Data investigations
Long-form analyses from 18.1M Amazon reviews. Each investigation combines several findings into one question, states its sample, and separates what the data shows from what it cannot.
What an investigation is
An investigation combines several findings into one question. It is longer than a single finding and shorter than a report, and it is not a product review or a ranking. Every number is a share of sampled reviews with its sample size attached; every page has a what this does not tell you section.
Why this format
A single finding can tell you that a complaint is concentrated. It cannot tell you what that means for a buyer. An investigation connects the concentration to the price band, the scenario, the material or the claim on the listing β and then stops where the data stops.
What we will not do
We will not turn a 0.3% mention rate into “8Γ more likely to fail”. We will not compare a coated pan with bare cast iron and call it a coating test. And we will not publish a page without the sample size.
Start with the cookware-set and Dutch-oven investigations; they are the clearest examples of the method.
The 'nonstick' label predicts complaints, not performance
Products advertising 'nonstick' draw coating-peeling complaints in 0.16% of reviews vs 0.03% without the claim (5.0Γ), and nonstick-failure complaints at 4.2Γ. The label describes a surface, not a lifespan.
InvestigationScenario mismatch: bread baking, deep frying and camping
Products matched to bread baking report uneven heating 3.9Γ the catalog rate; deep-frying products report 'too heavy' 4.5Γ; camping products 2.4Γ. The scenario you buy for changes the complaint.
InvestigationRating drift: the honeymoon ends β but not at the same rate
838 ceramic products lose 0.26 stars from early to late reviews. Stainless steel falls 0.31 stars; porcelain 0.10. The drop is a signal about disappointment, not a measured failure time.
InvestigationPrice inversion: when paying more makes complaints worse
Premium Dutch ovens report 'too heavy' 8.4Γ more than budget ones; budget cookware sets report loose handles 3.0Γ and misleading sizes 2.6Γ more. Price changes which complaints you get.
InvestigationPFAS-free is a materials claim, not a lab test
Tagine 100%, double boiler 97.5%, butter warmer 94.6% PFAS-free in our classification. That means the material/spec rules found no PFAS coating β not that a lab tested the pan.
InvestigationOur durability score has a blind spot: warping and nonstick failure
Cookware-set products in the top third of our durability score report warping 1.5Γ and nonstick failure 1.46Γ more than the bottom third. Here is what the score measures β and what it misses.
InvestigationHow cookware dies: 1,186,236 products, a 2-year median, and a volume effect
Across 1,186,236 products, the median review lifespan is 2 years and 46.9% are discontinued. The strongest predictor of ending is not type β it is how many reviews a product already has.
InvestigationFrying pans: the coating fails before the pan does
In 14,517 frying-pan reviews across 171 products, food sticking is 8.8Γ and nonstick failure 8.5Γ the catalog rate. The pan body is rarely the complaint.
InvestigationDutch ovens: 'too heavy' is 11.6Γ the catalog rate β and rust is 8.4Γ
53,969 Dutch oven reviews across 673 products. The category's defining complaint is mass; the second is rust. Premium ovens report 'too heavy' 8.4Γ more than budget ones.
InvestigationCookware sets: 229,181 reviews show what fails first
In 2,108 cookware sets, food sticking and nonstick failure appear 7.8Γ and 5.3Γ the catalog rate. The first failures are surface and fit β not the body of the pan.
ReportCookware Durability Report 2026: what 18.1M reviews say about how pans age
A data-only report on coating failure, price inversion, lifecycle and the blind spots in our own durability score. No product was purchased or physically tested.
InvestigationBrand portfolios diverge: OXO, Norpro, Hamilton Beach and Cuisinart
OXO has fewer products than Cuisinart but a longer median lifespan and a higher share still receiving reviews. Brand size is not brand longevity β and neither is a single reliability score.
Investigation'Oven safe' is a temperature claim, not a warping guarantee
Cookware advertised as oven safe reports warping in 0.21% of reviews vs 0.05% without the claim (4.3Γ). The claim covers a temperature rating; warping is about thermal shock, support and pan geometry.
Every figure is a share of sampled reviews from the 2023 Amazon Reviews snapshot, not a lab-measured failure rate. Read the methodology, scoring standards and editorial policy.