The One Pick exists because most buying guides don’t measure anything.
They rewrite spec sheets, rank whatever pays the best commission, and call it a review. We publish numbers we produced ourselves, with the conditions attached so you can check our work — or repeat it.
This page explains exactly how. If a claim on this site isn’t backed by something described here, we shouldn’t have published it.
The rule that governs everything
No measurement, no article.
This isn’t a promise we make in prose — it’s enforced in our publishing pipeline. A draft that cites zero measurements is rejected before a human ever reads it, and flagged again at review as no_primary_data. Numbers that appear in an article but not in a measurement file are caught as fabricated_number.
We built it this way because the temptation runs the other direction. It is far cheaper to write “the battery lasts about 20% longer” than to spend two days measuring it. The pipeline removes that option.
What “one pick” means
Most guides list twenty products because twenty affiliate links earn more than one. We name a single pick per category.
That constraint is the point. Naming one thing means we have to be right, and being right requires testing. When we genuinely can’t separate two products, we say so and explain what would change the answer for you — we don’t pad the list to hedge.
How we measure
Devices and computers
We use the same script on every machine, before and after, and record the conditions alongside the result. A number without its conditions is not a measurement.
| What | How |
|---|---|
| CPU throughput | AES-256 single- and multi-core, fixed duration, plus a fixed compute loop |
| Battery health | Cycle count and maximum capacity, read from the system, not estimated |
| Sustained load | Identical workload run to a fixed duration, temperature state recorded |
| Storage | Read/write throughput under controlled file sizes |
Every run records the exact OS version and build, the power source, the machine’s uptime, and whether thermal throttling was active. When we compare two machines, they are configured identically — same memory, same storage, same OS — so the difference we report is the difference you’d feel.
Phones
iOS does not expose temperature or power APIs to apps, so phone testing is manual and deliberately boring. We fix screen brightness, volume, and network conditions, then run an identical sequence — video playback, browsing, sustained 4K recording, idle — and record battery percentage at each step. Thermal behavior is logged observationally: whether the device throttled, whether recording resolution dropped, whether the system showed a heat warning.
OS updates
When a major OS update ships, we measure the same device before and after.
The “before” number is perishable — once the update installs, it’s gone forever. So we take baselines ahead of release day and keep them.
We wait at least 24 hours after updating before re-testing. Right after an update, the system is re-indexing and finishing background work, and battery drain looks dramatically worse than it actually is. Publishing that as “the new OS ruins your battery” would be wrong, and a lot of sites do it every September.
What we don’t do
- We don’t review products we don’t have. If we haven’t physically tested something, we don’t publish a tested claim about it. Where we discuss a product we haven’t measured, we label it plainly as not tested.
- We don’t chase news. Launch coverage is answered better and faster elsewhere. We write the question that comes after the announcement — whether it’s worth your money, and whether it does what it claims.
- We don’t give financial, medical, or legal advice. These are blocked in our review process regardless of how much traffic they’d bring.
Affiliate links and how we’re paid
This site does not currently carry affiliate links. When it does — most likely Amazon Associate links — every one of them will be disclosed, and buying through one will never cost you more than going direct.
Commission never determines our pick. Three things follow from that, and they’re checkable:
- We test before we look at commission rates.
- We recommend against buying when that’s the honest answer, including “keep what you have.”
- Every affiliate link is disclosed. An undisclosed one is treated as a publishing defect, not a style preference.
Corrections
We get things wrong. When we do, we fix the article and say what changed and when, rather than quietly editing it.
If you can show that a measurement of ours is wrong — especially if you’ve repeated it and gotten a different answer — we want to hear it. That’s the most useful mail we get.
Questions about a specific test? Contact us.