TL;DR. Role plays in training rarely fail because of the method itself. The problem is that a training room struggles to provide both safety and a realistic level of difficulty. An audience reliably reduces performance on the complex tasks that role plays are designed to practise (Bond & Titus, 1983). The usual solutions come at a cost: either fewer repetitions or less expert feedback. In a four-hour session with twelve people, each participant gets roughly one active turn. This article looks at the numbers, the trade-offs involved and what a practice format needs to provide so that safety and repetition no longer compete.
Key Takeaways
- The presence of others reliably improves performance on simple tasks and reliably degrades it on complex ones, across 241 studies with roughly 24,000 participants (Bond & Titus, 1983). A difficult conversation is the complex task.
- The effect is real but small: mere presence accounts for 0.3 to 3 percent of performance variance (Bond & Titus, 1983). What drives it is not the audience itself but the expectation of being evaluated (Cottrell, 1968).
- In a four-hour plenary block with twelve participants, each person gets about one active turn. Pair work multiplies that number and removes trained feedback in the same move.
- No format inside a training room provides both frequent repetition and expert feedback, because both depend on the limited attention of one experienced person.
- At Energieversum, participants used to reach the final onboarding day having practised between one and ten sales conversations; today the average is around 35. These figures come from one company's rollout and should be checked against your own setup.
The trainer announces a role play and the mood in the room changes. Someone suddenly becomes very interested in their handout. Someone else reaches for their phone. Then somebody asks whether the group could simply talk through the situation together instead. Most facilitators know this moment. It is usually interpreted as resistance: the group is not open enough, not brave enough or simply not ready yet.
That interpretation places the problem with the participants, but it misses the point. No one would expect a pilot, surgeon or musician to perform a difficult sequence correctly on the first attempt while a group of peers watches. The issue is not the attitude of the people in the room. It is the way the situation is set up. This article looks at how an audience affects performance, how many practice turns each person actually gets, what is lost with the usual remedies and what needs to change before practice can build real competence. The uncomfortable truth is that better facilitation alone will not solve the problem.
Why do people resist role plays in training?
The resistance is not aimed at the method but at the evaluation it creates. Someone who refuses a role play is refusing to be watched by colleagues while doing something they cannot yet do.
These reasons are well documented in the adult education literature: fear of becoming the centre of attention and having to prove oneself, concern about revealing more than intended, pressure to be creative on demand and fear of showing emotion and being judged for it. Then there are the objections people voice directly, such as role plays feeling childish or taking too much time.
So it is not true that nobody talks about the discomfort. The field has discussed it and proposed ways to address it. The more interesting question is what those measures do to the practice itself.
What does an audience do to performance in a role play?
An audience measurably degrades the exact class of task a role play exists to rehearse. The meta-analysis by Bond and Titus pooled 241 studies covering roughly 24,000 participants and confirmed a consistent pattern: the presence of others improves performance on simple, well-learned tasks and degrades it on complex ones (Bond & Titus, 1983).
A pricing objection, a declined promotion or an escalating service call is a complex task. Each one requires people to listen, weigh up what they have heard and formulate a response in the moment. That is precisely the kind of task that suffers when someone is being observed.
Two qualifications are important here. First, the effect is small. Mere presence accounts for between 0.3 and 3 percent of the variance in performance, so it would be an exaggeration to say that an audience ruins performance. Second, presence alone is not the decisive factor. Cottrell showed that what matters is the performer's concern about being evaluated by the people present (Cottrell, 1968). A room full of people who are not there to judge feels different from a room full of people who will give feedback afterwards. A training room is usually the latter.
How many practice turns does one person actually get?
In a plenary role play, each person gets roughly one active turn in half a day. This figure is rarely calculated, yet it explains much of why role plays often fall short, regardless of facilitation technique.
Consider a typical training block with twelve people and four hours, or 240 minutes. A single round, including a short introduction, the conversation and the debrief, realistically takes 20 minutes. That leaves time for twelve rounds. With one person in the active role each time, everyone gets one turn. Even that assumes that nothing overruns and there are no breaks or longer discussions.
The obvious alternative is pair work. Everyone practises at the same time and no group is watching. With six conversations running in parallel, the number of turns increases considerably. But there is no trained observer in most of those conversations. The facilitator can follow only one of the six pairs, while everyone else receives feedback from someone who is still learning the same skill.
| Format | Active turns per person | Trained feedback | Exposure to the group |
|---|---|---|---|
| Plenary, one round at a time | about 1 | yes | high |
| Pair work, all pairs at once | about 6 | no, peer feedback only | low |
| Fishbowl with a rotating pair | 1 to 2 | yes | high |
The problem is easy to see: no format in the training room provides both frequent repetition and trained feedback. Reduce the exposure and you lose the expert evaluation. Keep the expert evaluation and there is little time left for repetition. Both depend on the same limited resource: the attention of one experienced person.
Why does every safety measure weaken the practice?
