Methodology How MobileIQ works
Where every figure comes from, how the one number we compute is computed, and what we have decided not to publish.
Sources
Every figure on this site carries a chip naming the class of source it came from. The chip names the class; it does not grade the truth. A manufacturer's own specification is not a suspect document.
- Official
The manufacturer's published specification for its own hardware. If Apple says an iPhone weighs 199 g, that is what the iPhone weighs, and we record it as the specification rather than as a claim awaiting adjudication.
1836 of 2265 figures
- 3rd party
A named independent publication or database — cited on the page it appears on, with its own link and retrieval date. We use one where the manufacturer publishes nothing, not to second-guess a figure the manufacturer does publish: where a maker and a spec aggregator such as GSMArena print different figures, the maker's is the one shown and the other is printed beside it. Benchmark scores are the exception — the benchmark's publisher is the source.
429 of 2265 figures
- Measured
A named lab's own bench result — someone put the device on a rig and reported what it did. It is the one class that earns the accent, because it is the one class that is genuinely different from reading a spec sheet.
0 of 2265 figures — MobileIQ has no lab, so nothing here is measured by us and nothing is relabelled as if it were
Two rules hold across all three. Every figure links its source and prints the date we read it, because a source without a date is a rumour. And a figure we do not have renders as missing — named, in the space it would have occupied — never estimated, never interpolated from a sibling model, never carried over from last year's phone.
The Endurance index
It measures how long a phone runs, how long its battery survives, how fast it refills, how fixable it is, and what it weighs — and nothing else. Not cameras, not speed, not screens: a £430 phone can beat a £1,249 one here, and when it does, that is the point.
One number, 5 attributes, weights published below. It isnot an overall score, and we will never publish one — an overall score is a ranking with no stated question behind it, which is how a phone site ends up ranking the phone it happened to like.
Watches have no index. The Endurance index covers phones only. A watch index was tried and withdrawn: normalised across the watch catalogue, the multi-week battery claims of watches we could not score set the range, so it ranked the shortest-lived watches first. Watch pages and the watch catalogue show the figures and their sources, and rank nothing.
| Attribute | Weight |
|---|---|
| Battery per charge | 35% |
| Battery lifespan | 20% |
| Weight | 15% |
| Fast charging | 15% |
| Repairability (EU class) | 15% |
| Total | 100% |
- Min–max normalised across the current catalogue. For each attribute the best figure among the phones we have sourced scores 1 and the worst scores 0. So 100 means "best of the 55 phones we have sourced", not "perfect", and every score can move when a phone is added.
- A device missing more than 40% of the weighted attributes is not scored at all. It gets no number rather than a low one. A thinly sourced phone is not a bad phone, and scoring it as one would print our data gaps as its faults. Below that threshold a device is scored out of the weight it actually has. 34 of 55 phones currently carry an index.
- Repairability is scored absolutely, not against the field. The EU class is a regulatory grade — a C is a C whoever it stands next to — so it uses a fixed scale A 100 · B 75 · C 50 · D 25 · E 0 instead of being stretched across the catalogue. Min–maxing it would promote the best of a bad field to full marks.
- Beta The weights are a judgement, not a discovery, and they will move. When they do, every score on the site moves with them and this table changes on the same commit.
Perceptual verdicts
A comparison page does not just print two numbers; it says whether the gap between them is one you would notice. These are the thresholds it judges against.
