use on real calls. A practice-only scorecard creates a number nobody trusts and nobody acts on.

The metrics your org already reports on are the right ones to watch. Ramp time and weeks to first closed deal tell you whether earlier discovery reps are shortening the runway. Meeting-to-opportunity conversion and stage progression tell you whether the calls themselves got better. Certification pass rate tells you whether the bar is set somewhere meaningful, and win rate is the lagging number everyone will ask about eventually.

At the call level, watch how often a rep asked a follow up to their own question, how often they moved to product within one turn of a stated problem, and whether the next step they set was specific enough to be calendared. Those three are observable, scoreable, and directly coachable.

Clari used Yoodli AI roleplays to improve GTM conversation quality by 36%, which is the kind of measurement that only exists when practice and real calls are scored against the same definition of quality.

Start with the three questions your strongest reps ask that your newest reps skip. Build a drill around each one, with follow ups, and put it in front of the next onboarding cohort before they take a live call. To see how those drills fit into a full onboarding and certification program, the sales roleplay page walks through the structure.