How the Specdex score works.
Specdex is an editorial layer on top of publicly available product information. Specs come from the manufacturer's own spec page where they publish one. The 0–100 score is a curator's weighted assessment — not a measurement, not a manufacturer claim. This page documents both halves and how to flag a mistake.
Operating principles
Specs from the manufacturer
Where a manufacturer publishes a value (chip, display, storage tiers, weight, IP rating), we use it verbatim from their official spec page. Where they don't publish (e.g. Apple does not publish RAM or battery mAh), we omit the field rather than extrapolate from rumours, leaks, or third-party teardowns.
Score is editorial
The 0–100 Specdex score is our own weighted assessment built from public specs, hands-on use, and reading every authoritative review we can find. It is not a manufacturer claim and it is not a measured lab result. Treat it as a curator's opinion, not a fact.
Independence
No paid placements affect rankings. Affiliate links, where present, are clearly disclosed and isolated from the score path. We don't accept money to move a product up the list.
Corrections welcome
If you spot a wrong spec, a missing variant, or a product that shouldn't be on the site at all, email the editorial team. Past corrections include removing a fabricated GPU SKU and renaming a misnamed iPhone variant — both surfaced by readers.
Per-category rubrics
Different categories prize different things. Below are the weights we apply when computing the overall Specdex score, plus the test that drives each sub-score.
Phones
total 100%Public Geekbench / chipset benchmarks + manufacturer-claimed performance gen-on-gen
Public 3DMark / mobile-GPU benchmarks + sustained-load behaviour from established review outlets
Manufacturer sensor specs + sample reviews from outlets we trust on imaging
Manufacturer-published video-playback hours + charging wattage + reviewer-measured screen-on time
Manufacturer-published peak brightness, refresh, resolution, panel type
Build quality, IP rating, ergonomics, software polish — informed by reviewer consensus
CPUs
total 100%Public single-thread benchmarks aggregated from multiple sources
Public multi-thread + sustained-load benchmarks
Reviewer game-benchmark composites at common resolutions
Public performance-per-watt comparisons under sustained load
GPUs
total 100%Public raster benchmarks at 1440p / 4K from multiple reviewer datasets
Public ray-tracing benchmarks + path-traced game results
Public Stable Diffusion / LLM inference numbers; manufacturer-published TOPS where applicable
Public performance-per-watt under sustained game load
Laptops
total 100%Public CPU benchmarks for the configured chip on-battery + plugged-in
Public sustained gaming and AI workloads at the chassis's thermal envelope
Manufacturer-published panel specs + reviewer brightness / colour measurements
Manufacturer-claimed hours + reviewer real-world tests at common brightness levels
Weight + dimensions from spec sheets, plus reviewer-noted thermal noise behaviour
TVs
total 100%Reviewer measurements: colour volume, contrast, peak HDR brightness, motion clarity
Reviewer assessments of Dolby Vision / HDR10+ tone-mapping accuracy
Reviewer-measured input lag, VRR, ALLM, refresh ceiling, near-black uniformity
OS responsiveness, app catalog, app crash rate from reviewer notes
Speakers, design, port count, longevity
Headphones
total 100%Reviewer measurements vs Harman 2018 target; aggregated outlet consensus
Reviewer real-room attenuation measurements 100 Hz – 8 kHz
Clamping force, weight distribution, hot-spot fatigue from long-term reviews
Manufacturer-claimed hours + reviewer ANC-on tests
Codec support, multipoint, software polish, mic quality
Where the data comes from
| Source | What it contributes |
|---|---|
| Manufacturer spec pages | Primary source for chip, display, storage, weight, IP rating, charging, battery hours |
| Manufacturer press releases | Authoritative for release date, official MSRP, regional availability |
| Established review outlets | Cross-checked for hands-on observations, photo samples, and real-world battery / sustained-load behaviour |
| Public benchmark databases | Aggregated CPU / GPU performance signals (Geekbench Browser, 3DMark, manufacturer-published numbers) — never a private rerun |
| Reader corrections | Any field a reader can show is wrong against a primary source gets fixed — see corrections log on changelog |
When entries change
- When the manufacturer changes a specIf the manufacturer publishes a corrected or updated spec on their official page, we update the field on the next catalog refresh.
- When a reader sends a correctionWe accept any correction backed by a primary source (manufacturer page, press release, regulatory filing). The fix lands as a public entry in the changelog, with the reporting source credited if requested.
- When the editor revisits the scoreEditorial 0–100 scores can shift when a meaningful new data point arrives — a competing product launches, a long-term review surfaces a flaw, a price drops. Each shift is dated in the changelog so readers can see the trail.
Frequently asked
Is Specdex a benchmark lab? +
No. Specdex is an editorial layer that summarises, weights, and ranks publicly available product information. We do not run our own benchmarks, calibrate displays in-house, or measure power draw on a bench. Earlier versions of this page implied a lab — that was wrong and we've removed those claims.
Why a 0-100 score? Isn't that arbitrary? +
A weighted composite is necessarily one number. We chose 0-100 because it matches the implicit anchoring of school grades — readers intuitively know that 95 is exceptional, 80 is good, 60 is mediocre. It's a curator's shortcut, not a measurement.
Where do the spec values come from? +
Wherever possible: the manufacturer's own published spec page. The "Manufacturer-sourced where verifiable" badge on each review page marks that section. If a manufacturer does not publish a value (e.g. Apple does not disclose iPhone RAM), we leave it out rather than guess.
How do I report a wrong spec or a fake product? +
Mail the editorial team with the slug and a link to a primary source that contradicts the page. We've accepted reader corrections that removed a fabricated GeForce RTX SKU and a Galaxy product that wasn't actually shipping yet.
Are review samples provided by manufacturers? +
No. We do not maintain a review programme with any manufacturer.
Can I see the machine-readable record? +
Yes. Each product's structured record is at /api/product/<slug>.json — slug, brand, category, score, sub-scores, public specs, release date.