Read article in:
Our analysis shows that 29% of products list fewer than 5 ingredients. As typical cosmetic products contain between 15 to 50 ingredients, this suggests incomplete ingredient lists (Science.org.au). Furthermore, 15% of the products analyzed provide no ingredient lists at all.
This is not in line with EU regulations that require full ingredient disclosure in descending order of concentration (EU Cosmetics Regulation (EC) No 1223/2009).
An example of how important the correct order in the ingredient list is concerns Methylparaben, a preservative with strict use restrictions. Among the 600 products, 123 were found to contain Methylparaben, with 120 having it among the top 3 ingredients. This suggests concentrations likely to exceed the EU maximum permitted levels of 0.4% when used alone or 0.8% when combined with other parabens.
Among the 600 products analyzed, 41% were classified with a high-risk chemical profile, 37% with a medium-risk profile, and 7% with a low-risk profile. 15% of products did not receive a rating due to incomplete data, as in cases where an eyeliner listed only one ingredient, such as water.
Now let's dive into the ~500 chemicals identified across cosmetics. In the data visualization below, the size of each bubble reflects the frequency of a chemical's occurrence across products, while its color indicates the average risk score for that chemical across the products in which it occurs.
Our analysis shows that while only 9 chemicals are classified as high risk and 26.25% as medium risk, the majority (72%) are considered low risk. However, many products still receive a chemical score as high risk due to the presence and frequency of these 9 chemicals classified as posing significant health or environmental risks. The 'medium risk' category (yellow) includes products with a neutral, medium and high risk profile.
How we did the analysis
Complir has automated processes for collecting data. And normally, collecting and analyzing the products on Temu would take a few hours to set up.
However, this didn't prove possible on Temu due to various verification tests (e.g. 'I'm not a robot' checks) and other tools that block data collection. We could have developed an AI-driven data collection tool here, which would take 1-2 weeks. But since such a tool would never be 100 percent generic, but would have to be customized to a website's specific blocks, we chose to collect data manually for the purpose of this analysis.
How we analyzed the data: We used the European Commission's cosmetic ingredients database (CosIng) to identify and verify chemical names, including potential misspellings. CAS numbers (unique identifiers for chemical substances) of these chemicals were cross-checked against the European Chemicals Agency (ECHA) database for hazard information.
To enhance our analysis, we applied the Complir engine, which uses Generative AI to describe chemicals, identify relevant legislation, assess risks based on ECHA hazard data, and justify risk assessments.
Once Complir is set up - including automated data collection - any webshop will have a full overview of potential risks and non-compliance with regulations and legislation in an instant - and of course be able to continuously ensure that all products comply with standards.
