Charlie Songhurst

If you were sort of modeling Nabisco and Barbarians at the Gate, it's obviously the right conceptual approach.

Maybe that's just not right for network effect businesses, because instead, literally, how do you model in Excel the concept of in year six, something becomes a standard and therefore gets sustained or accelerating growth?

There was a sort of joke in the 80s of you'd never get fired for buying IBM.

Well, maybe in 2006, it suddenly became you never got fired for buying Salesforce.com.

How do you model in that as a concept suddenly kicking into revenue growth?

Maybe what you should actually be doing is writing a sort of 3,000-word essay on revenue growth drivers as opposed to sort of trending it down over time as an automatic default.

So there's sort of certain things where I think the industrial era protocols actually just mislead you.

Conversely, I think the counterpoint to that, which is the sort of too product-driven mistake, is when you look for network effects everywhere and you sort of assume they exist and actually you're just getting into a standard competitive market.

Trying to turn that to startup investing, I think a very good rule of thumb is to think it's very hard to extract economic lengths to get revenue without solving someone's problem.

So go right back to a sort of Jeremy Bentham utilitarianism.

Why are you making people happier?

How much happier are you making them?

And what percent of that can you extract as economic surplus?

And what percent goes to them?

So an interesting one is, you pick on Google, there's lots of charts saying people would pay thousands and thousands of dollars for search.

Yet Google's ad revenue per person is much lower.

So potentially, the utilitarian output of Google is 10x, maybe even 20x, the economic value derived from that.

Whereas I imagine if you looked at something like Oracle, you would actually find the utilitarian benefits...

The economic grants are a much higher percentage of the utilitarian benefits.

There's more value capture as a percentage of the total sort of common wheel created, the total common utility.