Explain a Threaded-Replies Launch That Lifted Comments but Cut Group Posts
Company: Meta
Role: Product Manager
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Other
A product team adds threaded replies to the comments on posts inside groups, so members can answer a specific comment directly instead of adding to one flat list. After the change, comments per user rise 10% while posts fall 2%. Why might this have happened, and how would you work out which explanation is right?
```hint Follow the effort
Members spend a limited amount of time and attention in a group. Think about what threaded replies make easier or more rewarding, what that might replace, and how the change alters what members see.
```
### Clarifying Questions
- Were these numbers measured in a randomized experiment against a control group or by comparing before and after the launch, and over how long?
- Does "posts fall 2%" mean total posts in groups, posts per user, or the share of members who post?
- How is comments per user defined: total comments divided by all group members or by active members only, and do replies inside a thread count as comments?
- Which groups and surfaces received threading, and is the 2% drop statistically significant?
### What a Strong Answer Covers
- Checks that both movements are real: experiment design, significance, novelty effects, and how replies are counted.
- Plausible mechanisms connecting threading to fewer posts, covering both changes in member behavior and changes in what members are shown.
- Decomposition of each metric into how many members act and how much each does, plus segmentation by group type, size and member role, to tell the mechanisms apart.
- A judgment on whether the trade is healthy for groups, measured against the feature's goal and longer-term outcomes.
### Follow-up Questions
- If you could look at only one more cut of the data to explain the drop in posts, which would it be, and why?
- Would you ship threading to all groups as it is, and what result would change your mind?
- Suppose the posts that disappeared were mostly repeat questions already answered in older threads. How does that change your conclusion?
Overview: A product analytics question about adding threaded comment replies to group posts, after which comments per user rose 10% and posts fell 2%. It tests whether a product manager can check that the result is real, explain how the two metrics could be linked, and design analyses that tell competing explanations apart.