Most sales roleplays are theatre. Here's how to design one that's fair to candidates, consistent across interviewers, and genuinely predictive of who will close.
Almost every sales team says they run roleplays. Very few run them in a way that produces comparable, trustworthy signal. The common version — a candidate pitches whatever they want to whoever's available, followed by a vibe check — is closer to an audition than an assessment.
A predictive roleplay has four properties. It's standardised (same scenario for everyone), adversarial (the buyer pushes back realistically), scored against a rubric, and reviewable (you can watch it again). Miss any of the four and the signal degrades badly.
The scenario should mirror a real deal your team runs. Not a dream deal — a typical one, with a typical complication. If your AEs usually sell into mid-market ops teams against an incumbent, that's your scenario.
Write it down completely: the company, its size, the buyer's role, the business context, the competing priority, and the hidden complication the candidate has to uncover. Then use it identically for every candidate.
The fairness trap
If you improvise a different scenario per candidate, you can't compare them — and candidates who got an easier buyer were graded on luck. Freezing the scenario isn't rigidity, it's the minimum for a defensible decision.
The buyer is the most important variable. A roleplay where the "buyer" answers every question helpfully and says "that sounds great" tests nothing. You need a buyer with a position.
Build your buyer with five things:
A stated priority that isn't yours
"Right now we're focused on renewals, not new tooling." The candidate has to find the bridge.
A hidden real problem
The thing they'd never volunteer in the first ten minutes. Only good discovery surfaces it.
An incumbent they're defensive about
"We already have something." Watch whether the candidate attacks the competitor or explores the gap.
A budget constraint that's real
Not a bluff. A genuine limit that requires the seller to scope or sequence rather than discount reflexively.
One moment of genuine hesitation
The "let me think about it" moment. This is where closing capability actually reveals itself. More on that in this breakdown.
A 45-minute unstructured roleplay produces mush. Break it into defined segments so you know what you're evaluating at each point.
Prepare
Candidate reads the brief and plans. Give them the scenario, not the answer.
Discovery
First conversation. No product allowed. Only questions and qualification.
Demo / value
Tailored walkthrough. They must reference what they learned.
Objections + close
Pushback, pricing pressure, hesitation. Do they secure a next step?
Deal review
"How did that go? What did you miss? Would this close?"
That final deal review is worth its own attention. Self-assessment is heavily predictive — candidates who can accurately critique their own performance are usually the ones who improve fastest after coaching.
Rubrics fail when they score impressions. "Good energy" and "seemed knowledgeable" aren't behaviours. Replace them with things you could count.
Observable
Not observable
Two interviewers watching the same roleplay should reach roughly the same conclusion. If they routinely don't, the rubric is too vague or the buyers are behaving differently.
Practical fixes:
This is the step almost nobody does, and it's the one that turns a process into an instrument. After you've hired a few people through the assessment, compare their roleplay scores to their actual performance at 90 and 180 days.
If high scorers are performing and low scorers aren't, your rubric works — keep it. If there's no correlation, you're measuring the wrong categories and should rebuild. If low scorers are outperforming high scorers, you've likely over-indexed on polish over substance.
This is also where the economics get compelling: the cost of a mis-hire, which we broke down in The True Cost of a Bad AE Hire, is large enough that even a modest improvement in predictive accuracy pays for the process many times over.
The friendly buyer
The interviewer wants the candidate to succeed, so the objections soften. Scores inflate and the assessment stops discriminating.
The product-pitch roleplay
If the candidate can spend the whole session demoing, you're testing presentation skill, not selling skill.
No artifact
Debriefs held hours later reconstruct the session in the candidate's favour. Record it, or at least take structured notes live.
Ignoring the debrief
The candidate's own self-review is often the most revealing part of the whole exercise. Score it.
Everything above is achievable internally. It's also a project: scenario design, buyer training, rubric development, calibration sessions, and ongoing maintenance. Teams usually do it well once and then let it slip as hiring volume rises.
Showcazz provides the simulation as a running service — candidates go through a full deal cycle and emerge with a scored record you can compare against a benchmark. If you'd rather build your own, this guide is the blueprint. Either way, the underlying argument holds: see the behaviour before you make the offer, not after. We make that case fully in Know Who Can Sell.
Showcazz runs a full five-stage AE deal cycle for every candidate and returns a scored, comparable record — no scenario design required.