← AI Lab
PrototypeFeedback ops · Claude skill

Signal

Turns a firehose of raw customer feedback into ranked, RICE-scored themes — so the loudest request isn't automatically the next thing you build.

Role
Product & build
Stage
Prototype
Stack
LLM · Sheets
Focus
Feedback triage
The problem

Feedback is a firehose

Support tickets, sales calls, app reviews, Slack threads — feedback arrives faster than any team can read it, and the items that get acted on are usually just the ones shouted loudest or most recently.

Signal ingests all of it, clusters it into themes, and scores each theme against RICE so prioritization starts from evidence instead of the last conversation you had.

How it works

Three moves

01

Ingest

Takes raw feedback from any source — pasted, exported or piped — and normalizes it into one list.

02

Cluster

Groups items into themes by underlying need — not keyword — and counts how many voices sit behind each.

03

Score

Applies RICE to each theme and returns a ranked table you can defend in a prioritization review.

What I learned

Notes from building it

Clustering by need beats clustering by wording. Two customers can ask for opposite features to solve the same underlying job — grouping on the job is where the insight lives.

Confidence is the honest part of RICE. Signal shows its working for every score so a low-confidence theme reads as "go learn more," not "go build."

The output is a starting point for a conversation, not the decision. It moves the debate from "who remembers what" to "here's the evidence — now let's argue."