Tree‑shake unused CSS Module classes by analyzing imports & usage

Many frontend teams today — from single-repo startups to large organizations with monorepos — accumulate non-trivial amounts of unused CSS through churned components, feature flags, and conditional imports, which inflates bundle size and slows time-to-interactive; this pain is acute for the estimated 900,000 frontend teams that buy build/perf tooling, a market we size at $2.7B with an average $3,000 ACV. The people who feel it most are frontend engineers, performance leads, and platform teams responsible for keeping bundle budgets tight across many apps. You could build a developer tool that statically and heuristically analyzes ESM import graphs and CSS Module export usage to deterministically remove unused classes at the export level, offering a Vite/esbuild/Turbopack plugin, a CLI for CI gating, and an optional code-mod or safe-removal mode that preserves source maps and provides rollbacks. Key capabilities would include cross-file analysis to map imports to class tokens, runtime-aware heuristics for dynamic className patterns, and a clear report/dashboard that quantifies bytes saved and potential regressions; practical challenges include handling highly dynamic patterns and minimizing false positives. This is an attractive time to enter the market because teams are prioritizing web performance, modern bundlers expose module graphs that make export-level shaking possible, and componentized styling (CSS Modules/CSS-in-JS) is widespread — all factors behind the market score of 88/100 and revenue potential of 76/100. With relatively low competition, you can differentiate by focusing on provable correctness, deep bundler/runtime integration, conservative safe-defaults for enterprise workflows, and an easy CI integration path; the main obstacles will be building robust heuristics for dynamic usage and earning trust through reproducible, minimal-risk removals.