CDG · Paris
ALE Q+A
Shipped a Q+A surface for Rainbow, a web communications platform with 100k+ MAU, turning scattered conversations into a queryable insight library.
- Role
- Product Designer (Intern, Paris)
- Timeline
- MAY 2025 – AUG 2025
- Technical Dialect
- Angular · RxJS · Elasticsearch · Figma
- 100k+Monthly active users
- 2.4×Knowledge re-use rate
Problem
Rainbow's power was its chat density — and its weakness was the same thing. Every new joiner re-asked questions a quiet expert had already answered three channels over.
Solution
I surfaced a Q+A layer that promotes the best answers out of channel noise, giving teams a memory layer without asking them to change behavior — the canonical answer lives where the conversation already happens.
§ Shipped
What I built
The Process
Design architecture
Optional deep dive — how the problem became the shipped product.
Channel archaeology with 12 teams
Twelve teams across three locales surfaced the same pattern: every channel had a quiet expert whose answers were rediscovered weekly by new joiners.
Promotion model for answers
I designed a lightweight promotion model — any message can become a canonical answer with one click, indexed and surfaced contextually on related questions.
Inline Q+A inside the chat rail
Inline panels render canonical answers inside the existing chat rail. No new tab, no new mental model — the memory layer lives where the conversation already is.