Cold-chain food recalls in France: the RappelConso figures

Cold-chain notices, month by month

By year

2021 (Apr-Dec) and 2026 (Jan-Aug) are partial. Comparing years on notice counts exaggerates the differences: a year with one large group looks high for a single event. The groups column partly corrects this.

By the product's storage temperature

The class is the one stated on the notice. The last column relates the cold-chain notices to all food recalls in the same class.

By product category

The categories are RappelConso's, as entered on each notice. "Other" is a category of the dataset, not a grouping made on this page.

The latest recalls concerned

The 10 most recent groups at the extraction date, including after August 2026. The product named is that of the group's first notice, with its official label in French; the link opens the RappelConso notice, which states what to do.

Method

Source and licence

The dataset "RappelConso V2 - Rappels de produits", published on data.economie.gouv.fr by the French administrations behind RappelConso (DGAL for food, DGCCRF), under the Licence Ouverte v2.0 (Etalab). One row per recall notice. Extracted on 27 September 2026, dataset last updated on 27 September 2026. The dataset on data.economie.gouv.fr.

Scope

The notices in the dataset's "alimentation" (food) category, which includes animal feed. The series cover the complete months from April 2021 to August 2026: the 8 earlier food notices (the first month of the series is partial) and the 180 notices of the month under way at extraction are left out; the list of latest recalls runs to the extraction date.

The classification rule

A notice is counted when its recall reason or its additional risk description, lower-cased and without accents, meets one of these three conditions (a notice meeting several is counted once). The rule was set after reading every food notice in which "froid", "température", "congel", "réfrig" or "ambiant" appears. The dataset is in French, so the rule matches French wording.

1. The text mentions the "chaîne du froid" (cold chain) together with a word saying it failed: rupture (break), impact, problème, non-respect, incident. (128 notices)/chaine d[ue] froid/ + /rupture|impact|probleme|non respect|incident/
2. The text describes a temperature excursion in other words: a temperature rise, a temperature anomaly or fault, a temperature threshold exceeded, freezing during transport. (6 notices)/remontee en temperature|anomalie de temperature|defaut de temperature|seuil de temperature[^.]*depasse|congelation lors du transport/
3. The text describes a product left out of refrigeration or a failing cold step: "hors froid" (out of the cold), fresh product stored at ambient temperature, technical problem in a cold room, thawing procedure not followed. (5 notices)/hors froid|produit frais en ambiant|probleme technique dans une chambre froide|processus de decongelation[^.]*non respecte/

Not counted:

• the health advice attached to notices (botulism, Bacillus cereus) that mentions the cold chain without it being the cause of the recall;
• "provided the cold chain is respected" and other storage advice;
• insufficient heat treatment or cooking, which are manufacturing faults, not storage faults;
• recipe faults that "affect keeping quality" (salt, acidity);
• labelling errors about storage conditions, and instability of a shelf-stable product at room temperature.

What these figures do not say

• A notice is not an incident. A single break can produce many notices, one per product reference (the largest group counts 36); that is why the page also gives the number of groups published on the same day with the same reason. The grouping is an approximation: two incidents worded alike on one day merge, one incident worded differently splits.
• A recall is not a food poisoning. The dataset describes withdrawal measures and holds no data on cases of illness.
• The classification depends on the words chosen by the recalling company. A notice citing only a "microbiological risk" without naming its cause is not counted, even if the cause was a break: the figure counts notices that name a temperature-related cause, not every break.
• The share is a share of published recall notices, not of products sold nor of all the cold-chain breaks that happened.

How to cite this page

KelvinTrack, "Cold-chain food recalls in France: the RappelConso figures", based on RappelConso (DGCCRF, DGAL), data.economie.gouv.fr, Licence Ouverte 2.0, data extracted on 27 September 2026. https://kelvintrack.com/en/shop/resources/cold-chain-food-recalls

The aggregates can be downloaded with the rule, the totals and the extraction date: JSON file. Any reuse of the source data must credit RappelConso and the date of last update, as the Licence Ouverte requires.

Frequently asked questions