Engine v3: Expanded Coverage, Confidence Scores, and a Hardened Platform

A few weeks ago we published a deep dive into the architecture of our LCA scoring engine — how it implements the EU Product Environmental Footprint methodology, what public datasets feed it, and why multi-layer estimation matters. That was Engine v2.

Today we are shipping Engine v3. This is the largest update since launch: broader product coverage, more destination countries, confidence scores on every analysis, per-plan rate limiting, a security-hardened API platform, and completely rewritten documentation. Here is what changed and what it means for teams building on the GreenMetric API.

---

Broader Coverage: More Products, More Countries

The v2 engine shipped with 31 product category templates covering electronics, apparel, and furniture. That was enough to score most consumer products, but it left gaps — phone chargers were classified as generic electronics, laptop cases were lumped with backpacks, and small kitchen appliances had no dedicated template.

Engine v3 expands the taxonomy to over 50 product categories, with eight entirely new category types:

• Charger — wall adapters, wireless chargers, multi-port USB chargers
• Cable & Adapter — USB-C, Lightning, HDMI, and display cables
• Phone Case — silicone, leather, and plastic phone and tablet cases
• Power Bank — portable battery packs of all capacities
• Smartwatch — wearable devices with dedicated lifecycle profiles
• Speaker — portable Bluetooth and home speakers
• Keyboard & Mouse — wired and wireless input peripherals
• Small Appliance — coffee makers, toasters, blenders, and similar countertop devices

Each new category carries its own lifecycle template — manufacturing energy profiles, material composition defaults, use-phase power consumption, and end-of-life recyclability rates. A phone charger now gets scored with the right weight, the right materials, and the right use-phase energy draw rather than being treated as a generic electronics product.

The engine also now supports 20 destination countries, up from 13 in v2. This means the scoring engine can model country-specific electricity grid carbon intensity for more markets — important because a product used in Sweden (almost entirely renewable grid) has a meaningfully different use-phase footprint than the same product used in India or China. The new countries include Mexico, Thailand, Indonesia, Turkey, Cambodia, Portugal, and Sweden, covering the major manufacturing and consumer regions that our users operate in.

We also improved how the engine extracts material and weight information from product descriptions. The classifier now detects more material types, handles compound descriptions like “recycled polyester blend” more accurately, and produces better weight estimates for products where weight is not explicitly stated. The result: the activity-based estimation layer activates more often and produces higher-confidence results.

---

Confidence You Can Trust: Uncertainty Quantification

In v2, every analysis returned a score — but it did not tell you how confident you should be in that score. A product with a full bill of materials and a well-studied category got the same type of result as a product with a vague two-word description and an estimated price.

Engine v3 changes this. Every score now ships with confidence bands and a data quality grade.

The data quality grade — high, moderate, low, or very low — reflects how much of the score was derived from identified materials and measured inputs versus estimated defaults and sector averages. A product where the engine identified three specific materials and knew the exact weight gets a “high” confidence grade. A product where only the price and category were available gets a “low” grade. Both products still receive a score, but now you know the difference.

Confidence bands provide a numerical range around the score. Rather than just “72 out of 100,” the API now returns something like “72 ± 8,” meaning the true score likely falls between 64 and 80 given the data quality. The grade range is also reported — “B to A” — so you know whether the letter grade is stable or might shift with better data.

This matters for EU compliance. The Green Claims Directive and the Empowering Consumers for the Green Transition regulation both require that environmental claims be substantiated and transparent about their basis. Reporting confidence alongside scores is how GreenMetric helps businesses meet that bar — you are not just showing a number, you are showing how that number was derived and how reliable it is.

---

Better for Developers

Rate Limiting with Per-Plan Tiers

Every API response now includes standard rate limit headers: RateLimit-Limit, RateLimit-Remaining, and RateLimit-Reset. When you hit a limit, the Retry-After header tells you exactly when to retry.

Rate limits are tied to your plan tier. Free accounts get 10 requests per minute — enough for development and testing. Starter, Pro, and Enterprise plans scale up to 500 requests per minute. Monthly quotas are also enforced per organization, with clear feedback when you approach or exceed your allocation. No more guessing whether a 429 response means “slow down for a second” or “you have used your monthly quota.” The response body now tells you which limit you hit.

Error Tracking and Monitoring

We invested heavily in production observability for v3. Every API request is traced end-to-end with structured logging that includes organization context, authentication type, and request timing. Errors are captured and triaged automatically — when something fails in the scoring pipeline, our team is alerted within seconds with full context.

For API consumers, this means faster incident response. If you report an issue, we can trace it to the exact request, the exact scoring step, and the exact data condition that caused it. We also added enriched error responses — instead of a generic “Internal Server Error,” the API now returns structured error objects with error codes, human-readable messages, and, where applicable, suggestions for resolution.

Rewritten API Documentation

The API documentation has been rewritten from scratch. Every endpoint now includes complete request and response schemas — no more guessing what fields are required or what the response shape looks like. The docs also include a rate limit reference table by plan, a complete error code catalog, authentication guides for both API key and JWT flows, and realistic example responses.

We removed endpoints that were documented but not yet implemented, and added endpoints that existed but were undocumented. The docs now reflect exactly what the API does today — nothing more, nothing less.

---

Security Hardened

Engine v3 is not just a scoring update. We rebuilt the platform’s security layer to be production-grade.

Session management now includes automatic token refresh. Access tokens are short-lived and rotate silently in the background — if a token expires mid-session, the platform refreshes it without interrupting the user. Logout revokes tokens server-side, so a compromised token cannot be replayed after sign-out.

API key restrictions give you fine-grained control over how your keys are used. You can restrict a key to specific IP addresses, specific request origins, or specific API endpoints. A key scoped to /v1/analysis from a single IP address cannot be used to access billing endpoints or make requests from an unauthorized server. Restricted requests are blocked before they count toward your usage quota.

Server-side route protection ensures that every authenticated page in the dashboard is protected at the middleware layer — before any page content is rendered. Unauthenticated requests are redirected to the login page server-side, not via a client-side JavaScript check. This closes the gap where protected pages could briefly flash content before redirecting.

These are not features we bolted on. They are how the platform works now — session rotation, key scoping, and server-side auth checks are active on every request, for every user, on every plan.

---

What’s Next

Engine v3 is the foundation. Here is what we are building on top of it:

• Shopify integration — automated product scoring at import, with an embeddable sustainability widget for your storefront product pages
• Embeddable widgets — drop a score badge on any product page with a single script tag, configurable theme and size, powered by your API key
• Client SDKs — official libraries for JavaScript and Python so you can integrate product scoring in a few lines of code instead of raw HTTP
• Compliance reports — downloadable, audit-ready reports per product that substantiate environmental claims for EU Green Claims Directive and Digital Product Passport requirements

The EU’s Empowering Consumers for the Green Transition Directive takes effect in September 2026. The Digital Product Passport begins rolling out for textiles and furniture in 2027. We are building the data infrastructure that helps businesses meet these deadlines with real, science-backed environmental data rather than vague marketing claims.

---