Start from the demand unit that actually drives the market, then work backward to a measurable proxy.

Population is only one possible base. Depending on the market, useful alternatives include businesses, transactions, locations, installed devices, vehicles, employees, or usage occasions.

Choose a proxy with a clear relationship to demand and state the conversion assumptions explicitly so the interviewer can challenge them.

If several proxies are possible, pick the one that creates the shortest defensible chain. A simple model with transparent assumptions is stronger than a complicated model built around weak data. For a related reference, see market sizing.

I had the opposite problem and moved too fast. Slowing down for five seconds before answering actually made me sound more confident.