Methodology: wattage pairs, fuel support and listing provenance

Ranking fuel-burning generators honestly starts with an admission: we work from listing documentation, not from a test bench. Nothing below is a measurement of ours. What we promise instead is that factual inputs have recorded listing or exact-model manufacturer provenance, and every rating starts from the formulas below.

Extraction before scoring

For each generator we scan the listing title, then the feature bullets, for running (rated) watts and peak (starting/surge) watts stated with an explicit wattage term. A bare wattage without an explicit running or peak role is not treated as evidence of either. Exact-model manufacturer documentation can supply a missing role or fuel-specific value; we retain its source. Fuel support — gasoline, propane, natural gas — is taken from the listing text alone. Each extracted attribute stores its source location, and a listing that never states both watt figures explicitly is scored on documentation completeness instead of spec fit; we do not fill gaps by analogy with similar models.

The three formulas

Documented spec fit = 100 × [0.45 × min(running watts / 9,500 W, 1) + 0.35 × min(peak watts / 12,500 W, 1) + 0.20 × min(supported fuel count / 3, 1)], rounded to a whole number. Running and peak watts require explicit roles and units in the exact listing or linked documentation for the exact manufacturer model; manufacturer values here are for gasoline. Bare marketing wattage is not treated as peak output; contradictory values are omitted, not swapped; the fuel count reflects gasoline, propane and natural gas support the listing itself claims. This index summarizes listing documentation for outage sizing — it is not a load test, a noise measurement or a durability judgment.

Documentation completeness = 100 × [0.5 × min(feature bullets / 8, 1) + 0.3 × min(listing images / 5, 1) + 0.2 × (brand stated ? 1 : 0)], rounded to a whole number, computed from the stored Amazon listing record when the listing text does not state both running and peak watts explicitly. It measures how completely a seller documents a listing — nothing about how the generator performs.

For the generator review blend, the captured Amazon average and review count enter this equation: 100 × [0.45 × (Amazon average / 5) × min(review count / 50, 1) + 0.35 × min(log10(1 + review count) / log10(1 + 5000), 1) + 0.20 × documentation score / 100]. The weights remain 45% for the confidence-adjusted average, 35% for logarithmic review volume and 20% for the documented component. Confidence reaches its cap at 50 reviews; volume credit reaches its cap at 5,000. Neither term establishes that reviews describe the same fuel, load or starting conditions a household will use. A gasoline running-watt figure must not be treated as propane output just because the listing supports both fuels. No verified review evidence means documentation-only scoring in the catalog, not eligibility for a comparator; each generator pool requires at least five review-backed candidates. This is an editorial evidence index, not a generator load test or a measurement of starting a particular appliance.

From the numbers to one rating

The formulas give us a starting point on a 100-point scale. The rating you see, out of 10, is the one we stand behind. It weighs what buyers say about the exact listing, how many of them have said it, the specs and the warranty the listing states, and our editors' judgement of how the generators compare with each other.

A rating doesn't move just because a stronger generator lands on another list; we don't grade on a curve. Identical results stay tied, and the listing slug settles their order without claiming one is electrically better. Read the running watts, the starting demand and the supported fuels next to the rating, because a spot on a list is not a wiring or load-compatibility decision.

Boundaries we hold

Standby generators requiring professional installation are excluded from the catalog entirely. No comparison on this site crosses the combustion/battery boundary — the two classes answer different questions, as our guide article explains. Prices and stock render from separately fetched offer data and are omitted when that snapshot is expired. Amazon photos have their own validity deadline. Review counts are sourced from captured exact-ASIN public-page evidence; they are not live customer totals. Noise, vibration, cold-start behavior and long-term durability are real buying factors we simply cannot score from listings — which is why our editorial ratings are not physical test results or Amazon customer ratings.

Corrections

Spotted a wattage we misread or a fuel claim that drifted? Write to [email protected] with the page URL and the listing text; provenance records make these checks fast. The affiliate disclosure covers the commercial relationship.

Spec-based ratings

Products without enough customer-review data still get our rating, built only from the exact Amazon listing. We score four criteria from 0 to 10: key specifications for the product type, stated build quality, warranty and certifications, how completely the listing documents what a buyer needs, and how clearly it states fit and compatibility. The rating runs from 4.5 to 8.5: 4.5 + the average of the four × 0.4. A thin listing lowers the rating but never below 4.5, because a short listing is not proof of a poor product; the 8.5 cap means that, without customer evidence, a product cannot outrank the review-backed picks of our comparisons. These are editorial judgments from documented facts, not from our own use, and they are labelled "Spec-based" wherever they appear.