Our brands · USA CPG
Honest rankings for the U.S. CPG market.
Independent research and one explainable score per product, across 30 categories, built to be trusted by readers and cited by AI answer engines alike.
Visit usacpg.comarrow_forward
Most “best of” lists are a black box. A pile of products, a vague sense that someone decided they were good, and no way to check the work. USA CPG flips that around: every score breaks down into the signals behind it, and the editorial independence is stated plainly. Transparency isn't a footnote here. It's the product.
Two pillars
Editorial and rankings, reinforcing each other
The Brief: editorial
Long-form, sourced analysis of the U.S. CPG market, where research meets policy and the macro picture. It's the engine that earns reach and makes the rankings worth trusting in the first place.
Product rankings
Top-5 lists across 30 categories. Every product gets one explainable Quality Score, 0.0–5.0, computed by a scoring engine, then refined by human editors. Algorithm first, judgment on top.
The scoring engine
How the Quality Score works
| Signal | What it captures | How it's handled |
|---|---|---|
| Rating | Star rating | Bayesian-adjusted, so a 5.0 from a handful of reviews can't outrank a strong, heavily-reviewed product. |
| Review volume | Social proof / popularity | Log-scaled, rewarding proof with diminishing returns. |
| Value | Bang-for-buck | Quality-per-dollar, ranked against the rest of the set. |
| Sales rank | Real-world demand | Best-Seller-Rank converted to a points curve. |
- The score is always 0.0–5.0, one decimal.
- A product's score never drops below its own honest star rating.
- Only products clearing a quality bar get promoted. Weak products simply don't appear.
- Brand diversity is enforced; color/variant duplicates collapse into one entry.
- Human override on top: editors can name a staff pick, pin, or exclude a product, always applied after the algorithm and always recorded.
The data pipeline
Where the rankings come from
Rankings are only as good as their data, so there's a real pipeline behind them, not hand-typed lists. Catalog, pricing, star ratings, review counts, and best-seller rank get pulled from Amazon marketplace data. Every pull is saved as a snapshot, so the price-and-rating history builds over time and pages re-render without re-hitting paid APIs. Only U.S.-based sellers are eligible, and the refresh rate is tiered against a budget so the whole thing stays economical.
What makes it distinctive
Five angles on the same idea
- visibility
Transparency as the product
A single explainable score, a published methodology, and stated editorial independence, set against the opacity of typical "best of" affiliate content.
- balance
Algorithm + human judgment
A real scoring engine for objectivity, an editorial layer for taste, and an audit trail that captures both.
- smart_toy
Built to be cited
Engineered for AI answer engines and search as much as for human readers: structured data, clean semantics, sourced editorial.
- trending_up
A learning system
Today's rankings generate the telemetry that will tune tomorrow's: a path toward a self-improving recommendation engine.
- verified
Genuinely independent
A two-pillar property that stands on its own brand, infrastructure, and editorial voice.
Tech Stack
Built to be legible to readers and search alike
| Framework | Next.js 16 (App Router), React 19, TypeScript (strict) |
| Styling | Tailwind CSS v4 with a custom design-token theme; Fraunces + Inter |
| Data | Google Cloud Firestore (server-side, environment-scoped collections) |
| Content rendering | Markdown with GitHub-flavored syntax + sanitized rendering |
| Media | Google Cloud Storage for editorial and product imagery |
| Pipeline | Node.js ingestion/publishing scripts; commercial product-data APIs + Amazon's official product API |
| Hosting | Containerized (Docker, standalone build) on Google Cloud Run |
| CI/CD | Cloud Build: automatic staging and production deploys from Git branches |
| Analytics | Google Analytics 4 + Google Tag Manager (production only) |
Status
Live, and built for what's next
USA CPG is live across 30 categories. And every recompute already writes an audit trail: the score, the rank, any human override. That trail is the seed for what comes next: a self-tuning loop that learns which placements actually serve readers, and adjusts the weightings over time. The rankings get smarter the longer they run.
Our brands
Explore what else we've built
Want something like this for your business?
Scoring engines, data pipelines, the editorial layer on top. We build this kind of system for you, too. Let's talk it through.
Let's Talk