Estimating the value seems pretty much impossible, but estimating a length seems more reasonable. Clever software trick, quick to invent, easy to capitalise -> 2 years. Ground breaking new drug/material with slow market deployment -> 20 years.

There's still some subjectivity involved (but that's already the case with the current patent system), but you could try to base it on some objective metrics.

It would be easy to discriminate between 20 years for drugs vs. 5 for anything software, but I guess it's tricky to decide whether something like e-paper should count as fast-moving consumer-facing tech or hard-won physics/chemistry knowledge.

No method is perfect, but I think that in most cases a 5-year protection is enough to encourage invention.

In general, it is a business decision on whether to patent a technology that it is not ready for the production: the risk of not making a product in 5 years vs the risk of competitor getting there first. It is not perfect, but I would prefer that inventions that the owner wants to own (and suppress competition for) but cannot commercialize in 5 years enter public domain. My 2c.