Our whole promise fits in three words: what’s worth buying. Honestly. The first two are easy to say — every shopping site on earth claims to know what’s worth buying. The third is the hard one, and it’s the reason this page exists.
“Honestly” isn’t a claim you can just make. You have to show your work. So here is ours: exactly how a product travels from the noise of the internet onto a page where we look you in the eye and tell you to buy it — or not to.
Why this page exists
Most product recommendations you read online are one of two things. They’re either an affiliate link dressed up as advice, written to make a sale rather than a case, or they’re a genuine opinion with nothing behind it — one person’s taste, presented as a verdict. Both can be right by accident. Neither is trustworthy on purpose.
We wanted a third thing: a repeatable way to decide what’s worth recommending, transparent enough that you can check our reasoning and hold us to it. What follows is that method. It isn’t a secret and it isn’t magic — it’s a discipline, applied the same way each time.
The four signals we track
Every product that reaches a Nemus page has been scored against four separate signals. Each one answers a different question, and each carries a different weight, because they’re not equally good at predicting whether something is actually worth your money.
Genuine quality
The heaviest signal, because it’s the one that defends the word “honestly.” We look past the star rating to the real pattern underneath — the shape of the rating distribution, how many reviews are verified purchases, whether the negative ones all describe the same failure, and what owners say months after the novelty wears off. A product that photographs well but frustrates its owners never clears this bar.
What the AI assistants recommend
More people now start a purchase by asking an AI assistant than by opening Google. So we ask them too — the same three questions every time, across four different assistants — and track what they recommend and how consistently. It’s an early read on what high-intent buyers are being pointed toward, before it shows up anywhere else.
Social momentum
We watch how fast a product is being talked about, not how loudly. Rate of change matters more than raw volume: something that’s been viral for a month is already everywhere, while something climbing this week is the actual opportunity. This is the signal we trust least on its own — buzz is the easiest thing to manufacture — which is exactly why it never decides a pick by itself.
The quiet winners
The sleeper hits — things quietly selling and quietly loved, with no viral moment to explain it. These are the hardest to surface and the most satisfying to find, because a good one makes a recommendation feel like a discovery rather than a repeat of what you already saw on TikTok. This is the signal that funds our editorial soul.
Why we don’t use a review-checking tool
For years the honest answer to “are these reviews real?” was a browser extension. Fakespot graded Amazon listings A to F; ReviewMeta recalculated ratings with the suspicious reviews stripped out. Between them they did the heavy lifting on quality checking for millions of shoppers, us included when we were planning this.
Mozilla shut Fakespot down on 1 July 2025 — the website, the extensions and the apps all went dark, with no export, no handoff and no named successor. ReviewMeta had already gone quiet. A method that had been the industry default for a decade simply stopped existing.
We could have swapped in one of the replacements that appeared afterwards. We decided not to, for the reason above: a signal worth 30% of the score shouldn’t depend on a startup that might not be here next year. So our quality checking is now a method rather than a tool — rating distributions, verified-purchase share, whether complaints cluster around one repeated fault, brand-level reputation, and long-form reviews from people who kept the thing for six months. It’s slower. It can’t be switched off by somebody else’s board meeting.
How we score, and what actually gets through
Each signal contributes points; we add them up. A product needs to clear 60 out of 100 to even become a candidate. That threshold is deliberately high — most weeks, only a handful of products across every category we cover make it that far. We’d rather show you a short list we believe in than a long one padded with mediocre options.
But the score doesn’t get the final word. A person does. The number tells us what’s worth investigating; a human editor decides what’s worth recommending. Automated systems get gamed — reviews get faked, trends get bought — so the last step is always judgement, not arithmetic. If something clears 60 and still feels wrong, it doesn’t run. If a product we love falls just short on paper, we’ll say why we’re backing it anyway.
The algorithm surfaces. The editor decides. Neither one gets to skip the other.
Where this stands today
Nemus launched in July 2026, and we’d rather tell you where the method actually is than imply years of history we don’t have. The four signals and the 60-point threshold are what we score against, and the scoring record starts from August 2026 — every product we assess is logged with its score, whether or not we end up covering it.
That record is the part that makes this more than a description. In six months it will show us which signals actually predicted a good recommendation, and we’ll re-weight them from evidence rather than reasoning. When we do, we’ll say so on this page.
What we test ourselves, and what we don’t
Here’s a line we’ll always be straight about: some of our recommendations are hands-on tested, and some are research-led. Both are honest work, but they’re not the same thing, and you deserve to know which is which.
At the moment, most of what we publish is research-led — built from manufacturer specifications, independent reviews, and the pattern of what real owners report. Every one of those pieces says so, in a notice at the top, before you read a single recommendation. When we have lived with a product, we’ll tell you exactly what we did, for how long, and what happened. We will never blur the line to sound more authoritative than we’ve earned.
Before anything gets published
Clearing the score only earns a product a closer look. Before a recommendation goes live it also has to survive a checking pass, and this is where most of our picks actually die.
We open every link before we publish it. Not check that it exists — open it, and look at the page. We confirm the product is genuinely on sale in the UK, under the name we’ve given it, at the price we’ve quoted. It sounds obvious. It isn’t: in one recent audit, six of seven links in our early articles went somewhere wrong, and we’ve since cut recommended products from live pages after discovering they were discontinued here, sold only in America, or listed under a model number that had quietly been superseded.
We publish the corrections at the foot of the articles concerned, including the ones nobody would have caught.
How the money works — and what it never changes
Nemus makes money through affiliate links. When you buy something after clicking through from us, we may earn a small commission from the retailer, at no extra cost to you. That’s the business, and we’re not coy about it.
Our commitment
No brand can pay to be recommended. There is no pay-for-placement, no sponsored verdicts, no ranking bought with a bigger commission. The wall between what we recommend and how we earn is the whole point of the business — the day we sell it, we have nothing left worth selling.
Where two products are close and one pays us more, we’ll point you to the better one and eat the difference. Every page that contains affiliate links says so, clearly, at the top — and where a link earns us nothing, we say that too, next to the link. Several of our current picks earn us nothing at all.
Where we’ll get it wrong
We will get things wrong. A pick will age badly, a product will change after we praised it, a better option will appear the week after we published. That’s not a risk we’re admitting reluctantly — it’s a certainty we’re planning around.
So we treat our recommendations as living things. We revisit the “best of” guides on a schedule, not just when it’s convenient. When something we recommended stops being the right call, we change the verdict and note that we changed it — we don’t quietly edit history and hope nobody noticed. A recommendation you can’t trust to be current is worse than no recommendation at all.
If you think we’re wrong
Tell us. If you bought something on our word and it let you down, or you know a product that beats one of our picks, that’s the most useful thing you can send us — it makes the next reader’s decision better. Genuine corrections get made in public, with the change noted on the page.
You can reach us at hello@getnemus.com. We read everything, and the good challenges change what we publish.
That’s the method. Everything else on this site is just us doing it, in the open, one product at a time.