how we rank
The short version: we don't lab-test products. We compare the people who do, add what owners say, and show you how strong the evidence is for every pick.
- Gather sources. Expert reviews and tests, each labeled by type and dated.
- Collect owner threads. We read Reddit and forum threads by hand and count real owners, positive and negative.
- Apply the gates. No open safety recall, plug-in electronics must carry a UL or ETL listing, no suspicious review patterns, not discontinued.
- Score. The rubric below, the same way for every product.
- Label the evidence. Strong, good or limited, so a pick with thin evidence is never dressed up as a sure thing.
the rubric (v1.0)
| Part | Points | How it's worked out |
|---|---|---|
| Expert evidence | 35 | Type A sources (a published test method or lab) earn 14 points, type B (hands-on review) 7, type C (roundup or survey) 2. At most 2 per type, and one per publisher: sister sites count once. Sources older than 18 months count half, older than 36 months not at all, and undated sources count half. |
| Owner consensus | 25 | Reddit and forum threads we read ourselves. It only counts with 3+ threads and 8+ distinct owners, scored with the lower bound of a Wilson interval so a few glowing reviews can't beat many good ones. |
| Value within its tier | 20 | Compared only with picks for the same need in the same budget tier; the cheaper one scores higher. |
| Reliability and returns | 10 | Warranty and long-term owner reports where we have them. A product often returned on its store listing loses points (we don't display that store data). |
| Fit and availability | 10 | Fits the need (Twin XL, dorm-safe, size) and isn't a one-season or hard-to-find listing. |
These weights are our best estimate and will change as we learn what helps people choose. Every change gets a new version below.
evidence labels
- ●●● Strong evidence: 2+ type A sources from different publishers dated within 24 months, owner consensus met, and no unresolved drawback.
- ●●○ Good evidence: 1 type A source or 2 type B sources from different publishers (without owner consensus, this is the highest a pick can reach).
- ●○○ Limited evidence: everything else.
Right now 0 picks have strong evidence, 98 good and 428 limited, out of 526. Owner threads are still being collected, which is why none are "strong" yet.
roles
- Top pick: the best-ranked pick for a need. A pick with limited evidence can never be a Top pick.
- Budget pick and Upgrade pick: the best-ranked pick in the budget and splurge tiers.
- Skip it: where the evidence says not to bother.
how often picks are re-checked
Every card shows when it was last checked. After 90 days it's due for a re-check; after 180 days it can't be a Top pick until someone checks the listing again. We check the safety recall database every month.
what we don't do
- We don't lab-test, and we never say "tested" or "we tried it" unless it's true.
- We don't show store star ratings or review counts.
- We don't track prices or send price alerts.
- Commissions never change a rank. When a product is only sold on Amazon, we say so.
what changed
- 2026-10-27: Ranking rubric v1.0 released. Picks now show evidence strength instead of a number.
- 2026-10-27 (dorm): Mattress topper, water bottle, laundry bag and backpack moved from essential to recommended after the coverage audit (fewer than 3 of the reference checklists include them). See docs/COVERAGE.md.
- 2026-10-27: rubric v1.0. First release. Weights are estimates; they will be tuned after we see which picks people actually choose.
Why small samples don't win: How Not To Sort By Average Rating (Evan Miller).
See a pick that's wrong, out of stock or recalled? Tell us. And here's how we make money.