Every common remedy for resistance reduces exposure by making the exercise easier. That comes at a cost. The recommendations make sense, but their effect on the quality of the practice is rarely considered.
Four measures appear in almost every practitioner guide: pair work instead of performing in front of an audience, which reduces the quality of the feedback; a gentle start with simplified warm-up exercises, which lowers the difficulty; permission to stop at any time, including at the point where the real learning might begin; and an error-friendly atmosphere with appreciative feedback, which is helpful but can make the response less clear when appreciation becomes the only rule.
Remove the pressure entirely and the exercise no longer resembles the situation it is meant to prepare people for. A conversation in which the other person gives in politely does not practise what a real conversation demands. This trade-off is not caused by poor facilitation. It is built into the format: in a training room, increasing safety often means reducing difficulty.
What would let someone practise without an audience and without losing the feedback?
What is needed is a counterpart that remains challenging and an evaluation that does not depend on someone watching. The trade-off disappears only when both elements no longer rely on the limited resources in the room.
Voice-based practice conversations with a virtual counterpart are designed to do exactly that. At Sleak, this is called Training Mode. A person chooses a scenario, has a complete conversation with a virtual persona and then receives an evaluation from the AI Coach based on a Scorecard. The Scorecard defines the expected standard in advance, while the feedback refers to evidence from the transcript instead of offering general praise. In this context, the AI Coach is the software that applies the standard, not a person. No colleague is watching, so there is no concern about being evaluated by others in the room. At the same time, the persona does not simply give in, so the task remains challenging. The product page explains how these voice-based practice conversations work.
More than 77,000 conversations have been run on the platform to date, across over 600 organizations. Those counts include test and demo accounts, so they describe platform usage rather than a customer number.
This does not replace the training room. It changes what the room is used for. Repetition can happen elsewhere, leaving the training room for the work it genuinely does better.
In practice: what the numbers at Energieversum show
A real rollout shows the cost of practising in the training room, both in time and in the number of turns. Before the change, the sales onboarding programme at solar provider Energieversum relied on one-to-one role plays with a team lead.
Each session could take up to 90 minutes. During that time, a team lead had to play the customer, observe the representative and provide detailed feedback afterwards. Running enough of these sessions manually required so much capacity that a team could realistically onboard only one new representative per month to the required standard. The limiting factor was not willingness, but available time.
After the change, practice volume increased. Before Sleak, many participants reached the final onboarding day having practised between one and ten sales conversations. Today, the average is around 35. Marc Weißbrenner, Head of Field Sales, describes the previous situation clearly: "The one-to-one roleplays used to consume the time and expertise of our best team leads. Now we no longer need them to run every practice conversation." These figures come from one company's rollout under its own specific conditions and should be compared with your own setup.
Isn't this just role play with a chatbot?
The important difference is not the medium, but who is present and what the performance is measured against. The objection is fair. At first, talking to software may seem like a reduced version of talking to a person, and in some respects it is.
What changes is the link between repetition and the availability of an experienced observer. The number of turns no longer depends on that person's time, and the evaluation no longer depends on who happens to be watching. Once those two things are separated, the original trade-off disappears. The conversation remains difficult because the counterpart does not give in out of politeness, while the exposure is removed because nobody from the team is listening.
The training room still has clear strengths. It is the right place for genuine human unpredictability, working together to find the right wording, exploring the group dynamic that emerges when twelve people view the same difficult conversation from different angles and building trust in a new team. These are not repetition tasks, and they belong in the room. A twelfth attempt at getting a standard phrase right does not.
FAQ
Why do role plays in training so often fail to work?
Because they are expected to provide two things that depend on the same limited resource: plenty of repetition for each person and trained feedback on every attempt. The presence of others can also reduce performance on complex tasks (Bond & Titus, 1983). The method itself is rarely the main problem. The available time usually is.
Does running role plays in pairs instead of in front of the group help?
For the number of turns, yes. For the quality of the feedback, no. With six pairs running at once, a facilitator can be inside exactly one conversation. The other five receive peer feedback from people who are learning the same skill at the same time.
Isn't participant discomfort just a question of trust within the team?
Trust lowers the threshold, but it does not solve the capacity problem. Even in a high-trust team, one person still gets roughly one active turn per afternoon. And the mechanism behind the discomfort is not the audience as such but the expectation of being evaluated (Cottrell, 1968).
How many practice conversations are enough?
There is no universal number, as it depends on the scenario and each person's starting level. Real rollouts can still provide a useful sense of scale. At Energieversum, practice volume increased from between one and ten conversations to around 35 per person by the final onboarding day. The absolute figure matters less than moving well beyond a single attempt.
Related reading
- Practicing Difficult Employee Conversations: 8 Scenarios Where AI Makes Leaders More Confident
- AI in Sales Training: How AI Role-Plays Empower Sales Reps Directly
- Scaling Sales Training: Why Traditional Approaches Fail and How AI Fixes It
- The Coaching Gap: Why Managers Are Accountable for Outcomes They Have No Tools to Produce
