How Hange works

Last updated October 2026

Hange asks you a few questions and recommends laptops. This page explains how it decides, because a recommendation you cannot check is not worth much.

The ranking is arithmetic, not an opinion

The same answers always produce the same laptops in the same order. Nothing is asked to judge which laptop is best at the moment you press submit — the ordering is computed by code that does the same thing every time, and it is finished before any text is written about it.

The sentences on the results page are written by the same code, from the same numbers: a card that says a laptop is “1.8 lb lighter” is reading the two weights. That is deliberate: it means “why is this first?” has a real answer rather than a plausible one.

We score from what nearly every laptop has

Detailed reviewer ratings — measured screen brightness, build quality scores — exist for a small slice of the market. Building a recommender around them would mean scoring that slice properly and making up numbers for everything else.

So we score from the chips instead. A laptop’s processor and graphics are listed almost everywhere, and an RTX 4060 performs like an RTX 4060 whatever chassis it is bolted into. We keep a hand-curated table of 367 processors and graphics chips, most of them with published benchmark figures, and map each laptop onto it. Where richer per-machine data exists, it raises our confidence in that row — it is never a prerequisite for appearing at all.

Six things we measure

Every laptop gets six scores, shown as bars out of 10, computed once and identical for everyone:

• Processor — multi-threaded and single-threaded performance, kept separate — a 24-core chip is not better for writing documents
• Graphics — 3D performance, with integrated graphics scored honestly rather than as zero
• Screen — resolution density, refresh rate and panel type
• Build — estimated from the kind of laptop, its brand and its price, or a reviewer's rating where one exists; marked “est.” when estimated
• Portability — weight and thickness
• Battery — a runtime timed in an independent review where one exists, otherwise estimated from battery capacity against how much power the machine draws

Your answers change how those six are weighted. They never change the underlying numbers, which is why two people with different priorities see the same laptop scored identically and ranked differently.

What we will not do

• Invent a number — a visible gap is better than a confident wrong figure, because you can see the gap
• Treat missing data as zero — a laptop we know less about is not a worse laptop, and it is not a better one either
• Pass an estimate off as a measurement — a battery life no one timed, a build score from brand and price, a weight we don't hold, a chip with no published benchmark: each is marked “est.” where it appears. Our battery estimates are typically within about a sixth of a timed runtime, and where a runtime was timed we use it instead
• Rank Chromebooks against Windows laptops — the benchmarks we use do not run comparably on ChromeOS, and scoring a Chromebook on whether it can export 4K video answers a question nobody asks of one

Every answer you give produces a visible outcome

Each recommendation comes with a list of what you told us and what happened to it: which answers narrowed the field and by how much, which ones changed the weighting, and which ones we could not act on.

That last category is the point. If we know whether a laptop has a touchscreen for only a fraction of the catalog, filtering on it would either throw away hundreds of machines nobody checked or do nothing at all. Neither is honest, so we say so — on the option itself, before you pick it, not buried in the results.

Where our data comes from

Three kinds of source, each used for what it actually measures:

• Benchmark databases — processor scores are PassMark results (its multi-core CPU Mark and its single-thread rating), and graphics scores are 3DMark Time Spy graphics results, averaged per chip across published runs. Every laptop is matched by its chips against our table of 367 processors and graphics chips. A chip too new or too rare to have a reliable result is placed against comparable chips instead, and the card marks its Power “est.”: 123 of the 367, found in 67 of the 651 laptops we rank
• Independent reviews with lab tests — full reviews that list every specification and time the battery in a standard Wi-Fi browsing test. 204 of the 651 laptops we rank have a runtime timed this way. For the rest we estimate it from battery capacity and power draw, with a model checked against 530 timed runtimes, and the card marks it “est.”
• Manufacturer and retailer listings — screen, weight, memory, storage and US prices. Where a listing is wrong or blank, we correct it by hand and record where each correction came from

Hange isn’t affiliated with these sources: not with PassMark Software, not with UL, which makes 3DMark, and not with any reviewer. None of them endorses Hange.

We update the catalog by hand, not automatically: a new one replaces it once newer laptops have been gathered and checked. The current catalog went live on 21 October 2026.

Rankings never depend on money. No brand, retailer or anyone else can pay to be ranked or shown, and the commission we may earn when you buy through a link plays no part: the order is settled before any shop link is attached. The only money in the ranking is each laptop’s price, against your budget.

One more rule for prices. We rank a laptop only when we hold a US dollar price for it. Some laptops we hold only at a price in another currency, and converting that into dollars would mean guessing at tax and exchange rates, so we don’t rank them until we have checked their US price.

How it is paid for

Hange is built and run by one person. When you buy a laptop through a link here, we may earn a commission from the retailer at no additional cost to you.

Commission does not affect ranking, and it cannot: the ordering is computed from the six scores and your answers before any retailer link is attached to a result. See the affiliate disclosure for the full statement.