Feature request intelligence dashboard
Build a "Feature Request Intelligence" workflow: a browsable intelligence app over our customer feature requests. The goal is to let a CS or product person search every request for a concept, see which themes are heating up across the customer base, drill into a single customer, and compare customers side by side. It is strictly read-only: nothing is ever written back to the source system. Structure it as three steps: A sync step on a daily cron (06:00 UTC) that is also callable on demand via its own route. It pages through the Tines Records API — or your own request/ticket source — using an API key connector, trims each record to the fields the app needs, and writes a compressed dataset plus a small metadata file into a persistent named volume. It must refuse to overwrite a good cache with an empty or failed result, so a bad API call leaves yesterday's data in place. A data route step that streams the cached dataset (gzipped JSON) straight from the volume to the browser. This is called by the app, not by people. A React app step served at its own route as the human entry point. All analysis happens client-side after one fetch of the cached dataset, so exploring is instant and sends no further requests: normalization of raw status values into a small set of sensible groups, customer deduplication (group by company id, falling back to a normalized company name so lookalike spellings stay one customer), full-text search with relevance ranking and synonym/fuzzy handling, topic "temperature" scoring that weights unique customers more heavily than raw volume so one noisy customer can't fake a hot topic, clustering, and charts. Include a methodology panel that surfaces unmapped/unknown values and states data caveats honestly rather than guessing. If the cache is missing (e.g. right after a publish, since volume data doesn't carry over from a draft), show a friendly empty state with a button that calls the sync route to build the dataset. Keep all routes private to space members. Give the app a polished, dense-but-legible dashboard design with cross-filtering and shareable URL state. Tools used: Tines, 3B, React + Tailwind, TypeScript
What this prompt builds
A read-only intelligence dashboard that syncs feature requests from Productboard (or any source) into Tines, then lets Customer Success and Product teams search, analyze, and visualize thousands of requests in real time. All analysis runs client-side after a single data fetch, so exploring trends, comparing customers, and identifying hot topics is instant. It surfaces which features customers want most, tracks request status, and helps prioritize product decisions without writing anything back to the source system.
The problem
Customer Success Managers and Product teams struggle to manage, search, and track thousands of feature requests scattered across systems like Productboard. Finding related requests, understanding which customers are asking for the same features, and spotting emerging product trends requires tedious manual work. This workflow solves that by syncing feature requests into a single, searchable intelligence dashboard. It provides instant full-text search with fuzzy matching, analytics that weight unique customers over raw volume so one noisy customer can't skew priorities, cross-filtering visualizations, and the ability to drill into a specific customer or compare customers side by side—all without writing anything back to the source system.
Solution and impact
The workflow delivers a read-only dashboard that pulls feature requests into Tines daily (or on demand), caches them locally, and streams the dataset to a React app where all analysis happens client-side. Customer Success and Product teams get instant access to powerful search, trend detection, status tracking, and customer insights without sending repeated API calls or waiting for queries to run. This saves significant time hunting through records, improves the quality of customer conversations by surfacing what matters to each account, and helps prioritize features based on real customer demand. The methodology panel also surfaces data caveats and unmapped values honestly, so teams can trust the intelligence they're acting on.