Watt & Warden
How We Work This Out
Everyone in this category says they tested twenty units. We have not tested any, and we say so. Here is what we did instead, and where it is weaker than a publisher with a test bench.
The position, plainly
Watt & Warden runs no test bench. Nobody here has plugged in a power station, put a generator through a winter, or measured an inverter's real efficiency under load. We own none of the equipment we write about.
We could say otherwise — plenty of sites in this category do, and the claim is essentially unfalsifiable from the outside. We are not going to, because a testing claim we cannot substantiate is exactly the thing a competitor or a Google quality rater can most easily disprove, and because it would be a lie.
What we actually do
The work is real and it is a different kind of work. In order:
- Compile published specifications from manufacturer product pages and spec sheets, and arrange them so figures line up and can be compared honestly.
- Compute the numbers manufacturers leave out — runtime against a real appliance load, cost per hour of running, cost per usable kilowatt-hour across a rated cycle life.
- Cite every input, with its date, so a reader can open the source and check it.
- Publish the arithmetic itself, not just the result, so anyone can substitute their own numbers and get a different answer.
- State what is not published rather than estimating it. “Not published by the manufacturer” is an acceptable cell value on this site and it appears in several tables.
The assumptions we use, stated once
Every calculation on this site rests on a small number of assumptions. They are listed here so you can find them in one place and disagree with them.
| Assumption | Value we use | Why |
|---|---|---|
| Inverter efficiency | 85% | Conservative for a good inverter at moderate load, generous for a cheap one near its limit. Manufacturers rarely publish a curve. |
| Refrigerator average draw | 150 W | A modern fridge with the door shut, averaged across compressor cycling. Your EnergyGuide label gives a better figure. |
| Generator fuel consumption | 0.75 gal/hr at half load | Planning figure for a mid-sized machine, used only where the manufacturer publishes no comparable number. |
| Gasoline price | $4.071/gal | EIA Weekly Retail Gasoline Prices, week ending August 31, 2026 |
| Propane price | $2.674/gal | EIA Weekly Residential Propane Prices, week ending March 30, 2026 |
| Electricity price | 18.34 ¢/kWh | EIA Electric Power Monthly, U.S. residential average, June 2026 |
| Solar harvest | 400–500 Wh per 100 W panel per good summer day | Roughly 50–80% of rated output across daylight, allowing for temperature, angle and controller losses. Given as a band because the real figure varies too much for a point value. |
Change any of these and every dependent figure moves proportionally. That is the point of stating them.
Where our method is genuinely weaker
This section is not modesty. It is the part that makes the rest believable, and a reader who needs what is in this list should go elsewhere for it.
- Build quality. We cannot tell you whether a hinge feels cheap, whether a handle flexes, or whether a panel's legs survive a season.
- Real-world noise. We quote published dB(A) figures, measured under the manufacturer's conditions. We have not stood next to any of these machines.
- Long-term reliability. Whether a unit still works after three years in a garage is a question about units in the field, and we have no field data.
- Measured efficiency. We use 85% as a working figure. The real number for a specific unit at a specific load requires a bench.
- Customer service and warranty experience. We can read the warranty terms. We cannot tell you how a claim actually goes.
If build quality and long-term reliability are your deciding factors, a publisher with a genuine testing program will serve you better than we can. We would rather say that than pretend a specification table answers it.
Where our method is stronger
The trade is real in both directions. A test bench cannot tell you what an hour of generator costs at today's fuel price, or how many watt-hours a battery delivers per dollar across its rated cycle life, or what capacity a three-day outage requires. Those are arithmetic, they are computable from published data, and they are what this site is for.
They are also reproducible. Every figure we compute is shown with its inputs, so you can redo it with your own fuel price, your own fridge and your own outage history — and get a better answer than ours.
The honesty stat block
- 0
- Units we claim to have tested
- 0
- Sponsored placements accepted
- 0
- Free products accepted
Three categories where zero is the good number. We are not counting things we lack — we are counting claims we have declined to make.
How we pick products
Products enter a roundup by meeting the requirement the page is written for, not by being popular. The sequence:
- Establish the load. What has to run, and what starting surge it demands, from published wattage references.
- Eliminate on inverter or output. Anything that cannot supply the surge is out, regardless of capacity or price.
- Compare on published specification — capacity, cycle life, weight, charge rate, fuel consumption — with the endpoints of each claim checked, because a cycle rating quoted to 70% is not the same claim as one quoted to 80%.
- Name who should skip it. Every pick carries a stated disqualifying condition. It is a required field in our content system, not an editorial choice made page by page — a roundup where every product is great is an advertisement.
A product is excluded when its specifications cannot be sourced, when it cannot meet the page's requirement, or when a cheaper item on the same page does the same job.
Prices
Prices on this site render from a live data layer or not at all. Where a live price cannot be verified, the button reads “Check price” rather than showing a number.
There is no code path on this site that can display a remembered or estimated price, because there is no price source other than that layer. That is a structural guarantee rather than a promise.
Who writes this
Sawyer V. — an enthusiast who is genuinely into this. No lab coat, no test bench, no van full of generators. What there is: manufacturer manuals read end to end, published specifications compiled into tables you can actually compare, and the electrical arithmetic done properly and shown in full. If a number is on this site, you can check it. More about the site.
Corrections
If a figure here is wrong, we want to know, and we would rather hear it from you than leave it standing. Email through the contact page with the page and the figure. We correct errors as soon as we can verify them, note the correction on the page, and update the last-modified date — which on this site reflects real content changes and nothing else.
There are currently 46 content pages on the site. Every one carries its sources.