Tell me about explaining a model's limits to a non-technical stakeholder

Use the STAR structure and translate model behavior into business terms.

Situation. A product lead wanted to launch a churn-prediction model and assumed it was about 90 percent accurate for every customer, because that was the headline number.

Task. Set realistic expectations without killing the project.

Action. I replaced accuracy with a question the lead cared about: if we contact the top 1,000 customers by score, how many real churners will we reach? I showed examples of correct and wrong predictions, explained that the model was weaker on new customers because it had little data on them, and described what the numbers meant in cost per wrong decision. I proposed a pilot with a human review step and a clear success metric.

Result. The lead chose a limited launch. The pilot reached the agreed target, and we expanded with confidence and no surprises.

The lesson: avoid jargon, use concrete examples, be honest about uncertainty, and offer a safe next step.