The one thing to know first
PickPrism does not physically test products. We do not buy them, handle them, time them, or measure them. Every number on this site comes from a document someone else published, and every judgement is analysis built on top of those documents. We say this loudly because the internet is full of "hands-on reviews" that were never hands on anything, and the difference matters when you are about to spend money.
What we do instead is the part most buying guides skip: work out which specifications actually decide the outcome for a specific situation, then apply them consistently.
Four kinds of statement
Everything we publish falls into one of these, and we keep them separate:
- Verified specification
- A documented fact, traced to its source — a manufacturer specification sheet, a product manual, a standards or certification database, or a regulatory filing. If we cannot find the document, the number does not appear. Comparison tables show Not published rather than a plausible guess.
- Calculated metric
- Something we derived arithmetically from verified specifications — runtime per unit of weight, capacity relative to footprint, coverage per hour. The inputs are cited; the calculation is deterministic and identical across every product in the comparison.
- Editorial analysis
- Our reasoning about what those facts mean for a buyer: which trade-off matters more, who a product suits, when a cheaper option is enough. This is a judgement, and we label it as one.
- Human-tested observation
- Reserved for structured testing data from a named person who actually used the product. This currently appears nowhere on PickPrism. If it ever does, it will be explicitly marked as tested, with who tested it and how.
Where evidence comes from
We work down this order, and prefer the highest tier that answers the question:
- Manufacturer specification sheets and technical documentation
- Official product manuals and installation guides
- Standards bodies, certification programs, and government databases
- Established technical publications with a stated test methodology
- Major retailer listings, used only to confirm identity, availability, and configuration
- Credible independent sources, weighted by how clearly they describe their own method
Every article records the documents it relied on, when we retrieved them, and what each one supported. The Sources section at the bottom of each guide is that record, not decoration.
We do not build articles by summarising other buying guides. Reading competitors tells us what questions buyers ask; it does not tell us anything true about a product, so it never becomes evidence.
The PickPrism Research Score
Each pick carries a score out of ten. It is a weighted rollup of factors, and both the factors and the weights are printed on the page underneath it. The score is a summary of the analysis above it — it is not, and cannot be, a measurement of how a product performs in your home.
How it is built:
- Start from the category's factor library. Robot vacuums can be judged on suction, threshold clearance, mop lift, carpet detection, brush design, navigation, runtime, dock maintenance, filtration, dimensions and warranty, among others. The library is the full set of things that could matter.
- Choose and weight the factors this guide is about, before scoring anything. A guide about homes with both carpet and hardwood weighs threshold clearance and mop lift heavily and charge time barely at all; a guide about small apartments would weigh them differently. The weights are fixed before any product is scored, because weights chosen after seeing which product wins are not weights, they are a justification.
- Rate each factor from published specifications. Zero to ten, with the reason recorded next to it.
- Treat a missing figure as an evidence problem, not a performance one. Where a specification would decide the purchase and the manufacturer does not publish it, the score is reduced and the reason is printed: not that the product performs badly, but that a buyer cannot check it. Where the missing figure would not have changed the decision, the factor sits at neutral rather than being punished.
- Publish the whole breakdown so you can override our priorities with your own.
What a low score means when nothing is published
This is the part most likely to be misread, so it is worth stating flatly. When a guide scores a product down on a specification the manufacturer does not publish, that is not a claim that the product is bad at it. It is a statement that you cannot verify it before buying, and neither can we.
A robot vacuum that publishes no threshold-climbing height may climb thresholds perfectly well. What it does not do is let a buyer with a tall door saddle check before spending the money — and for that buyer, in that guide, the gap is a real cost. Where the same manufacturer does not publish a charge time, and the guide is not about charge times, nothing is deducted at all.
The score breakdown says which of the two happened, on every factor, in the article's own words.
What "best" means here
A guide may only call something the best overall if it screened a real field to get there. Where we screened the current models in a category and shortlisted the strongest, the guide says how many were screened and when. Where we compared a handful of well-known models without surveying the market, the award says exactly that — our top pick among the models compared — and the difference is deliberate.
We never print a number of candidates we did not actually check.
Why the same product gets different scores in different guides
Because "best" is meaningless without a situation. In a guide for small apartments, weight and storage footprint carry heavy weight and total runtime barely matters — you are never more than twenty feet from an outlet. In a guide for large multi-storey homes, runtime and battery configuration dominate and a few extra pounds is a fair trade. Same specifications, different question, different answer. A site that gives a product one universal score is answering a question nobody asked.
How picks get chosen
A product has to clear an evidence bar before it can be recommended at all: confirmed identity (brand, model, current configuration) and enough published specification detail to justify the recommendation on its merits. A product we cannot verify is left out. An incomplete guide is better than a confident one built on invented facts.
Awards are specific — best for small spaces, best lightweight option, longest published warranty — and we only hand one out when the analysis actually supports it. If two products are genuinely equivalent for a use case, we say so instead of manufacturing a winner.
Prices
We do not print prices, and we do not describe them either. Retail prices change daily and a stale number is worse than no number, so our buttons send you to Amazon, where the current price is.
That includes calling a product budget, mid-range or premium. We hold no price data — our product records carry specifications and identity, not cost — so any such label would be a guess dressed as a finding, which is the one thing this method exists to prevent. If we ever have a sourced basis for describing where a product sits in its market, we will say what the basis is before we use the words.
Keeping guides current
Each guide records when it was published, when it was last updated, and when its research was last verified. Those are three different dates and we do not touch any of them to fake freshness. On review we re-check specifications against their sources, remove discontinued products, repair links, and update the analysis if the conclusion has moved. If nothing has changed, the update date does not change either.
Where this method falls short
Being straight about the limits:
- Published specifications are marketing documents. Manufacturers choose favourable test conditions, and we cannot audit them.
- Specifications say nothing about long-term reliability, and we do not speculate about it.
- Some qualities — how a handle feels, how loud a motor is in a small kitchen — cannot be read off a datasheet. Where that is the deciding factor, we say the specification cannot answer it.
- Our weightings are judgements. Reasonable people would weight differently, which is why we show ours.
Corrections
If a specification on this site is wrong, we want to know. See corrections for how we handle them, and editorial policy for the rules we write under.