In Theoryland, which resembles the real world but entirely resides in theory, you could adjust your prices periodically until you slowly integrate to the point where your profits (or revenues, if you wish) are maximised.

Why this is in Theoryland is because you cannot adjust prices and get feedback so easily in real life. But hey, the logic checks out!

> It seems like a really powerful way to price products, but no one appears to do it.

In a closed system. There are too many factors obfuscating the landscape in real markets (most notably consumer access).

The only tidbit is, in a new market you can overcharge.

> There are too many factors obfuscating the landscape in real markets

On the one hand, yes, but on the other hand, machine learning gives us a powerful set of tools to analyse these factors, which we could use to come up with a model that learns from the market to predict consumer surplus (but you're right about the naive economist approach).

Test different price points. Derive price elasticity of demand. Optimize price. That's the theory.

In practice? Pick a random number that doesn't sound too stupid. Try raising your price. Keep doing that as long as it works.