Diagnose a 10-Point Fall in Twitter's Daily Sign-In Rate Using Only Data Questions
Company: monday.com
Role: Product Manager
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Other
Twitter's daily active rate has dropped from 70% to 60%. For this case, a user counts as daily active on any day they signed in to the product. Find out why the rate fell.
The interviewer acts as a data analyst. You may ask only for information that can be pulled from the company's own data, such as counts, time series, event logs and breakdowns by segment. A question whose answer would not be in the data (for example, what a competitor launched or what was in the news) will not be answered.
```hint Two numbers make a rate
Write the rate as a fraction and find out which part of it moved before you start cutting the data by segment.
```
### Clarifying Questions
- What is the base of the percentage: users active at least once in the past 30 days, all registered accounts, or some other population?
- Did the rate move from 70% to 60% in a single day or over several weeks, and is each figure one day's value or an average over a period?
- Does "signed in" mean an explicit login event, or any day on which a user with a saved session used the app or website?
- Has the query, the event that records a sign-in, or the way accounts are filtered (for example bot removal) changed during this period?
### What a Strong Answer Covers
- Splitting the rate into numerator and denominator and establishing which one moved, and by how much.
- Checking the metric itself (definition, logging, authentication and session behavior) before reaching for product explanations.
- An ordered set of data-only questions (timing, platform and app version, geography, account age and type, engagement), with what each answer would rule in or out.
- Respecting the scenario's rule: outside events are inferred from their footprint in the data, never asked about directly.
- A conclusion that names the most likely cause, how much of the 10-point drop it explains, and the next step.
### Follow-up Questions
- The number of daily signed-in users is flat, but the base grew by about 17% over the same weeks. What do you examine next?
- The drop is confined to Android users on the newest app version, yet login failure rates are unchanged. What could be going on?
- How would you change the metric or its monitoring so that a shift like this is caught within a day?
Overview: A metric-diagnosis case for product managers: Twitter's daily active rate, counting anyone who signed in that day, fell from 70% to 60%. The candidate may only ask questions a data analyst can answer from the data, which tests splitting a ratio into its parts, validating the metric, segmenting the drop and inferring causes from data patterns.