What 1,126 unique formula-review URLs reveal
The chart shows lexical mention counts after deduplication and removal of one standalone applicator-kit listing. It does not measure sentiment, prevalence in the market, or what a representative shopper thinks.
Read this before the chart
The scraper intentionally combined recent-all reviews with rating-stratified helpful-review slices. Rows were then deduplicated by review URL, and one standalone tanning-mitt kit was excluded from formula findings. Theme tags are lexical matches, not sentiment labels. One review can mention several themes, so counts overlap.
The most frequent lexical tags
The source capture contained 1,684 rows and 1,175 unique review URLs. We removed 58 captured rows (49 unique URLs) from one standalone applicator-kit listing. The resulting formula subset contains 1,626 captured rows, 1,126 unique URLs, 1,125 with text, and 1,116 marked verified.
Sampling mix and rating distribution
After deduplication, 866 formula-review URLs (76.9%) appeared only in rating-filtered helpful slices, 92 (8.2%) appeared only in recent-all or validation slices, and 168 (14.9%) appeared in both. This three-way membership view avoids assigning an overlapping URL to whichever slice happened to be retained first. The deliberate stratification supports issue discovery, not natural rating prevalence.
| Rating | URLs | Share |
|---|---|---|
| 1 star | 198 | 17.6% |
| 2 star | 187 | 16.6% |
| 3 star | 207 | 18.4% |
| 4 star | 237 | 21.0% |
| 5 star | 297 | 26.4% |
What changed in our editorial model
Format comes before brand
Mousse, drops, lotion, mist, and express formulas ask for different technique. Our finder now resolves that choice before it recommends a product.
Fit beats a universal winner
Skin tone, face or body use, timing, and tolerance for scent or transfer can outweigh a two-point difference in the overall score.
Failure modes stay visible
Every review page now shows a skip condition and evidence status near the verdict instead of burying drawbacks below a promotional summary.
Evidence boundary
- This is: a stratified, structured analysis of public formula-review text and product surfaces.
- This is not: a clinical study, representative survey, natural rating distribution, sentiment model, sales ranking, or hands-on wear test.
- Unit of analysis: unique review URLs, not a verified count of unique people.
- Accessory exclusion: the standalone mitt-kit listing is reported above but excluded from formula theme counts.
- Retailer totals: displayed only as retailer context and never substituted for our analyzed corpus.
- Reproducibility: the public theme summary is available as JSON; the scoring method is on the methodology page.
Editorial implication
The best-X page should work like a decision system: answer the shopper's use case, show the comparison logic, expose the evidence, and state where the evidence stops. That is the July standard across Tan List.