Not a mission statement — a working document. This page lists the sourcing rules, the tie-breaking rules, and the update schedule that every comparison on this site is supposed to follow.
Revision: September 20, 2026 · Applies to all published comparisons
We use the word "comparison" rather than "review" on purpose throughout this site. A review typically implies hands-on testing of one product in isolation; a comparison, as we use the term, is an explicit head-to-head built from documentation, independent testing where it exists, and owner-reported patterns — and the word choice is meant to set the right expectation about what kind of evidence sits behind the page, rather than implying a lab test that didn't happen.
Name the specific model, generation, and regional version being compared — not a product family. "The 2024 base trim sold in this market" and "the 2024 top trim sold elsewhere" are treated as two different subjects, not interchangeable data points.
Manufacturer spec sheets, official manuals, warranty terms, and regulatory filings where they're public. This sets the floor: what's claimed, not yet what's true in daily use.
Testing outlets, teardown reports, and long-term owner data are checked against the primary documentation. Where a manufacturer's claim and an independent measurement disagree, the disagreement gets written into the article rather than quietly resolved in the manufacturer's favor.
Specifications are translated into questions a reader actually has: does this last a full workday, does this fit the space, does this cost more to run than it looks. Specs that don't answer a real question get cut rather than padded into another paragraph.
The draft states plainly who each option suits and who it doesn't, instead of collapsing the comparison into a single score that hides which factors were weighted more heavily.
Before publishing, every number that could be wrong — battery figures, prices, warranty lengths, release dates — is checked a second time against its original source, not against the first draft's memory of that source.
Every page gets a review date based on how fast its category typically changes (see the cadence table below), not left indefinitely on a "published" status.
None of these steps are unusual on their own — most competent reporting follows something similar. What tends to go missing on comparison sites generally isn't the individual step, it's step 6: the second-pass check that catches a number quietly copied wrong from step 2 into the final draft. A surprising share of factual errors on review sites trace back to that exact gap rather than to bad sourcing in the first place.
Not every source carries equal weight. When two sources disagree, the higher tier generally wins the write-up — but the disagreement itself is still mentioned, because a reader deciding between two products benefits from knowing where the uncertainty actually sits.
| Tier | What it includes | How it's used |
|---|---|---|
| Tier 1 | Manufacturer spec sheets, official manuals, government safety/regulatory filings | Baseline facts — treated as claims to verify, not as settled conclusions |
| Tier 2 | Independent testing labs, established technical publications, teardown analysis | Used to confirm or challenge Tier 1 claims; disagreements are flagged in the article |
| Tier 3 | Long-term owner reports, repair-community threads, verified forum patterns | Used mainly for durability and real-world fault patterns that reviews rarely catch in the first month |
| Tier 4 | Single anecdotal reviews, unverifiable social posts | Never used alone; only considered when a pattern repeats across many independent instances |
A worked example of how this actually plays out: a battery-life figure published by a manufacturer (Tier 1) claims a certain number of hours under typical use. An independent testing outlet (Tier 2) publishes a lower figure using a fixed brightness and a defined workload. Rather than picking one number, the comparison states both, explains that the manufacturer's figure likely reflects a lower brightness setting and lighter background activity, and tells the reader which number to expect closer to under normal daily conditions. The reader ends up with a usable range instead of a single number that turns out to be optimistic in practice.
This list exists mainly to explain omissions readers sometimes ask about. If a widely circulated claim about a product doesn't appear anywhere on this site, it's frequently because the claim traced back to exactly one of the patterns above and didn't survive the check.
On disagreement between sources: the article states both figures and explains the likely reason for the gap — different test conditions, different regional units, or a firmware version difference — rather than silently picking whichever number looks better for one side.
A comparison isn't finished at publication; it's finished until the next scheduled check, or until a reader flags something sooner. Roughly, the review interval by category:
Fast-moving categories where pricing, firmware, and competing releases shift quickly enough to date a page within a couple of months.
Slower product cycles, but pricing and firmware updates still move often enough to warrant a look every few months.
Model years change less often; the main trigger for an early review is a reported recurring fault pattern.
Model years are the natural checkpoint, though a mid-cycle recall or facelift can trigger an earlier revision.
