Tell me about a time you had to make a decision without enough data

What They’re Really Asking

Can you balance risk and action when perfect information isn't available?

Strong Sample Answer

At Amazon, we were launching a new fulfillment feature but lacked historical data on peak holiday load for the specific warehouse configuration. The deadline was rigid. I gathered what proxy data existed from similar facilities and convened a working session with engineering and ops leads to identify our 'one-way door' risks. We agreed that a full rollout was too risky, but a controlled beta with 5% of traffic allowed us to gather real-time data while limiting exposure. I defined clear stop-loss metrics beforehand. Within 48 hours, the beta revealed a latency bottleneck we hadn't anticipated. Because we had limited the scope, we paused, fixed the issue, and relaunched two weeks later with zero impact on customer promises. This approach demonstrated 'Bias for Action' while respecting 'Customer Obsession' by preventing a broader failure. It taught me that lacking data isn't a blocker if you structure small, reversible experiments to generate that data quickly.

Common Mistake to Avoid

Company-Specific Variants