How we pick

Why we do not test

Because we do not have a lab, we do not own these machines, and claiming otherwise would be a lie that a competitor or a search quality rater could falsify in an afternoon. The ranking field here leads with genuine testing operations — publishers who buy every unit, run studios, and measure suction with manometers. We cannot imitate that honestly and we are not going to try.

So we compete on the axis they leave open. Not one publisher we researched answers the question a value buyer is actually asking: if the flagship costs several times as much, what specifically is the difference? That question is answerable from published specifications and arithmetic, it is reproducible by anyone who wants to check us, and it is uncontested.

Step 1 — Define the buyer and the ceiling

Every ranking states who it is for and what it must cost. Price ceilings are hard: a pick on a page titled "under $200" that costs $215 is removed, not rationalized. Because prices move daily and this site refreshes them daily, our price-ceiling pages check their own picks against their own headline and flag any that have crossed it — rather than leaving the title quietly wrong for a month.

Step 2 — Compile published specifications from primary sources

For every machine we pull the manufacturer's own published figures: rated runtime and the conditions it was measured in, motor power, battery voltage, bin or tank capacity, weight, filtration class and any test standard named, brushroll type, noise rating, warranty length and the consumable list.

Every figure is attributed. Where a manufacturer publishes no figure, the specification table says "Not published" — never an estimate, and never a number carried over from a similar model in the same range. A marked gap is a fact about the product; a quietly removed row is a fact you were not given.

Step 3 — Compute the running cost

Consumables are where budget machines quietly become expensive, and no competitor we researched publishes this. For each machine we compute what it costs to run for three years:

Where an interval is not published, we write that it is not published rather than modeling one. That is why several running-cost tables on this site have rows reading "unknown": those gaps are the honest output, and filling them would be the exact fabrication this method exists to avoid.

Step 4 — Compute the value delta against the flagship

The core of what we do. For each category we tabulate the budget pick against the category flagship on every published specification, then state the delta in plain arithmetic. Not "the Dyson is better" — it often is — but how much better, on which published number, and whether that difference does anything in your house.

Sometimes that arithmetic favors the flagship, and we say so. Sometimes it does not: the Dyson V8 and the Shark IX141 publish the same 40-minute rated runtime, and the Shark's bin is larger. Both of those are the manufacturers' own numbers.

Step 5 — Name what we cannot know

Every page carries an explicit limits section. Things published specifications cannot tell you, stated plainly rather than glossed over:

• Real-world pickup. No two brands publish a common test standard, and pascals, air watts and motor watts are three different quantities.
• Durability. Nobody publishes an expected lifespan, a cycle rating or a failure rate for anything on this site.
• How it feels. Handle balance, maneuverability, whether a trigger gets tiring — none of it is on a spec sheet and we have not held these machines.
• App longevity on connected products. The largest unpriced risk in the robot category, and no manufacturer commits to a support period.
• Noise in your house. Ratings are comparable between machines and tell you nothing about your specific hallway.

Step 6 — Name at least one machine to skip

Every ranking on this site names a machine the reader should avoid and the specification that disqualifies it. A roundup where every pick is a winner is a sales page, and naming a model not to buy is the cheapest credibility available to a publisher that cannot test anything.

The scoring rubric

Each machine is scored 0-10 on five category-appropriate metrics, and the overall figure is their average. The scores are editorial judgments about published specifications. They are not measurements, they are not test results, and they are deliberately never emitted as structured data — marking an editorial score up as a product review rating would contradict everything on this page.

How prices work here

Prices come from a live lookup against the retailer, refreshed daily, and they carry the date they were verified. A price older than 48 hours is not shown at all — the buy button falls back to "Check price on Amazon", which is honest, rather than displaying a number we cannot stand behind. There is no fallback to a remembered figure anywhere on this site, because there is no code path that could produce one.

How we make money, and how that is contained

We earn a commission when you buy through a link here, at no extra cost to you. Commission scales with price, which means recommending cheaper machines earns us less — that is the honest shape of the conflict on a site built around value picks.

What contains it: rankings are set from published specifications before any link is built, every ranking names a machine to skip, and we recommend renting or buying nothing at all where that is the better answer. The editorial policy states the rule in full, and the affiliate disclosure covers the commercial relationships.

How to tell us we got something wrong

Email us. If a figure on this site does not match what the manufacturer publishes, we want to know, and we will correct it and log the correction publicly on the editorial policy page. Everything here is traceable to a document, which means every error is findable — that is the point of working this way.