It would be easier, from a purely presentational standpoint, to average every criterion into one number out of ten. We've chosen not to, because a composite score hides the weighting decision inside the arithmetic. Two readers who care about different things — one prioritizing running cost, one prioritizing raw performance — deserve to see those categories separately rather than pre-averaged into a conclusion that quietly assumes their priorities match the writer's.
Where a page does show a rating, it's presented as a summary of the written comparison, not a replacement for reading it, and the categories behind that summary are visible rather than folded into a black box.
This also means two products can each be the "right" pick for different readers on the same page, without one of them being labeled a loser. A page comparing a compact, efficient laptop against a larger, more powerful one isn't obligated to name an overall winner if the honest answer is that the compact one wins for someone traveling constantly and the larger one wins for someone doing sustained heavy work at a desk. Forcing a single winner in that situation wouldn't make the comparison more useful — it would just make it less honest about where the real answer actually depends on the reader.
A specification true for one market isn't automatically true everywhere. Engine tunes, warranty terms, accessory bundles, and even color options can differ by region, and a comparison written primarily with one market in mind says so explicitly rather than presenting a single global answer. Where the site covers a market with specific practical quirks — locally common fuel-conversion practices for vehicles, regional parts availability, import-versus-local pricing gaps — those quirks are treated as part of the comparison rather than a footnote.
The seven-step pipeline above is the same for every page, but the weight given to each criterion changes by category, and it's worth being explicit about that rather than implying a single universal formula.
| Category | Weighted heavily | Weighted lightly |
|---|---|---|
| Phones | Software update commitment, real-world battery behavior, repair cost and part availability | Peak performance score scores under artificial, non-representative conditions |
| Laptops | Sustained (not burst) performance, keyboard and thermal comfort over long sessions, port selection | Marketing names for chip generations that don't map to a clear performance jump |
| Vehicles | Independent crash-test results, real-world fuel or energy use, verified reliability patterns, local service network | Trim-only styling packages that don't touch cost, safety, or reliability |
| Appliances | Motor duty cycle, verified energy rating, noise under full load, parts availability | Program count beyond the handful most owners actually use |
| Software | Actual pricing at typical team size, data portability, update frequency history | Feature-count comparisons that ignore whether either product's users touch most of those features |
It's worth being direct about the limits of this process rather than letting the page imply more rigor than actually exists. This site does not operate a physical testing facility, does not run controlled repeatable performance scores on every product mentioned, and does not have staff in every country whose market quirks get discussed. Where a claim rests on a source we haven't independently verified, the article says so rather than presenting it with unearned confidence.
The methodology is also not static in the sense of being beyond revision. If a reader points out a real gap in how a category is being evaluated — a criterion we're underweighting, a source category we've been too quick to trust — this page gets updated along with the practice it describes, not just quietly ignored in favor of the comparisons that already exist.
It's also worth naming the practical limit on scope: this document describes the standard applied to product comparisons specifically. General buying advice that doesn't compare named products head-to-head — the kind covered on the How to Choose page — follows the same sourcing honesty but doesn't run through this exact seven-step pipeline, since there's no specific pair of products to define, source, and fact-check against each other.
A useful test of any methodology is what happens when it's followed and still produces a wrong first draft. On more than one comparison, an initial verdict favoring a product on paper was revised after step 3 turned up a recurring, independently reported hardware fault that hadn't yet shown up in first-week reviews — the kind of pattern that only becomes visible once enough units have been in use for six months or more. The published version reflects the revised conclusion, and the original reasoning isn't hidden — it's acknowledged as the reason the page changed, consistent with how the Corrections page documents this kind of update.
The methodology above only means something if it's applied the same way regardless of who the product's manufacturer is or whether an affiliate commission is attached to a given link. A few concrete rules keep that true in practice rather than as an aspiration: no comparison's verdict is edited after a commercial partnership begins with either brand mentioned in it; affiliate links are added to an already-finished verdict, never the other way around; and no single advertiser or affiliate partner accounts for enough of the site's revenue that losing that relationship would change what gets published about them.
Where a conflict does exist — for instance, a product sample provided free by a manufacturer — the article discloses it directly rather than folding it silently into the "independent testing" category described above.
