How Our Safety Scoring Works

From data to decisions

Every ingredient in our database goes through a rigorous 5-step scoring pipeline. The goal: translate complex scientific data into a simple, trustworthy safety score that any parent can understand.

Here's exactly how it works.

Step 1: EU CosIng Import

We start with the EU Cosmetic Ingredient Database (CosIng) — the official registry maintained by the European Commission. This gives us the definitive list of ~30,000 cosmetic ingredients with their regulatory status.

Europe has the strictest cosmetic regulations in the world. The EU has banned over 1,300 ingredients, compared to just 11 in the US. Starting with EU data means we're building on the most protective regulatory framework.

Step 2: ECHA Hazard Classification

Next, we cross-reference each ingredient against the European Chemicals Agency (ECHA) database. ECHA classifies chemicals using standardised hazard statements:

• H360 (May damage fertility or the unborn child) → Automatically flagged as High Risk
• H361 (Suspected of damaging fertility or the unborn child) → Flagged as Medium Risk
• H362 (May cause harm to breastfed children) → Flagged as Medium Risk

These regulatory flags are treated as hard constraints — our scoring algorithm cannot downgrade an ingredient that has an official hazard classification.

Step 3: Scientific Literature Review

For each ingredient, we search PubMed — the world's largest biomedical research database — for pregnancy-relevant studies. We look for:

• Reproductive toxicity studies
• Teratogenicity (birth defect) research
• Prenatal exposure studies
• Breastfeeding transfer studies

The number and quality of relevant studies affects the confidence level of our score.

Step 4: Three-Axis Scoring

Every ingredient receives three sub-scores (0-3 scale):

Hazard Score (H)

How intrinsically dangerous is this substance?

• 0 = No known hazards
• 1 = Minimal concern
• 2 = Moderate concern (some evidence of harm at high doses)
• 3 = Known hazard (regulatory flags, strong evidence)

Mechanism Score (M)

Does it have a biological pathway to cause harm during pregnancy?

• 0 = No known mechanism
• 1 = Theoretical concern only
• 2 = Demonstrated mechanism in animal models
• 3 = Demonstrated mechanism in humans

Exposure Score (E)

How much actually reaches the body from normal cosmetic use?

• 0 = Negligible absorption
• 1 = Low absorption (most cosmetic use)
• 2 = Moderate absorption (leave-on products)
• 3 = High absorption (>20% systemic availability)

Step 5: Final Score Calculation

The three sub-scores combine into a final Safety Score (0-100) using our severity-blending algorithm:

1. Band assignment first: The highest sub-score determines the risk band

2. Severity blending: Final score = 0.7 × peak severity + 0.3 × average severity

3. Hard constraints: ECHA H360/H361 flags override any calculated score

Risk Bands

Why band-first logic?

A single high hazard score should place an ingredient in a dangerous band, even if the other scores are low. For example, an ingredient with a hazard score of 3 (known teratogen) but low exposure (0) is still classified as high risk — because if exposure conditions change (different product type, damaged skin), the hazard remains.

This is the precautionary principle in action.

Product-level scoring

When you scan or analyze a full product, we:

1. Score each ingredient individually

2. Weight by position (ingredients listed first are present in higher concentrations)

3. Flag the worst offenders prominently

4. Give an overall product safety score

The product score is only as good as its worst ingredient — a single high-risk ingredient can significantly lower the overall score.

Transparency and limitations

We believe in showing our work:

• Every ingredient page shows the source data and sub-scores
• Our methodology page explains the full pipeline
• We link to PubMed studies where available

Important limitations:

• Concentration matters — an ingredient at 0.01% behaves very differently than at 5%
• Individual health conditions may change risk profiles
• New research can change risk assessments
• This tool does not replace medical advice

Our methodology is continuously refined as new research emerges. We update scores when significant new evidence is published.