How we rank devices (and why it's not by score)

A few people have asked when we're going to add a numerical score to device pages. "Like a 7.8 out of 10 — it's useful for comparing." We're not going to do that, and this post explains why.

The problem with numeric scores

A score collapses several incomparable axes into one number. A camera is good or bad. A battery is good or bad. A processor is good or bad. To produce one summary score you have to pick weights — say, 25% camera, 20% battery, 20% chipset, 15% screen, 10% build, 10% software. You apply those weights to every phone, and you get a number that can be ranked.

The weights are wrong for almost everyone.

If you take 200 photos a week, the camera should be 60% of the score. If you commute six hours a day, battery should be 40%. If you play games on your phone, chipset should be 35%. The 25/20/20/15/10/10 weighting is wrong for all three of those people, and they're all real.

What numeric scores actually measure is "how closely does this phone match the spec preferences of the publication's editor." That's useful when the editor is honest about it — Wirecutter basically does this and is upfront. It's misleading when the score is presented as objective.

The other problem: scoring incentivizes spec inflation

When publications score phones, manufacturers optimize for the spec sheet rather than the user experience. We saw this in the megapixel race (200MP cameras that take worse photos than 12MP ones because the sensor is too small to capture good light per pixel). We see it now in the RAM-inflation race ("8GB physical + 8GB virtual = 16GB!" marketing where the virtual RAM is just slow swap space).

If we publish a score that rewards "more RAM," we're contributing to the problem. If we publish data and let you make your own call, we're not.

What we do instead

Three things:

1. Computed insights grounded in peer data

Every device page has a "Quick verdict" and "Where this phone stands today" section. These are generated by a rule engine that compares the device to its release-year cohort. Examples:

• "Above-average battery — the 5,000 mAh battery is in the top quartile of phones released in 2024."
• "Limited memory for its release year — 4GB RAM is in the bottom quartile of phones released in 2024, which can limit multitasking and the longevity of OS updates."
• "Released 5 years ago. Without a recent OS update, this device is unlikely to receive further security patches from Google."

These statements are deterministic. They're computed from data, not opinion. They tell you something specific instead of compressing everything into a single number.

2. Comparison tables that surface deltas, not scores

The /compare page shows two phones side-by-side and highlights which has the better value in each spec category. The output is a verdict like "Pick the Pixel 6 for newer Android version, more RAM. Pick the Galaxy A51 for larger battery."

You'll never see the page say "Pixel 6 wins 7-3." It says which one wins on each axis and lets you weight them.

3. Hand-written buying guides

For specific use cases — best budget phone, best for battery, best Samsung under $500 — we publish hand-written guides where we explicitly state our weights and reasoning. The buyer guide for "best for battery life" weights battery 60%; the one for "best camera" weights the camera 60%. We tell you which guide is for which buyer.

What we won't ever publish

• An aggregate "Soft9 score."
• A "best phone of 2026" superlative without specifying for whom.
• Anything that suggests a phone is objectively bad. Phones are bad at things, and good at others. The Galaxy A02s is a bad camera phone and a fine "I want a phone that doesn't break" phone.

What we will publish

• Lists like "best for X under $Y in 2026" with explicit reasoning about who the list is for.
• Tradeoff analyses ("here's what you give up if you spend $300 instead of $500").
• "Don't buy this" warnings when a manufacturer's update commitment is bad enough that a phone shouldn't be bought even at zero dollars.

The third one matters. Most spec sites don't tell you when a phone is genuinely a bad idea, because saying that loses access to manufacturer review units. We don't have access to lose, so we say it.

— Marcus