Investigate an 8% YouTube DAU Drop While New Active Users Rise
Company: Google
Role: Data Scientist
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Technical Screen
YouTube's daily active users (DAU) are 8% lower than on the same weekday last week, while the number of new active users has increased. How would you investigate?
```hint Split the total
Write DAU as a sum of user groups and ask what a rise in new users implies for everyone else.
```
```hint Check the counting before the cause
Before searching for a product reason, ask whether the way active users or new users are counted could have changed.
```
### Clarifying Questions
- How is an active user defined (any app open, a video view, signed-in activity), and how is a new user identified (a new account, a new device or a new cookie)?
- Did the drop appear suddenly on this one day, or has DAU been falling for several days?
- Is the drop spread across platforms, regions and app versions, or concentrated in one of them?
- Were there launches, experiments, app releases, outages or logging changes in the past week?
- Was anything unusual about last week's day, such as a holiday or a major live event?
### What a Strong Answer Covers
- The decomposition of DAU into new and existing users, and what an 8% total drop with rising new users implies for existing users
- Data-quality checks first: data completeness, logging and definition changes, bot filtering, and identity changes that can relabel returning users as new
- A check of the baseline against several prior weeks and the calendar
- Segmentation by platform, app version, region, sign-in status, traffic source and user tenure, with each segment's contribution to the drop
- Hypotheses that explain both movements together, a way to confirm each, and a clear next action
### Follow-up Questions
- You find that the rise in new users comes almost entirely from one app version on one platform. What does that suggest, and how would you confirm it?
- How would you tell a one-day anomaly from the start of a sustained decline?
- If the cause turns out to be a recommendation change that reduced returning users, how would you estimate its long-term cost?
Overview: A metric investigation case: YouTube daily active users are 8% below the same weekday last week even though new active users increased. Tests decomposing DAU into new and existing users, ruling out logging, identity and baseline issues, segmenting the drop by contribution, and forming hypotheses that can be confirmed.