This page describes how a product evaluation should be conducted and what it must disclose. It is not itself a review, and it does not rate, rank, or recommend any product.
01What an evaluation has to disclose
A review is a claim about a product made by someone with access the reader does not have. That asymmetry is the entire value of a review, and it is also why the conditions have to be stated. Without them the reader cannot tell a considered assessment from an impression.
The minimum disclosure set:
- How the unit was obtained. Purchased at retail, loaned by the manufacturer, supplied
permanently, or provided as a pre-release sample. Each of these creates a different pressure, and the reader is entitled to weigh it.
- Any commercial relationship. Affiliate links, referral arrangements, advertising relationships,
or a competitive interest in the category.
- How long it was used, and for what. A week of ordinary use and an afternoon of unboxing
produce different levels of confidence and should not be presented identically.
- The exact configuration tested, including firmware or software version, and the date. Products
change materially after release, and an undated review silently becomes a claim about a version that no longer ships.
- Test conditions, for anything measured. A number without its method is not a measurement.
- What was not tested. Long-term reliability, support quality, and durability are the attributes
buyers care most about and the ones a review period rarely covers honestly.
02Evaluating against use, not against a specification sheet
Specification comparison is the easiest form of evaluation and the least informative, because specifications are chosen by the manufacturer to be favourable and are frequently measured under conditions that do not resemble use.
A more useful evaluation starts from the task. Define what the product is for, define what successful performance of that task looks like, then assess whether the product achieves it. This reorders the findings around the reader's decision rather than around the manufacturer's marketing structure, and it exposes the things specification sheets omit entirely: setup friction, failure behaviour, whether the software is required and whether it is any good, what happens at the edges of the intended use, and how the product behaves when something goes wrong.
Where measurement is used, it should be reproducible. State the method, the conditions, the number of runs, and the variation between them. A single run reported to three significant figures implies a precision that was not achieved.
03Reading a review critically
Most readers arrive at a review with a decision already partly formed, which is the condition in which confirming evidence is most persuasive and least useful.
Useful questions to hold while reading:
- Does the review state what it did not or could not test?
- Does it describe a specific reader, or is it addressed to everyone? A review that suits everyone
has usually not committed to anything.
- Are the criticisms specific and consequential, or are they token balance?
- Is the date visible, and does the version tested still ship?
- Would the conclusion change for a different use case, and does the review say so?
A single review is one data point from one unit under one set of conditions. Manufacturing variation, firmware changes, and individual fit or preference all mean that agreement across several independent evaluations is worth considerably more than depth in any one of them.