A simulation objective describes an observable behavior the learner performs inside the scenario, something a fly on the wall could check off. Objectives are the spine of the design: they drive the scenario, the character's behavior, the evaluation criteria, and the end conditions.
Start from the business goal, work backward
- What outcome does the organization need? (fewer escalations, higher close rate, safer handoffs)
- What must people DO differently to get there? Those actions are your candidate objectives.
- What makes those actions hard in real life? That friction becomes the character's resistance and the scenario's complications. Content the learner merely needs to reference is knowledge-base material, not an objective.
Make each objective observable
- Use action verbs describing conversation behavior: asks, acknowledges, proposes, summarizes, de-escalates, quantifies, confirms.
- Avoid verbs you cannot see in a transcript: understands, knows, learns, appreciates, is aware of. If a subject-matter expert hands you one of these, ask "what would I see them do if they understood it?" and write that.
- Weak: "Understand the importance of empathy in customer calls."
- Strong: "Acknowledges the customer's frustration in their own words before offering a solution."
How many, and at what grain
- 1-4 objectives per simulation. More than that means the scenario is testing too much at once. Split it into a series.
- Each objective should be demonstrable multiple times, or at least clearly once, within the length of the simulation (its max turns, if you enable that setting).
- If two objectives always co-occur ("asks discovery questions" and "doesn't pitch early"), consider merging them into the stronger single behavior.
Objectives → evaluation criteria → end conditions
- Every evaluation criterion should trace to an objective, and every objective should have at least one criterion. Orphan criteria measure noise; orphan objectives go unmeasured.
- End conditions should let the learner finish by achieving (or clearly failing) the objectives, not only by running into the max turns limit.
- When reviewing an existing simulation, mismatches here are the highest-value fixes to surface.