Editorial policy

How a part earns a recommendation

Ranking starts from proven quality per dollar rather than from price alone: a part's rating, weighted by how many people rated it, divided by the square root of its price. That formula is why a well-reviewed mid-priced item usually outranks both the cheapest option and the flagship, and it is applied identically across every category so the ordering is comparable.

On IRL paths a camera is scored differently, because a field camera and a desk webcam are not the same product. Low-light ability, motion handling, power draw, and runtime on a standard battery bank replace the webcam spec columns entirely. A webcam that wins at a desk can rank last on a street, and it should.

Nothing is ranked by commission rate. Where a category has a paying affiliate option and a free or cheaper one that is genuinely better, both are listed and the better one is the pick.

What we do not do

We do not lab-test hardware. There is no test bench here, no measurement rig, and no sample loans. Every ranking is computed from published specifications, tracked prices, and retailer rating data, and every stream-quality verdict is a conservative planning budget produced by a published model, not an observed result. Saying that plainly is more useful than the alternative, which is letting the layout imply testing that never happened.

Where a number comes from testing someone else did, the page says whose.

What earns a part its own page

Part pages ship only when there is real data behind them: a real image plus either a rating or at least three tracked price points. Everything else stays unpublished rather than becoming a thin spec dump with a headline. Parts cross the threshold automatically on the next data refresh, which is why the number of published part pages grows without anyone writing them.

A part page shows a trade-off only when someone has written one. It never generates a sentence about there being no catch: that is reassurance copy, and a review page with an empty con column is telling you something real about how much editorial is behind it. The same rule applies to the case for a part, which is why some pages carry only the intro paragraph.

What counts as a source for a streamer setup

Every item on a streamer page carries one of three labels, printed beside it. First-party means the streamer, their organisation, or the company that built the rig published it: a setup tour, a tech page, an interview, a clear on-stream shot. Reported means independent gear trackers agree and nothing first-party confirms it. Our pick means nothing has been published for that slot and the item is the one we would put there to match how they stream.

Gear aggregator and setup-tracker sites are cited, and they are tiered. For most streamers they are the only public trail that exists, so removing the links would not make a page better sourced, it would make it unsourced. What they cannot do is confirm anything: several of them contradict each other and none cite a primary source, so agreement between them supports a "reported" label and never a confirmation. Each citation at the foot of a streamer page says which tier it is.

Where nothing at all has been published, the page says so in its opening paragraph and every item is labelled as our pick. A page is never published claiming certainty it does not have. When better evidence appears, the label changes with it.

The picture beside a streamer is their own channel avatar, taken from the profile they publish on Twitch or YouTube and served from our servers rather than hotlinked from theirs. It identifies whose page you are on, nothing more. A handful of streamers instead carry a photograph under a Creative Commons or public-domain licence, and those are credited beneath the image. Any creator who would rather not appear here can say so and the page comes down.

Prices, ratings, and stock

Prices are curated estimates plus tracked history baked at build time, not a live feed, and they can differ from the listing you land on. The retailer page is always authoritative and the buy link goes straight to it. Each price is labelled with the date it was last observed; the newest observation anywhere in the current bake is 2026-08-26.

A price is only called below average against a bounded 12-month window ending at that part's own newest tracked point, and only when at least three points fall inside that window. Averaging an entire history instead would mark a product that simply drifted down over three years as a permanent deal, which is how the earlier version of this worked and why it changed. Every "below average" figure on the site is that same figure.

Ratings and review counts shown on part pages come from the retailer listing and are labelled as such. They are not our own testing scores, they are never presented as if they were, and they are deliberately left out of our Product structured data rather than claimed as ratings we collected.

Corrections

Errors get fixed rather than quietly deleted. If a claim was wrong in a way that changed a buying decision, the page says what changed. Gear that a streamer has since replaced, a platform spec that moved, a price that has drifted far from reality: all of it is worth an email, and corrections are the most useful mail this site gets.

What we will not publish

Content generated to fill a keyword with nothing behind it. Specifications copied from a manufacturer page and presented as a review. A quality verdict stated as a promise when mobile uplink cannot promise anything. A streamer's gear presented as confirmed when the trail ends at an uncited tracker. Compatibility advice we have not reasoned through, since the whole point of the compatibility engine is catching the purchase that will not work.