Ecobalyse score verification: having a collection checked
What does the French DGCCRF check?

As of 17 October 2026, 188 brands had declared 118,566 product references of clothing on the official environmental cost portal, more than 600 on average per brand. Each product reference carries a score calculated with the official French method for the environmental cost (coût environnemental, also known as the French eco-score), using Ecobalyse or software that reproduces it, and each one commits the brand. The portal's terms of use state that any information submitted is the sole responsibility of the declarant.

Checking a collection therefore means verifying hundreds of calculations and, above all, the data that feeds them. The calculation can be rerun automatically through the tool's application programming interface (API). What takes time is the evidence: the composition, mass and countries of manufacture of each garment, which the brand obtains from its suppliers. The sections below start from what an official inspection looks at, then set out the method for an entire collection.

1Who answers for the scores declared on the portal?

The brand, and the brand alone. Decree No. 2025-957 of 6 September 2025 requires the manufacturer, importer or other person placing the product on the market who displays an environmental cost to declare it on the portal before any communication. It states that this person is responsible for the data filed there. Labelling remains voluntary, but a brand that displays a score no longer chooses its own rules.

What the public sees, what the inspector sees

The decree separates 2 levels of information for each declared product reference, with the ministries and ADEME (the French Agency for Ecological Transition) using the detailed data only to monitor the scheme:

• The public sees the score, its breakdown by impact category, the durability coefficient, the date, the method version and the legal status of whoever did the calculation.
• Officers of the DGCCRF (the French consumer protection and fraud control authority) also see, for inspection purposes, the data used for each parameter: mass, materials, country of each stage, price.

This split shows where an inspection bites. The public score is only the result; what gets verified is each input data point, product reference by product reference, and the document that supports it.

A service provider does not carry the responsibility

A brand may entrust the calculation to software, an engineering consultancy or its supplier: the portal then shows the legal status of whoever did the calculation. Responsibility, however, cannot be delegated. The portal's terms of use point out that a false declaration exposes the declarant to the penalties of Article 441-1 of the French Criminal Code, up to 3 years' imprisonment and a €45,000 fine.

Since 1 October 2026, anyone may calculate and publish the environmental cost of a garment, based on available or estimated data (what changed that day). If the brand declares its own score, that score prevails, and the third party must align with it within 1 month. A checked and documented score therefore also addresses one of the risks of publishing nothing.

2What does a DGCCRF inspection check?

An inspection may start from the portal, where "Contrôleur" (inspector) accounts receive the declarants' calculation data and can recalculate each environmental cost, so that declarants can be warned of any error. They also check that each product is linked to the right manufacturer or the right importer. DGCCRF officers may then impose penalties on the basis of the checks and calculations carried out.

Recalculation does not tell you whether the data is true. For that, Article R. 541-246 of the French Environmental Code, inserted by the decree, requires anyone who calculates or communicates a score to keep the elements that justify the calculation available to officers. The text sets no retention period: the most prudent course is to keep the file for as long as the score can be inspected.

What the methodology notice anticipates for an inspection

The methodology notice of 1 October 2025, to which the ministerial order refers, itself anticipates 3 inspection situations:

• For a product reference manufactured in several countries, the country producing the largest share of the mass is used, a method the brand must be able to justify, in particular during an inspection.
• The number of product references declared for range width is a commitment: on any inspection date, the product references sold on that same day must remain below that number.
• A recalculation may differ from the declared score by around 1% because of rounding, with no systematic bias; the notice states that this gap is not a regulatory tolerance.

What a wrong score costs

A label that does not follow the method is prohibited by Article L. 541-9-15 of the French Environmental Code, on pain of an administrative fine of up to €15,000 for a company. If it misleads consumers, it may also constitute a misleading commercial practice under French consumer law, which expressly covers the environmental impact of a product. This is punishable by 2 years' imprisonment and a €300,000 fine, raised to 5 years and €750,000 when the practice is carried out through a website. The fine may reach 80% of the campaign's expenditure for an environmental claim, as for any banned claim. The details of these consequences are in who inspects and who sanctions.

3What evidence should you keep for each product reference?

