How it works

shouldibuythat answers one question — should I buy that? — with one verdict, drawn from what people who actually own the thing say on Reddit. Not marketing copy, not affiliate fluff, not a five-star average that hides the one-stars.

The four verdicts

Every product lands in one tier. The tier is a recommendation shape, not a score:

• Strong Buy — owners recommend it broadly; few regrets, durable over time.
• Buy If — worth it, but only under a stated condition — we tell you which.
• Risky Buy — real, recurring caveats; buy with eyes open.
• Stay Away — don't — unless a narrow exception applies, which we'll name.

That's why we publish a conditional line (“Buy if… / Buy only if… / Don't buy unless…”) instead of a star rating. A number averages away the exact thing you need: who this is right or wrong for.

The evidence floor

A product stays unpublished until there's enough real discussion behind it — by default at least 150 owner mentions across 12+ ownership threads from 60+ distinct authors. If a product hasn't cleared the floor, we'd rather show nothing than a confident verdict built on a handful of comments. Quiet is more honest than wrong.

The floor is set per category, because volume differs enormously between them: a popular headphone draws thousands of comments where a projector draws hundreds. Holding both to one number would either wave through thin verdicts in busy categories or silence quiet ones no matter how thoroughly they're covered. What a quieter category relaxes is the volume— how many mentions and threads — never the number of distinct owners behind the recommendation rate, because that's the figure the percentage actually rests on. Projectors run 60 mentions / 8 threads / 40 owners. Every verdict records the floor it was held to.

How a verdict is built

We read public Reddit ownership threads, weight them by upvotes and how long someone has owned the product, and extract recurring praise, complaints, failure modes, and how sentiment ages. Everything is paraphrased, never quoted — we summarize the consensus and the dissent, we don't reproduce anyone's words. The recommend ratio you see is the share of opinionated owner mentions that recommend it; it is not a review score we collected.

Where we draw the lines

Our numbers come from Reddit, and Reddit grades tougher than star ratings — two things worth knowing before you read any verdict.

First, the recommend percentage. We count one vote per person, and threads are full of people whose answer is “get something else instead.” Even the most beloved products in a category rarely clear 80% here. So when three-quarters of owners still say “buy it,” that isn't a B-minus — that's as close to consensus as the internet gets.

Second, the serious-failure number. It is not “the chance your unit breaks,” and it is not every gripe — comfort, setup, filter cost, and support complaints are tracked separately asconcerns, because they change whether people recommend a thing without being failures. This number counts only serious failures: the unit dying or a major defect, among long-term owners. And a raw failure percentage only means something next to similar products — headphones with batteries and hinges fail more than a steel kettle ever should. So we never call 10% (or any number) “background noise” in the abstract. We show the product's rate and its category baseline, and judge it against that baseline.

The verdict is a simple grid — how strongly owners recommend it, against how its serious-failure rate compares to its category:

“Elevated” means serious failures run above the category norm (or past ~20% outright); “Severe” means well above it (~35%+, or one part failing for a large share of owners). That's the whole rule — understandable, not a black box.

Both axes are judged against the category, not the world. Headphones with batteries and hinges fail more than a steel kettle ever should — so we flag a product's breakage only when it runs meaningfully above what its own category typically suffers, and we show you that category baseline right next to the number. Recommendation works the same way, because that's how a friend talks: “best of the over-ears” means something even when the whole category runs rough. A product clearly above its category's typical recommend rate can move up one band; clearly below, one down.

Two guardrails stop that from becoming grading on a curve: no product reaches the top band below 65% of owners recommending — the best of three bad options is still not a Strong Buy — and the comparison only exists at all once a category has five products with real data. We show the category average and the product's rank right on the page, so you can see the curve we saw. Every verdict publishes its inputs — the mention counts, the ratios, the failure math, the category baseline — so you can disagree with where we drew the lines, but not with the numbers.

The honesty wall

• The verdict and its tier are decided before any price or affiliate link is attached.
• We never reorder recommendations by commission. The better-reviewed pick wins even when it earns us less — and we'll point you to it.
• When we make money, we say so, and it changes nothing about the verdict.

Questions or a product we should cover? Join the club — that's where the weekly category breakdowns go.