Reviews are written on hardware supplied before general availability. That timing produces several systematic differences between what a reviewer tests and what a customer eventually receives.
Early production is not settled production
Manufacturing improves through a ramp period as tolerances tighten and assembly problems are resolved. Units built at the start of that period are not representative of the average.
They can be better, since early batches are often built with more supervision, or worse, since the process has not yet stabilised.
Component sourcing also changes over a product's life, and parts such as storage modules or display panels may come from different suppliers in later batches with slightly different characteristics.
Software is usually pre-release
Review units frequently ship with firmware that is not final, sometimes with a note that an update will arrive before customers receive the product.
Performance, battery behaviour and camera processing can all shift meaningfully with that update, which means a review may describe behaviour that no retail unit exhibits.
Continuous updates extend this problem in the other direction, since a product reviewed at launch may behave differently a year later after several revisions.
Conditions of supply shape coverage
Loan units are provided under agreements covering embargo timing and return, and access to future units depends on an ongoing relationship with the manufacturer.
This does not require anyone to write dishonestly. It creates a mild, persistent pressure that operates through what gets covered and how criticism is framed rather than through instruction.
Publications that buy their own units remove that pressure, at a cost that limits how many products they can cover and how early.
Testing windows are short
Embargo schedules often allow days rather than weeks, which is enough to assess design and performance and insufficient to detect problems that emerge with use.
Battery degradation, durability, thermal behaviour after months of dust accumulation and software stability over time all fall outside that window by construction.
Sample variation is invisible in a single unit
Any manufactured item varies, and a single review unit cannot distinguish its own characteristics from the product's typical ones.
Display uniformity, panel brightness, coil noise and processor performance under sustained load all vary between individual devices from the same line.
This is why aggregated owner reports become more informative than launch reviews after a few months, even though they are less rigorous individually.