The ministerial order of 6 September 2025 lists 17 calculation parameters. 6 are mandatory and rely on data specific to the product reference: the category, the mass, the materials, and the countries of weaving, dyeing and finishing, and garment making. The other 11 may be entered; otherwise a default value applies. For a collection, these parameters are best grouped by who holds the data.

The brand's data

3 parameters come from the brand itself. They are verified only once, from internal documents, and an error on the first 2 is repeated on every product reference:

• The maximum number of product references sold on the same day in each segment (women, men, children, babies, underwear), supported by a dated catalogue extract.
• The size of the company and its repair service, which only counts if it is approved by the producer responsibility organisation (PRO, éco-organisme in French) of the textile extended producer responsibility (EPR) scheme under the repair bonus.
• The reference price of each garment, incl. VAT, excluding promotions and sales, on the sales channel used for range width.

The product data

The category, mass and composition come from the product sheets. The mass is that of the finished product, and any material weighing at least 2% of the mass and 5% of the impact must be declared. A garment made of several pieces may be calculated on its main component, the heaviest one. The simplest evidence remains weighing one unit and the composition label, which must already comply with the EU regulation on textile fibre names.

The suppliers' data

The 3 mandatory countries, and those that avoid a default value, come from the supply chain. An unknown country is accepted only for the origin of materials and spinning. Each stage has its own document, and the questionnaire described in what labelling costs makes it possible to gather them once per supplier rather than once per garment.

A certified material, such as organic cotton or recycled polyester, is better supported by more than a supplier declaration: the transaction certificate for the batch. The logic is the same as for suppliers' carbon data or the evidence in a packaging technical file: an assertion without a document does not survive an inspection, any more than a claim without proof.

4How do you check a collection without redoing everything?

Reworking 600 files one by one makes no sense, and a random draw lets the product references that matter slip through. The method that holds up has 3 stages. You recalculate the whole collection, verify the brand and supplier data once, then prove the product data on a chosen sample.

Recalculate everything first

The regulatory version of Ecobalyse is still v7.0.0, and its API reruns the calculation for a product reference from the same parameters. Running the whole collection through the API costs almost nothing and catches 4 families of errors:

• Scores calculated on a version other than v7.0.0, for example in the "Textile" tab of the website, which opens a development version.
• Gaps of more than 1% between the declared score and the recalculated score, which point to an entry different from the one kept in the file.
• Product references missing a mandatory country, or whose spinning was left at the default value although the country is known.
• Internal inconsistencies, such as 2 garments of the same category, same material and same countries whose scores diverge sharply.

Where an error costs the most points

Not all data carries the same weight, and it is not always the data that textile eco-design works on first. On a 150 g cotton T-shirt made in Pakistan, an official Ecobalyse example at 742 points, we varied one data point at a time in the version 7.0.0 API. A wrongly declared material or a range width left at its default shifts the score by nearly 190 points, whereas the country of garment making shifts it by only around ten.

The gap does not always work against the brand. In the regulatory version, dyeing declared in Pakistan when it took place in China adds 52 points to the score. That error penalises the brand, but the reverse error would have flattered it by the same amount. This is why the sample concentrates on the data that carries weight, whichever way the error runs.

Choosing the sample

No legal text sets a sampling rule. A useful sample is chosen on risk rather than at random:

• Product references whose score is promoted, in store, online or in a campaign, and the best scores in each category, which draw attention.
• Garments in certified, organic or recycled material, where the possible gap runs into hundreds of points without a transaction certificate.
• Product references made in several countries, those from a new supplier and those with a wash, a print or several components.
• A few product references drawn at random in each category, to check that the data collection chain also works where nobody is looking.

When to start again

A score is updated no more than once every 3 months. A change of method requires it to be updated within 12 months, unless it is already printed on the product. The check is therefore repeated with each new collection, each change of supplier and each new regulatory version of the tool. Part of this file should also serve for the Digital Product Passport (DPP) planned for textiles.

5Who can carry out the check, and what should you ask a provider?

No body is approved to verify an environmental cost, and no legal text provides for certification of scores. 3 levels of checking exist, and they complement each other:

• An internal review, by someone who did not enter the data, is enough for data-entry and version errors if it follows a written checklist.
• The publisher of the calculation software knows its own configuration, but it is checking its own work, as with a life cycle assessment (LCA) carried out with software.
• A third party, engineering consultancy or consultancy firm, brings an outside view and a sampling method, provided it did not carry out the calculations it is checking.