| Attribute | Noticeable | Marginal | Basis |
|---|---|---|---|
| Weight | 30 g apart | 15 g apart | — |
| Display size | 0.5 in apart | 0.2 in apart | 0.2 in is inside bezel-and-corner variation; 0.5 in is the step between a maker's standard and Plus sizes. |
| Refresh rate | 60→90 Hz, 60→120 Hz | 90→120 Hz | 60→120Hz widely perceptible; 90→120 marginal for most users. |
| Peak brightness | 400 nits apart | 150 nits apart | Outdoor-visibility studies; differences under ~150 nits at these levels are not reliably perceived. |
| Typical brightness | 400 nits apart | 150 nits apart | Same reasoning as peak brightness: it is the same scale read at a different point, and a difference under ~150 nits is not reliably perceived. |
| HDR peak brightness | 500 nits apart | 200 nits apart | HDR highlights only — a few small bright regions of a film frame — so the bands are a little wider than for the whole-screen figures; 200 nits of highlight is inside the spread between two review units. |
| Storage (base) | ×2.0 or more | ×1.4 or more | A doubling is a different buying decision; 128 vs 256 GB changes what you delete. |
| RAM | ×2.0 or more | ×1.5 or more | 8 vs 12 GB is felt only when apps get evicted; 8 vs 16 changes it. |
| Battery capacity | 1000 mAh apart | 400 mAh apart | Capacity is a proxy for the hours rows above it; 400 mAh is inside the spread between two phones' own figures, 1,000 mAh is a fifth of a flagship battery. |
| Battery per charge | 6 h apart | 2 h apart | — |
| Claimed video playback | 6 h apart | 2 h apart | Same deltas as battery per charge: both are hours of use, and the maker's video-playback figure is a lab loop rather than a day, so a two-hour gap is inside test-to-test variation and six is a difference a full day of use would show. |
| Battery lifespan | 400 cycles apart | 200 cycles apart | 400 cycles is about a year of nightly charging. |
| Fast charging | 25 W apart | 10 W apart | — |
| Wireless charging | 15 W apart | 5 W apart | 5 W is the gap between two pads' ratings for the same phone; 15 W is the step from a basic Qi pad to a magnetic fast charger, which is a different overnight. |
| Geekbench 6 (CPU, multi-core) | ×1.3 or more | ×1.1 or more | 30% is a chip generation; 10% is run-to-run and thermal variation between two units. |
| 3DMark Wild Life Extreme | ×1.3 or more | ×1.1 or more | 30% is a chip generation; 10% is run-to-run and thermal variation between two units. |
- These are initial editorial estimates. They are our reading of what a person actually perceives, not the output of a study we ran. They are published here precisely so they can be argued with — a threshold nobody can see is a threshold nobody can correct.
- A difference below the marginal figure is marked as one you will not notice, and no winner is declared for that row. Between marginal and noticeable the row says so in words. Past noticeable, the row names the phone that is ahead — where the registry says which way is better; a bigger number is not always the better one. A gap the table lists the attribute for but does not cover — a refresh-rate step we have not tabulated — returns nothing at all rather than a guess: silence is the honest answer to a question we have not judged.
- Thresholds are of three kinds. A distance ("30 g apart"), where a gap means the same whatever the figures are; a step ("60→120 Hz"), where an attribute only takes a few values and each step is judged on its own; and a multiple ("×2.0 or more"), where a gap only means something relative to the figures — 128 against 256 GB is a different decision, 1,024 against 1,152 GB is not, though the second gap is bigger.
- Only the attributes above are judged. Every other row on a comparison page prints its two figures and says whether they match — "Same." or "Differs, but not ranked." — and nothing else. Chipsets, build materials and camera modules are strings: every reviewer picks a side, by an argument we do not carry, and ranking the letters would be a verdict dressed as data.
What we do not do
- No reviewer-score averaging. Averaging other people's out-of-ten verdicts produces a number whose meaning is nobody's — not theirs, and not ours.
- No affiliate-ranked results. Nothing on this site is ordered by what pays. The catalogue's default order is release date, which is a fact about the phone rather than a judgement of ours.
- No prices without a date and a source. An RRP is a figure retrieved from one shop on one day; printed bare it becomes a promise we are not tracking and cannot keep.
- No carrier compatibility matrix. Retired on 19 August 2026: every current mainstream phone works on every UK network, so a twenty-cell grid per device answered a question nobody was asking — the compatibility question is real only for imports, refurbs and grey-market budget devices, and that is an answer page rather than a table.
- No imagery we made up. Product photographs are the manufacturers' own press images, self-hosted, credited under each one with a link to the article they came from; where no suitable manufacturer image exists, the page carries our technical drawing instead — a to-scale outline from the maker's published dimensions, with rectangular corner rounding as our drawing convention and a round watch drawn as an ellipse at those figures. If a manufacturer ever objects to our use of an image, it comes down the same day and the drawing resumes.
- No prices we invented. The cost panel republishes CeX's own published used and trade-in prices, read daily by product ID, linked to the exact CeX page and dated. Where colours price differently we print the range, never an average. The eBay figure is a median of live asking prices — what sellers hope for, labelled as such, never dressed as what anyone paid. If CeX ever objects, the panels come down the same day.
- No benchmark score as a ranking, and none in our data files. Geekbench 6, 3DMark Wild Life Extreme, AnTuTu and DXOMARK scores are cited on device and comparison pages, each attributed to its publisher, linked and dated. Geekbench 6 (CPU) and 3DMark Wild Life Extreme are judged against the thresholds above, because each measures one thing a phone does. AnTuTu and DXOMARK are composite scores weighted by their publishers, so a comparison prints them side by side and says "Differs, but not ranked." — never a win. None of the four is in the /data JSON files, and none will become a column you can sort the catalogue by: a systematic extraction of someone else's scores is the substantial part of their database, and database right protects exactly that.