[mildchalupa]

In hardware all products are constantly changing.

Examples: A part gets obsoleted by vendor, sustaining engineering has to find a replacement. They conduct a small study to see how the new part works. The budget for conducting lengthy tests to determine if a simple change has any effect at the worst case tolerances is typically not there.

Suppliers quality changes over time. This is known as supplier quality fade. Suppliers will often underbid a part to win a contract. Then once volume ramps up they begin value engineering. Maybe they add more regrind material to an injection mold. Or they progressively begin to make the part slightly thinner on each subsequent batch. Maybe they begin to use a toxic filler material without notice.

An colleague literally had a situation in which a steel casting supplier was putting rocks in the castings.

Companies move injection mold tooling from vendor A to vendor B. The tooling might be the same but the settings are different, so the results are then different.

Without an excellent quality team that has money and time to check everything this will happen.

If you are competing in a marketplace with low margins there will never be enough money to check everything.

[coldtea]

Perhaps part of a product, each supplier, and each changed spec should contribute to a unique fingeprint.

If you have a product which you keep calling FOO but has changed some spec in the new version (without hw change, perhaps you enabled some firmware option and now it can record 25 minutes instead of 20), it should get a new fingerprint. If it has changed a supplier for a chip, new fingerprint. And so on.

And aside from the fingerprint, you should also give all the entries that contributed to it (which when hashed in a structured function should give the same fingerprint), plus things like its production date.

Name="FOO", cpu.model="x-dragon-v5", cpu.supplier="Fukimata Japan", ..., fan.supplier="Fans-R-Us",...

Reviews then are tied to the product+fingerprint instead of just the product.

[vianneychevalie]

I work in data management for large companies and I can promise you the problem is not a technical one. Which organization has the maturity to actually follow-through with quality data for such systems? Which legislators have the power to enforce such an approach?