# Answer Key: Feature Adoption and Retention

Use this as a calibration guide, not a single correct answer.

## Core Readout

The naive comparison likely shows higher retention among workspaces that used collaboration in week 1. That should be treated as an association, not a causal estimate, because collaboration users also differ on pre-treatment characteristics such as team size, baseline sessions, invites sent, plan, and acquisition channel.

## What A Strong Answer Should Say

- Name the causal question before estimating: "What would retention have been if comparable workspaces had or had not used collaboration in week 1?"
- Identify plausible confounders: larger teams, more engaged workspaces, paid plans, and invited teammates may all affect both collaboration use and retention.
- Avoid adjusting for variables that happen after collaboration use if they may be mediators.
- Discuss overlap: if only large team workspaces adopt collaboration, small free workspaces cannot support the same counterfactual comparison.
- Recommend a next design: adjustment plus sensitivity analysis, a randomized collaboration prompt, or an encouragement design.

## Common Mistakes

- Concluding that collaboration caused retention just because adopters retained more.
- Treating baseline engagement as an outcome rather than a confounder.
- Ignoring acquisition channel and plan.
- Using a complex model to hide weak identification.

## Instructor Notes

Ask learners to draw a small DAG with baseline engagement, team size, collaboration use, and retention. The best answers make the assumptions visible before proposing a model.