The question

Share a recent experience where you failed to achieve success on a key metric and why?

The approach

I used to work in a food delivery startup, so let’s quickly use the STAR method to structure our thoughts around a story:

Situation: We started showing estimated time to delivery. Conversion went down sharply.

Task: Understood what the drivers were. Did field research, interviewed customers, did a few A/B tests to understand the underlying reasons

Action: Identified the largest contributing factor (unexpectedly, it was the specificity of the prediction). Decided to show a range instead, even though it still hurt conversion.

Result: We finally understood the reason — the average time shown was much higher than most people subconsciously assumed it actually would be. We still decided to show it, since research made it clear there was value for the customer. But we showed it in a way that minimized impact on conversion.

The answer

I worked in a food delivery startup. Based on user research, we understood customers wanted to know how soon their food would arrive. We started showing them estimated delivery timings for each restaurant. In our A/B test, we conclusively saw a drop in conversion.

This caught us off guard, but we dug deeper into the behavioral metrics to try understanding the underlying drivers. We also did a bunch of field research and interviewed a lot of customers to isolate the root.

We eventually found that customers usually assumed a delivery timing of less than half an hour, whereas our specific prediction was nearly always in the 30–40 minute range. This jarred with their expectations and led to a drop in conversion. However, given that our prediction was only an average, we often had deliveries which were well short of 30 minutes, which were now not even being considered by customers. We started showing a range instead, which covered 70% of all delivery timings (this was usually in the 20–50 minute range).

Tl;dr: We showed what we thought would be useful information for a customer, since we thought it would drive more transactions. The information was useful, but in a way that convinced users not to use our product. We still ended up showing the information, in a way that was still useful but didn’t hurt the usage of our product as much.

Prepfully’s got a truckload of more Facebook PM interview questions — and for a real-world mock interview with a Facebook PM, you can schedule one for yourself here.