It is the same independence rule as for an LCA critical review or a carbon footprint assessment verified by a third party: whoever did the calculation does not check their own work.

The questions to ask

Before signing, 5 answers let you compare 2 proposals:

• Does the provider recalculate every product reference on the regulatory version, through the API, or only those in the sample?
• How does it choose the sample, and how many product references from each category, each supplier and each certified material will it include?
• Does it go back to the documents (invoices, transaction certificates, weighings) or does it rely on supplier declarations?
• Does it report each discrepancy in points, with the correction to declare and the date from which the 3-month rule allows it to be declared?
• Did it take part in calculating the scores, configuring the software or collecting the data it is going to check?

What the report contains

The useful deliverable is a list of discrepancies, product reference by product reference, with the document consulted and the effect in points. The example below was built for this article, with discrepancies calculated in the version 7.0.0 API. The costliest comes from a material described as organic without a certificate: more than 700 points on a jumper. None of these discrepancies shows up on recalculation, since each score was correct for the data entered and the error lay in the data.

The price

There is no public price list for this type of check. The price depends on the number of product references, the number of suppliers and the state of the data in the brand's product software. Recalculating the whole collection is quick when the declarations went through the API; the time goes into the suppliers' documents, as with auditing the claims across a product range.

At Projet Celsius, checking a collection starts with recalculating all its product references. It ends with the suppliers' documents, where discrepancies of several hundred points are hidden. A brand that keeps, for each product reference, the document proving each mandatory data point can answer the DGCCRF without reconstructing anything, and can set its own score against a third party's.

6Key takeaways

• No body is approved to certify an Ecobalyse score: the brand alone answers for each data point it declares on the portal.
• The DGCCRF may recalculate each score from the portal and demand the elements that justify the calculation, with no time limit set by the text.
• A certified material without proof, a wrong mass or a default range width shifts a score far more than the country of garment making.
• The method that holds up for a collection: recalculate everything through the API, verify the brand and supplier data once, then prove a sample chosen on risk.
• The check is repeated with each collection, each new supplier and each new regulatory version of the tool.

To read a score and use the tool, see our Ecobalyse guide; for the full framework, see French environmental labelling for textiles.

Further resources

Frequently asked questions

1Do Ecobalyse scores need to be certified?

No. No legal text provides for certification of scores or for an approved body to verify them. The brand declares its scores on the official portal and remains solely responsible for them. A check by an internal review or by a third party is voluntary; it serves to find errors before a DGCCRF inspection or before a third party publishes a different score.

2Can the DGCCRF inspect an Ecobalyse score?

Yes. Its officers can access each product reference's calculation data on the portal and recalculate the score. Decree No. 2025-957 also requires the brand to keep the elements that justify the calculation available to them. A non-compliant label is liable to an administrative fine of up to €15,000 for a company.

3How long should you keep the evidence behind an environmental cost score?

The decree sets no period. It only requires the elements that justify the calculation to be kept available to inspection officers. The most prudent course is to keep each product reference's file for as long as its score is displayed or viewable on the portal, and to keep every previous version when a score is updated.

4What margin of error is allowed on an Ecobalyse score?

The methodology notice accepts that a recalculation may differ from the declared score by around 1%, because of rounding, provided the gap does not reflect a systematic bias. It states that this is not a regulatory tolerance: a wrong data point remains an error, even if its effect on the score is small.

5What if a third party publishes a different score for one of my garments?

Declare your own score on the portal, with justified data. Since 1 October 2026, a third party may publish the environmental cost of a garment from estimated data, but the score determined by the brand prevails: whoever was communicating a different figure must update it within 1 month.

6How much does it cost to check the scores of a collection?

There is no public price list. The price depends on the number of product references, the number of suppliers and the state of the brand's data. Recalculating all product references through the Ecobalyse API is quick; the time goes into verifying the suppliers' documents on a sample of product references.

Ask Guillaume your questionA personal reply within 24 working hours

Send my questionor: guillaume@projetcelsius.com