How URLCarbon calculates a product's carbon footprint

Each URLCarbon report estimates the greenhouse gas (GHG) emissions associated with one unit of the analysed product, expressed in kilograms of carbon dioxide equivalent (kg CO₂e) using 100-year Global Warming Potentials (GWP100) consistent with the IPCC Sixth Assessment Report.

The default system boundary is cradle-to-grave, covering:

• Raw material extraction and processing
• Component and product manufacturing
• Primary and secondary packaging
• Inbound and outbound transport and distribution
• Product use phase (where relevant — e.g. energy-consuming products)
• End-of-life treatment (recycling, incineration, landfill)

The functional unit is a single retail unit as sold. Capital goods, corporate overheads, and employee commuting are excluded, in line with standard product-level LCA practice.

Reports include the seven Kyoto Protocol gases where they are material to the product's lifecycle: CO₂, CH₄, N₂O, HFCs, PFCs, SF₆ and NF₃. All gases are converted to CO₂-equivalent using GWP100 factors and reported as a single aggregated figure.

Emission factors are selected based on the best available match for the product's geography, technology and time period. Where multiple factors are applicable, we use the source most commonly cited in peer-reviewed product LCAs for that category.

1. 1

   Data extraction

   The product URL is fetched and parsed. Structured metadata (JSON-LD Product schema, Open Graph, retailer-specific fields) is extracted first, then supplemented with visible product copy — materials, weight, dimensions, country of origin, packaging, warranty and end-of-life instructions.

2. 2

   Material & component classification

   Extracted attributes are mapped to lifecycle inventory categories (e.g. cotton, recycled PET, aluminium 6061, lithium-ion cell, cardboard). Composite products are decomposed into their principal components by mass share.

3. 3

   Lifecycle stage modelling

   Each stage — materials, manufacturing, packaging, transport, use and end-of-life — is modelled independently using the datasets above. Transport is modelled from country of origin to destination market using GLEC modal factors. End-of-life uses country-specific waste treatment mixes.

4. 4

   Aggregation

   Stage-level emissions are summed to a total kg CO₂e per unit and normalised to the report's headline figure, category benchmark comparison, and per-order calculator.

5. 5

   AI review & consistency checks

   The draft footprint is cross-checked against peer LCA studies for the same category. Outliers (>2σ from category mean) are flagged, re-modelled, and either corrected or documented as legitimate outliers.

Product pages rarely disclose every input required for a full LCA. Where complete product information is unavailable, URLCarbon uses AI-assisted material recognition, engineering estimates, and category-specific modelling to infer missing values.

Typical inferred fields include material composition, product mass, packaging weight, country of manufacture, and typical use-phase duty cycle. Inferences are constrained by category priors derived from peer-reviewed LCA studies rather than free-form estimation.

These assumptions contribute to the report's confidence score and are transparently disclosed in the "Assumptions" and "Methodology & data sources" sections of every PDF. Increasing the underlying data quality — for example by supplying a bill of materials or a product spec sheet — raises the confidence score and can materially change the headline figure.

Every report carries a confidence score from Low to High, driven by four inputs:

• Reports are estimates suitable for internal decision-making, supplier screening, buyer communication and category benchmarking. They are not a substitute for a certified, third-party-verified LCA or an EPD.
• Emission factors carry inherent uncertainty (typically ±10–30% at the unit-process level). Aggregated product footprints inherit this uncertainty.
• Use-phase emissions for products with highly variable consumer behaviour (e.g. clothing wash cycles) are modelled on category-typical assumptions, not individual user behaviour.
• Biogenic carbon, land-use change, and non-GHG environmental impacts (water, biodiversity, toxicity) are not included in the headline figure.

Emission factors and category benchmarks are refreshed as source datasets publish updated versions (DEFRA annually, IEA annually, Ecoinvent on their release cadence). Each report records the methodology version used at the time of generation, so historical reports remain reproducible even as factors evolve.