The scheduled review cadence in the table above is a default, not a ceiling. A report involving a safety issue, a factual error that could lead to a bad purchase (wrong price, wrong compatibility claim, wrong capacity figure), or a broken link to a source moves ahead of the normal schedule and is typically addressed within a few days rather than waiting for the category's next scheduled pass. Cosmetic issues — a slightly dated photo, a stylistic preference — stay on the normal cadence, since fixing those doesn't change whether a reader makes a good decision.
The fastest way to trigger this fast lane is a specific, sourced report through the contact page: which page, which exact claim, and ideally a link or document showing the correct figure. A general "this seems off" report still gets looked at, but it necessarily takes longer to act on than a precise one.
Comparison length isn't a target on this site — it's a byproduct of how much genuine trade-off exists between two products. A close matchup between two well-matched options usually needs more explanation, because the deciding factors are subtler and depend more on the specific reader's situation. A lopsided matchup, where one option is simply the better choice for nearly everyone, gets a shorter, more direct verdict, because padding that kind of page with extra sections wouldn't add real information — it would just add length.
For products sold under different regional names, trims, or specification sheets, the article specifies which exact regional configuration is being described up front, rather than blending figures from multiple markets into one table. Where a manufacturer publishes documentation only in a language other than English, the translation is checked against the original technical terms rather than relying on a single machine translation pass — unit conversions in particular (fuel consumption figures, battery capacity units, engine displacement) are cross-checked against the source document's original units before being restated.
Does a manufacturer ever see a draft before publication? No. Nothing is shared with a manufacturer or retailer ahead of publishing, and no comparison is delayed to accommodate a company's preferred launch timing.
What happens when no independent testing exists yet for a brand-new product? The article says so plainly and relies more heavily on manufacturer documentation, clearly labeled as such, until independent data becomes available — the review-cadence table above then triggers an earlier-than-usual revisit once that data exists.
Can a reader see the sources behind a specific claim? Where practical, sources are linked directly in the article. Where a claim comes from a private communication (a support ticket reply, for instance), the article notes that origin rather than presenting it as a public, checkable source.
What if two Tier 2 sources disagree with each other, not just with the manufacturer? The article notes both figures and, where identifiable, the methodological difference that likely explains the gap — different test conditions, different sample units, or a different firmware version at the time of testing. Where no clear explanation can be identified, the page says the sources disagree without inventing a tidy resolution that isn't actually supported.
Does this page apply to older comparisons published before it was written? Yes — this is a description of the standing process, and older pages are brought in line with it as they come up for their scheduled review rather than all at once, since rewriting everything simultaneously would trade thoroughness for speed in exactly the way this methodology argues against.
None of this is meant to read as a guarantee of perfection — it's a description of the standard we're holding ourselves to and the specific mechanism for catching the times we fall short of it. A methodology page is only as good as the corrections record behind it, which is why that record is public rather than summarized away. If something on this site doesn't match what's described here, that gap is worth reporting through the contact page — it's the fastest route to a fix, and the reason this process exists in written form in the first place rather than as an unwritten habit.
If you're reading this because a comparison on the site made you doubt it, that's a fair reason to be here. Read the pipeline, check whether the page in question actually followed it, and tell us where it didn't. A methodology that nobody is allowed to test against real pages isn't worth much, and this one is written down precisely so that it can be tested.
A correction isn't a patch bolted onto an old page — it re-enters at step 3 of the pipeline above. The disputed claim is checked again against its original source and, where relevant, a higher-tier source is sought out before the article is edited. The Corrections page lists what changed and when. If a reported issue turns out to be a genuine gap in the sourcing, the same fix is checked against every other page that made a similar claim, rather than treated as a one-off.
In practice this means a single reader email can quietly improve a dozen pages. A wrong warranty figure reported on one comparison gets searched for across the whole site, because the same wrong number was probably copied into other pages from the same original source. It's slower than fixing only the page that was reported, but it's the difference between correcting one mistake and correcting the habit that produced it. Reports that turn out to be mistaken on the reader's side are answered too, with the source that settles the question, since a clear explanation costs us a few minutes and saves the next reader the same doubt.