Flagging Underperforming Meta Ads and Quantifying Their Impact
Company: Meta
Role: Product Manager
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Other
You are a product manager working on ads at Meta. Which metrics would you use to spot ads that are performing badly, and how would you measure the effect those ads have?
Explain what you mean by poor performance, which metrics flag it, and how you would quantify the impact of those ads.
```hint Say who the ad is failing
An ad can disappoint the advertiser who paid for it, the person who sees it, or Meta's own marketplace. Decide which failure you are measuring before you pick metrics, because the same ad can look fine from one side and bad from another.
```
### Clarifying Questions
- Does "poor-performing" mean ads that miss the advertiser's goal (few clicks or conversions for the money spent), ads that people dislike, or both?
- Which placements are in scope: Facebook and Instagram feeds, Stories, Reels, or every ad surface?
- Is the goal a monitoring system that flags individual ads, a measurement of the total effect poor ads have, or both?
- Should the metrics work for every advertiser objective (awareness, traffic, app installs, purchases), or one in particular?
### What a Strong Answer Covers
- A definition of poor performance from the advertiser, user and platform perspectives, and how these can disagree for the same ad.
- Objective-aware metrics for flagging ads, compared against similar ads rather than one global benchmark, with handling for low-volume noise.
- Negative user-feedback and attention signals that capture harm click metrics miss.
- A causal method, such as long-running holdouts, for measuring the effect of poor ads on user engagement, advertiser retention and revenue.
- What the metrics drive: ranking changes, advertiser guidance or policy action.
### Follow-up Questions
- An ad has an above-average click-through rate and an above-average rate of people hiding it. Is it a good ad?
- How would you tell whether a fall in feed time was caused by worse ads rather than worse organic content?
- Ads from new advertisers look poor on most of your metrics. How do you avoid penalizing them unfairly?
Overview: A metrics prompt for an ads product manager at Meta about spotting ads that perform badly and measuring their effect. It tests defining poor performance for advertisers, users and the platform, choosing objective-aware and fairly normalized metrics, and using holdout experiments to estimate the impact.