Ingredient Counts Run From 1 to 87 in Our Catalog. No Study Says the Number Predicts Anything.

The shortest ingredient list we sell has one line on it. Jojoba oil, and then nothing.

The longest has 87. It is a spicule ampoule, and reading it takes about as long as reading the first page of this article.

Both of those products passed our screening. The one-line one scored 9.8. The 87-line one scored 9.1. Somewhere in the gap between those two numbers is a belief that a lot of people hold and that we have quietly leaned on ourselves: that a shorter ingredient list is a safer ingredient list. We went to find the evidence for it this week. We came back with nine search queries and no paper.

What the distribution actually looks like

We recounted every row this week rather than trusting the figures in our own notes, and that turned out to be the right call: our catalogue grew from 54 products to 56 in the last few days, and the totals moved with it.

Look at the bottom two rows against the top three. The 66-ingredient cream and the one-ingredient oil got the identical 9.8. The lowest score in our entire catalogue, an 8.0, belongs to a product with 13 ingredients in it.

We ran the arithmetic properly rather than eyeballing it. Pearson correlation between ingredient count and our own screening score across all 56 products: 0.05. For a sample this size that is indistinguishable from no relationship at all.

Which is a slightly awkward thing for us to publish, because our score is supposed to be a considered judgment and it turns out to be completely uninformed by a variable most shoppers treat as decisive.

We went looking for the study. It is not there

Before anything else, a confession about method, because we got this wrong on the first attempt.

You cannot search PubMed for the phrase “number of ingredients”. The word “of” is a stopword in its phrase index, so the database silently declines to treat it as a phrase and returns a warning most people never read. Our first pass, untagged, returned 1,768 results, which looked like a thriving literature and was actually the search engine tearing our query into pieces and matching them separately. Anyone who tells you they searched that phrase and found nothing did not search that phrase.

The workable version uses field tags and a proximity operator. Here is what nine of those returned.

The 30 hits and the 6 hits are the interesting ones, because they are not empty and they are still not about this. Open them and they are surveys of which allergens turn up in which product categories, reviews of a single preservative, analytical method papers. The phrase matched because scientists write “a number of ingredients” the way everyone else does.

So the papers that count things are counting allergens, not ingredients. That distinction is the whole article.

The two hits with the phrase in the title are both about groceries. One of them, published in early 2026, is called “Ingredient count is not a reliable indicator of nutritional quality.” We could not get past the publisher's paywall to read it, so we are citing a title and nothing more, and we would rather say that than imply we read it.

What the regulators count instead

If ingredient count mattered, the people who write safety law would have built something around it. They built the opposite.

The European Commission's guidance on how to assemble a cosmetic product safety report, Implementing Decision 2013/674/EU, is unusually plain about the unit of analysis. The aim, it says, is to describe the toxicological hazard of each of the substances in the finished product, determine the potential exposure, and draw up a risk characterisation. A few lines later: the assessment of exposure to each of the substances is necessary in order to assess the risk associated with each individual substance. Then the threshold that actually governs whether something ships: a margin of safety of at least 100.

A hundred-ingredient product needs a hundred of those calculations. A one-ingredient product needs one. Neither is safer for the arithmetic being shorter.

The SCCS puts it in a single sentence in the twelfth revision of its Notes of Guidance: the safety of cosmetic products is based on the safety of the ingredients. Plural, and one at a time.

The American side is the sharpest illustration. The Cosmetic Ingredient Review sets its annual priority list, under section 30(a) of its procedures, on the basis of frequency of use, which it defines as the number of different products in which an ingredient is used. That is a count of ingredients, and it runs in exactly the opposite direction to the one shoppers make: it asks how many products contain this substance, not how many substances this product contains.

The number that does predict something

Something in this space is measurable, and it is not the length of the list.

The SCCS opinion on fragrance allergens identifies 82 substances as established human contact allergens, split between 54 single chemicals and 28 natural extracts. And it attaches a number to exposure rather than to inventory: a general level of exposure up to 0.8 µg/cm², which works out around 0.01% in a cosmetic product, is tolerated by most consumers. That is a concentration threshold. You cannot compute it from a count.

A 2025 review in Current Allergy and Asthma Reports lines up the actual predictors from the clinic. Roughly 30% of patients presenting with facial dermatitis have it from cosmetics. Carmine came back positive in 3.1% of 4,240 patients tested; carvone in 0.8%. Leave-on products carry more sensitisation risk than rinse-off ones, which is a statement about contact time. Fragrance, preservatives, hair dyes, acrylates and UV filters are named as the recurring culprits.

Read the list of risk factors in that review and it is chemical identity, concentration, product format, how often you apply it, and where. Ingredient count is not on it. It does not appear as a rejected hypothesis either. Nobody thought to test it.

What our screening found

The practical takeaway

Stop counting lines. Start scanning for names.

The things that have actual evidence behind them are specific: the 82 substances on the EU's established-allergen list, the fragrance declarations at the bottom of the panel, whether the product stays on your skin or goes down the drain, and how often you reapply it. A 40-ingredient moisturiser with none of those is a calmer proposition than a 12-ingredient one built around an essential oil.

If you have reacted to something before, the short list genuinely helps — not because it is safer, but because it is easier to audit. Twelve lines you can read in the shop. Eighty-seven you cannot.

And treat “minimal” as a design philosophy rather than a safety claim. It is a perfectly good reason to like a product. It is not a finding.

We have written about a related counting problem in the one product in our catalogue that says preservative-free, about the words that get printed on the front in clean, vegan and non-comedogenic, and about how the scoring works at all in how we screen K-beauty ingredients. The method itself, including the terms it is missing, is on our awards page.