The capstone project has become standard in enterprise training. Almost every serious technical programme now includes one. Most of them are useless as evaluation tools — not because of poor design intent, but because of the social dynamics that shape them.
The pressure on L&D teams to show high completion and pass rates is real. Programmes where participants fail look like programme failures, not participant gaps. Vendors whose learners fail capstones lose clients. The result is capstones calibrated to ensure that everyone, with reasonable effort, can complete them. A capstone that cannot fail anyone tells you nothing about capability.
Safe capstones have identifiable signatures. The task is narrow enough to be solvable by following the course material step by step. There is a clear template or example that participants can model. The evaluation criteria reward completion and presentation quality over technical correctness. Reviewers are reluctant to fail participants who have made a genuine effort.
None of these properties are wrong in isolation. The problem is that combined, they produce an artefact that signals effort and familiarity rather than capability. Passing a safe capstone tells you the participant attended the programme and can work within heavily scaffolded constraints. It does not tell you whether they can apply the skill to a problem they have not seen before.
A capstone that actually evaluates capability has three properties that safe capstones typically lack:
Novel problem structure. The task cannot be solved by re-applying examples from the course. It requires the participant to adapt, combine, or extend what they have learned to a context they have not encountered. This is uncomfortable by design — the discomfort is the evaluation.
Realistic constraints. Enterprise problems come with incomplete information, time pressure, ambiguous requirements, and tooling that does not behave as expected. A capstone that provides clean data, complete specifications, and unlimited time is not evaluating enterprise capability — it is evaluating tutorial-following ability.
Calibrated failure. A well-designed capstone should produce failure rates that reflect the actual distribution of readiness in the participant population. If 100% of participants pass, the capstone is not evaluating at the right level. A 10 to 20 percent failure rate in a reasonably well-designed programme is not a sign that something is wrong — it is a sign that the capstone is doing its job.
Implementing evaluative capstones requires decisions that feel risky. Agreeing in advance that some participants will not pass requires L&D to separate programme quality from participant outcomes — and to communicate that distinction clearly to business stakeholders who may not have made it before.
It also requires trainers and evaluators who are willing to give honest assessments. This is a hiring and training decision: the people who evaluate capstones need to be practitioners who can tell the difference between a technically correct solution and one that happens to look right.
The reward for this investment is measurement that actually informs decisions. Knowing which participants can apply a skill independently, and which need more practice, is the foundation of every meaningful downstream decision — about deployment, about further development, about programme effectiveness.
Browse our upcoming batches — live, instructor-led, delivered on Orbit.