How I Review

Most product reviews on the internet are written by someone who has never touched the product. That's not a hot take. They read the manufacturer's marketing page, rephrase the claims, sprinkle in "game-changer" and "must-have," slap a 4.8 out of 5 on it, and publish. The affiliate commission pays out whether the product works or not.

I don't do any of that. This page is the exact process I use for every review on this site. It exists so you can decide whether to trust me — and so you have a standard to hold me to if I ever drift from it.

Same site. Same checkout. Same shipping wait. No PR samples, no review units, no early access, no discount codes that aren't available to every other reader.

If I can't get a product the way you'd get it, I can't review it honestly. The supplement that shows up with special "reviewer's pricing" is a different product than the one in your mailbox — different service, different expectations, different experience. The review would be a review of the reviewer's version, not yours.

There's no fixed number of days here, and I'm skeptical of reviewers who claim there should be. A $15 pour-over kettle tells you everything it's going to tell you inside a week. A mattress, a sleep tracker, or a supplement needs a month or more. A piece of software needs enough real tasks to reveal its edges. A razor needs enough shaves for the blade to actually dull.

The honest rule is this: I use a product until I can answer "would I buy this again with my own money?" with conviction. If I can't answer that question yet, I'm not done testing. The testing window for every review is published at the top of that review, so you know exactly what you're looking at.

Some reviews take three weeks. Some take three months. None of them get published early.

Where measurement is possible, I measure.

Heart rate monitors get compared against a Polar chest strap. Sleep products get tracked alongside Oura ring data. Supplements making physiological claims get bloodwork — before and after. Blenders get timed. A standing desk's wobble gets recorded on camera. A knife's edge gets tested on the same set of tomatoes, same technique, same day.

Not everything is measurable. "This razor feels smoother" is a subjective claim. "This chair is more comfortable after six hours" is partly subjective. I tell you when something is subjective, and I try to triangulate — comparing to products I've already tested, asking other people to try the same thing, or running the same task with and without the product.

Anything reported as a number in a review has a source. If I can't source it, it doesn't get a number.

Before I finalize any review, I sit down and ask myself: what's the strongest case against this product? Then I write that case out in plain text and take it seriously.

• Who is this actually wrong for?
• What would a competitor's best engineer say about this design?
• What does the worst verified complaint look like, and is it a real problem?
• If I got this as a gift, would I quietly return it?

If the product survives the strongest objection, the review reflects that with confidence. If it doesn't survive — the review reflects that too, even when the affiliate payout is great.

This is the step most reviewers skip. It's also why most reviews read the same.

I use a 1–10 scale and I use the full range. A 7.4 means "actually good for most people, with specific caveats I'll explain." A 5.0 means "this exists and works, but I wouldn't buy it." A 3.5 means "don't." Below 3, we're in Graveyard territory.

If every review on a site scores above 8, none of them mean anything. The full rating scale — what each number actually signifies — lives at /editorial-standards.

Sub-ratings (Effectiveness, Value, Quality, Ease of Use) are independent of the overall score. A product can be high-quality and still be a bad deal. A product can work brilliantly and be miserable to use. The sub-ratings exist so you can weight what matters to you.

Products change. Formulas get reformulated. Companies get acquired and their QC collapses. Subscription prices creep up. Customer service gets outsourced. A product that was a 7.8 in February can be a 5.1 by October without a single word of the original review still being true.

Every review on this site shows a "Last Updated" date at the top. When something material changes, the review gets revisited — not quietly edited to look better, but updated with a visible changelog at the bottom. If a score drops by more than one full point, a banner at the top flags it. Corrections are public and logged at /corrections.

Nothing gets silently deleted.

Some products don't make it to the testing phase:

• Anything making medical claims that require FDA approval it doesn't have
• Products from companies I've previously caught using fake reviews, fabricated studies, or fake scientist endorsements (the list exists and it grows)
• Products sold only through high-pressure upsell funnels with no straightforward purchase option
• Multi-level marketing products, full stop
• Anything where I can't identify who actually manufactures it, where, or from what
• Products whose landing pages use fake countdown timers, fake stock counts, or fake testimonials from people I can reverse-image-search

I'd rather publish fewer reviews than review the wrong products.

• Accept a free product in exchange for coverage
• Accept payment for a positive review, ever, under any framing
• Take down a negative review because the brand asked, offered money, or threatened legal action
• Publish a review I haven't fully tested
• Use AI-generated content to pad a review, generate fake experience, or invent benefits
• Write affiliate-heavy "best of" listicles that hide commercial intent behind informational framing
• Run sponsored content, sponsored posts, or paid placements of any kind, including the kind that other sites hide under labels like "partner content"

If any of this changes — if I ever accept a sponsorship or a paid placement — it will be disclosed in a banner at the top of every affected page, and this section will be updated to say so.

Every real review on this site includes:

• A "Tested for X days" line at the top
• At least one original photograph of the product in actual use (not the manufacturer's stock image)
• Specific, concrete details you couldn't get from the product's sales page
• Identified failure modes — because every product has some, and if none are listed, the review isn't real
• A "Who shouldn't buy this" section as long as the "Who should" section
• A published date and a last-updated date, both visible

If a review on this site is missing any of these, email me. It means I missed my own standard, and I want to fix it.

Email jeff@jeffsreviews.com. Tell me what I got wrong, what your experience with the product was, or what I missed. I read every email that isn't from a PR agency.

If a review is inaccurate, I update it. If the correction is material, it gets logged publicly at /corrections with the original wording, the corrected wording, and the date of the change. You can see every correction I've ever made — and the fact that there aren't many is part of the point.

The full picture lives at /disclosure. The short version: I earn affiliate commissions when readers buy products through links on this site. That's how the lights stay on.

This never affects what I recommend. The reviews with zero affiliate availability sit alongside the ones with the highest payout rates, and the scores aren't correlated with commission. If they were, the Graveyard wouldn't exist — every product in it was on an affiliate program when I tested it, and the negative review went up anyway. See /graveyard for the full list of products that didn't make it.

The only way an affiliate-supported review site works honestly is if the reviewer is willing to publish negative reviews of products that pay well. That's the test. I take it seriously.

That's the process. It's not complicated, and it isn't secret. It's just rare, because most of the internet makes more money skipping these steps than running them.

If you catch me missing any of this, tell me. I'll fix it publicly.