Benchmarking methodology

The goal of the benchmarking functionality within Carbon+Alt+Delete is to compare the carbon footprint of a given company with the carbon emissions of its peers. The functionality provides an emission range in similar industries, as well as where the given company is on this range.

Benchmarking dataset

To create reliable benchmarks, we start by collecting and preparing a high-quality dataset from companies reporting emissions on the Carbon+Alt+Delete platform.

One of the key strengths of the benchmarking dataset is its combination of granularity and standardisation:

• Granularity: Emissions are captured at the level of individual activities (e.g., specific fuel types, transport modes, procurement categories). This provides a rich basis for accurate comparisons.
• Consistency: All companies on the platform report using the same underlying data structure. This structural consistency ensures that data is comparable across organisations, sectors, and reporting periods—something that is rarely possible with public, unstructured climate disclosures.

To ensure that the dataset reflects realistic and comparable emission profiles, we apply several filtering steps. Firstly, we remove incomplete footprints, such as partially reported scope 3 data or footprints still in progress. Secondly, we remove outliers with unrealistic or unreliable values—for example, implausible emissions, turnover, or employee figures. This cleaning process helps maintain a benchmarking dataset that is both credible and statistically robust.

The benchmarking dataset is updated on a regular basis to incorporate newly reported emissions data. This ensures that benchmarks remain relevant, reflective of current performance, and aligned with the latest reporting cycles.

Note that all benchmarking data is completely anonymised. No company names, site names, or identifiable metadata are included. Also, the dataset is not reverse-engineerable, ensuring full confidentiality for all participating organisations.

Benchmarking calculation

Once the dataset has been prepared, we calculate the benchmarking figures by applying a consistent and transparent methodology. The first step is to filter the dataset so that each company is compared only with relevant peers. This filtering is based on the company’s sector, using NACE classifications, and on the selected emissions scope—either Scope 1–2 or Scope 1–2–3. By narrowing the dataset in this way, we ensure that benchmarking results reflect meaningful comparisons within a similar group of organisations.

After filtering, we calculate emissions for each company included in the benchmark. This involves summing all emissions that fall within the chosen scope and then applying the selected type of normalization. Companies can be compared on total emissions, emissions per full-time equivalent (FTE), or emissions per turnover.

With normalized emissions calculated, we then derive the statistical values that form the basis of the benchmark. For every peer group, we determine the sample size and compute key percentiles, including the minimum value, the 25th percentile, the median, and the maximum value. These percentiles define the distribution against which your company is compared.

Using this statistical output, we create the graphical benchmark representation shown in the platform. The percentile values are transformed into the bar graph format, which is a 1-dimensional representation of a probability distribution.

Finally, we calculate your company’s own position on the graph. Your emissions—based on the same scopes and normalization options applied to the benchmark group—are plotted on the same relative scale. This allows you to clearly see how your performance aligns with, exceeds, or falls behind similar companies in your sector.

Notes Limitations

To ensure transparency, here are the key caveats to keep in mind:

• Scope 2 methodology: we always include location-based scope 2 emissions, even if market-based values are also provided.
• Reporting bias: benchmarks only include companies with complete reporting in our platform. This may introduce some bias toward more climate-active companies.
• Missing business data: where business data was missing, we used an LLM to estimate values. Only estimates with sufficiently high reliability scores were included.