Trust & methodology

Evidence should be easy to audit.

HomeLab Toolkit separates what we measured, what we observed, what a vendor claims, and what still needs verification. That distinction matters more than a long benchmark table.

Our five evidence levels

1. Measured: captured with stated hardware, software, settings, and test conditions.
2. Observed: confirmed during hands-on use, with screenshots, logs, photos, or video where useful.
3. Documented: supported by a primary source such as official documentation or a specification.
4. Inferred: a reasoned conclusion from available evidence, clearly labeled as an inference.
5. Unverified: plausible or reported, but not yet supported strongly enough to present as fact.

Hardware testing

Every review starts with an identity record: model, CPU, memory, storage, network controllers, BIOS version, operating-system build, and unit serial mapping. When two units are available, we compare their configuration and repeat short tests to expose unit-to-unit variation.

We record the exact tool version and important settings. Results are not silently combined across operating systems, firmware versions, power modes, storage states, or network paths. Screenshots and logs support the written conclusion; they do not replace an explanation of the conditions.

Practical workloads

Synthetic benchmarks are useful for comparison, but home-server decisions also need sustained and recoverable workloads. Depending on the product, testing may include networking, external storage, idle and load power, thermals, noise, media transcoding, containers, virtual machines, sleep and wake behavior, power-loss recovery, and long-running stability.

Tool formulas

Interactive tools show their assumptions beside the result. Calculations, pasted reports, and generated configurations stay in your browser. We prefer transparent formulas and useful ranges over a single overconfident number. Storage tools distinguish decimal TB from binary TiB; network tools account for protocol efficiency; power tools separate idle and load time; report analyzers leave unrecognized fields unknown.

Corrections

If a material claim changes, we update the page and note the correction when the original conclusion could have affected a purchase or configuration decision. New firmware or software may justify a retest rather than silently replacing the old result.

Read the editorial policy for publication standards, inspect the assumptions behind our interactive tools, or start with the EQi12 evidence hub to see how lab reports, guides and fixes connect.