How to Run a Live Discovery Call Inside the Interview Loop (With a Script)

Ask a sales candidate to describe a great discovery call and watch what happens. They reach for their best story — usually an outlier deal, retold with the polish of someone who's used it in three other interviews that same week. What you're actually scoring, whether you mean to or not, is how well they can construct a narrative under zero time pressure with unlimited prep. That's a real skill. It correlates with confidence and interview experience. It does not reliably correlate with whether the person can sit in front of a prospect who's mildly irritated about being on the call and extract something useful in twenty minutes.

The fix is not a cleverer version of the same question. It's replacing the question with the thing itself: a live, scored discovery call role-play, run inside the loop, against a buyer persona built to resist rather than cooperate. This takes about 30 minutes to run and tells you more than every other stage of a typical sales interview combined.

Why "tell me about a time" fails specifically for discovery

Discovery is a live-listening skill. It's the ability to hear an answer, notice what it implies, and ask the next question that follows from what you just heard rather than the next question on your prepared list. None of that is testable through a retold story, because the candidate already knows how the story ends. They're narrating a decision tree they've already walked. A live role-play forces them to walk one they haven't seen, which is the entire point.

Setting it up

You need three things before the interview: a buyer persona brief, a script for the interviewer playing the buyer, and a scoring rubric. Build all three before you're in the room — improvising the buyer's resistance on the fly produces inconsistent difficulty across candidates, which makes your scores incomparable to each other.

The buyer persona

Keep it realistic and mildly difficult, not hostile. A buyer who stonewalls everything tells you nothing except how someone handles stonewalling. You want a buyer who cooperates the way a real, moderately busy prospect cooperates: gives short answers unless asked a good question, volunteers nothing unprompted, and has one soft objection in reserve for the middle of the call.

Example brief for the interviewer to work from:

You are Sam, Head of Ops at a 90-person logistics company. You took this call because a colleague mentioned the category, not because you have an active problem you're chasing. You're moderately busy and give short, factual answers unless the candidate asks something that makes you think. You do not volunteer your budget, your timeline, or who else is involved unless asked directly and well. Partway through, if the candidate asks about current process, mention offhand that "we've got something that mostly works," and let them decide whether to dig into that or move past it.

The script

Run it in four beats, roughly 20 minutes total:

Open (2 min) — candidate introduces the call, sets an agenda, earns the right to ask questions. Score whether they ask permission or just launch into questions.
Discovery (10 min) — candidate runs the actual questioning. This is the bulk of the score.
The soft objection (3 min) — the buyer mentions "something that mostly works." Score whether the candidate treats this as a dead end or as the single most important sentence in the call.
Close (5 min) — candidate has to propose a next step based on what they actually heard, not a generic "let's schedule a demo."

The rubric

Score against named competencies, not overall impression. A workable starting rubric:

Competency	What "clears the bar" looks like	Common failure
Questioning depth	At least 3 questions surface a business consequence, not just a fact	Candidate collects facts (headcount, tools) but never asks "so what happens because of that"
Active listening	Follow-up questions clearly build on the previous answer	Candidate runs a pre-set list regardless of what the buyer just said
Objection handling	Digs into "something that mostly works" rather than gliding past it	Candidate hears it, nods, moves straight to next scripted question
Next-step relevance	Proposed next step ties directly to something surfaced in the call	Generic "let's get you in for a demo" regardless of what was discussed

Two interviewers watching the same 20 minutes should land within a point of each other on each row. If they don't, the rubric's still too vague — tighten the language until it's observable, not interpretive. The Discovery Call Scorecard for Managers is a solid base if you'd rather adapt one than build from a blank sheet, and the Sales Candidate Scorecard Calculator is useful once you're running this across a slate of candidates and need the scores to roll up into one comparable number per person.

What you'll actually see

Run this ten times and a pattern emerges fast. The candidates who interview beautifully on "tell me about a time" split roughly into two groups once you put them live: some hold up, and some visibly reach for the story-mode reflex — trying to narrate their way through a question they haven't actually thought about — and stall. That gap is the entire value of the exercise. It's invisible in a behavioural interview and impossible to miss in a live one.

The objection you'll get, and the answer

Hiring managers resist this exercise for one honest reason: it takes longer to set up than a question bank, and it feels riskier to run because a bad role-play can go visibly badly in the room. Both true. Set aside 45 minutes instead of 20 to interview each candidate, and brief whoever plays the buyer so their difficulty is consistent across candidates — that's most of the setup cost, and you only pay it once.

Where this fits in the loop

Run it after a screening call, not before. You want people who've already cleared a baseline threshold going into the exercise, and you want the live segment to be the deciding evidence, not the first filter — a bad phone screen already tells you enough without burning 45 minutes of a hiring manager's time on someone who wasn't close.

A candidate who talks a good discovery call and a candidate who runs one are not always the same person, and the only way to find out which one you've got is to stop asking them to describe it and put them in the room.