How we score

One framework, applied the same way on every page, with the limits of our evidence stated plainly.

What WhichSetup is

People who already know they need a category — an espresso machine, a monitor arm, pantry bins — get stuck between four and eight products that all look the same. We rank a short list against one public framework, say who should skip each pick, and date every check. That’s the whole product.

The scorecard (100 points)

Every pick is scored before it appears on a page.

BlockPtsWhat a high score means
Category integrity25The category is one we can cover honestly and link precisely — an exact product variant, a real buying decision, not a filler category we’d have to pad. We do not compare payouts between products on a page.
Demand durability15Steady availability and review activity — not a discontinued or dying product
Listing quality15Genuine reviews, a coherent brand, no bot-farm smell
Variant & return risk15The exact model is identifiable; fit and sizing rules are clear
What we can honestly say15We have real commentary to add without pretending to own a lab
Claim safety15No health, medical or legal claims needed to make the pick work

Top pick: 70 or higher with no hard fail.Worth comparing: 55 or higher with no hard fail.Add-on only: a small item that earns its place in a setup — a scale, a brush, a set of clips — rather than a headline pick.

Hard fail gates

Any one of these disqualifies a product as a pick, whatever it scores elsewhere.

  1. A category we can’t cover without padding — groceries, consumables and gift cards are not buying decisions we can add anything to
  2. We can’t confirm which product the listing actually is
  3. The pick only works if we make a medical, legal or safety claim we can’t support
  4. Counterfeit or trademark chaos in the reviews
  5. Only a parent listing is linkable when the variants meaningfully differ

How money does and doesn’t influence this

We’re an Amazon Associate: we earn a commission when someone buys through our links, at no extra cost to the buyer.

That affects which categories we cover — we don’t open a category we can’t fund the upkeep on, because a page nobody maintains is worse than no page. It does not affect the order of picks on a page. Payout is not an input to any score, and we don’t compare what different products on the same page would earn us. A cheap trial machine routinely beats a flashier one here, because for an undecided buyer it’s the right call.

Synthesis, not lab testing

Our pages are research synthesis: manufacturer specifications, published expert consensus, patterns across owner reviews, and our own scoring and commentary.

We do not claim to have tested dozens of units in a lab. We don’t own one. If we ever run first-hand tests, we’ll label them with the method and the date. Treating a synthesis page as a lab report would be the easiest lie on the internet and it’s the one we’re least willing to tell.

We use AI tooling to help research and draft, working from public sources: manufacturer specs, retailer listings and published reviews. The site’s owner approves each guide before it publishes. Scores come from the framework above, not from lab testing.

When we re-check a page

What changedWhat we doWithin
A top pick goes out of stock or is replacedRe-target or rewrite the verdict72 hours
Price moves more than 15% on two or more picksUpdate the page and log it7 days
A page passes 90 daysFull re-verification14 days
A page passes 180 days without a re-checkRebuild it or take it down14 days

We don’t bump a page’s date without a substantive change. If the log says we checked it, we checked it.

What we refuse

Questions

Corrections, disputes and “you got this wrong” email:hello@whichsetup.com. See also ouraffiliate disclosure and who runs this.

Last reviewed: September 19, 2026