Ranking Google Play Store ads: objective, install prediction signals, and auction
Company: Google
Role: Product Manager
Category: Machine Learning
Difficulty: medium
Interview Round: Other
In a product manager interview at Google, you are asked to design the system that decides which ads people see in the Google Play Store and in what order. Explain what the ranking should optimize, which signals it would use, how ranking relates to what advertisers pay, and how you would judge whether the algorithm is working.
```hint Three parties, one score
Every ad slot serves a person looking for an app, a developer paying for installs, and Google's revenue. Decide how one ranking score can reflect all three before you list signals.
```
### Clarifying Questions
- Which placements are in scope: ads in search results for a typed query, or also ads on the home page and app detail pages where there is no query?
- What do advertisers pay for: taps, installs, or later in-app actions, and is the ranking tied to an auction?
- Is the priority short-term ad revenue, or long-term revenue protected by user trust and install quality?
- Are there policy constraints to respect, such as a cap on ad slots per page or stricter rules for apps aimed at children?
### What a Strong Answer Covers
- A ranking objective that balances user relevance, developer value and revenue, rather than ranking by bid alone
- Concrete signals for predicting whether a user will install and keep an app (query and app relevance, app quality, user context, past ad performance) and how new apps without history are handled
- How the score connects to the auction and pricing, plus eligibility and policy filters
- Offline and online evaluation: prediction quality, A/B tests, metrics for each party, and guardrails
- Risks such as low-quality or misleading apps buying their way up, feedback loops that favor incumbents, install fraud, and privacy limits on user data
### Follow-up Questions
- A game with a high bid gets many installs, but most users uninstall it within a day. How should the ranking respond?
- How would ranking change for an ad on the home page, where there is no search query?
- How would you decide whether adding one more ad slot to search results is worth it?
Overview: A technical product question for a Google product manager on designing the ranking algorithm for ads in the Google Play Store. It tests defining an objective that balances users, developers and revenue, choosing signals to predict valuable installs, linking ranking to the auction, and evaluating the algorithm with experiments and guardrails.