This is only if you ignore why people moved to NoSQL in the first place.

PgSQL is great for a lot of things, but I would argue that if you're using it you're betting on your overall product/service having some other killer advantage than data processing. The competent wing of the NoSQL crowd are using it in strange ways that enable new classes of product and service that cannot be achieved with PgSQL.

I get the feeling you're comparing NoSQL to what PgSQL was 4-5 years ago. Modern PgSQL has a pretty rich set of tools for handling schemaless data, and combines that with its traditionally strong support for relational data. PgSQL has a direction that cares a lot more about data than doctrine.

Though your last line makes me curious. What is an example of one of these new classes of product and service that can't be achieved with PgSQL?

Please elaborate?

Can you give examples? I assume you're talking about hadoop-related tech, but you left it vague.

A lot of the Hadoop users are in the right ballpark. Not necessarily that stack, but that approach to things, and the problems they are attacking, especially graph based data.

My broader point is that if your project fits into pgSQL then you need another unique selling point and the data functions are just an implementation detail of some other aspect of your offering, whereas for many of the people not using that kind of thing their analytics and data systems are their selling point. (That's a backhanded compliment to pgSQL, in that it's easy enough to get right on small systems that any competent developer should be able to manage it, thus reducing the market value though). There is, of course, the blurred line of crazy MySQL deployments, many of which are barely relational.

I'm trying to follow this reasoning. I guess the idea is that if a system has a smaller market share, then it's less likely to be used by your competitors, and can therefore offer a competitive advantage. Is that right?

That doesn't make sense to me because it only really applies when there is a high likelihood of competitors using one product but not the other. Although postgres is doing great, in most markets it's still far more likely that your competitors are using oracle or sql server. So any advantage postgres has -- and I believe there are many -- offers a potential competitive advantage.

For instance, you could argue that data systems are a critical selling point of Heroku, and they use postgres.

I think your point ultimately boils down to: "postgres is not quite at the forefront of certain analytic use cases", which I agree with. It is at the forefront of many other use cases though.

That's not quite my reasoning, so I'll try again!

pgSQL does what it does easily enough that it reduces the barriers to entry to such a level for traditional RDBMS workloads (which there are plenty of) that such workloads are simply not economically worth pursuing (especially for startups) except as small components in larger systems where the value add is elsewhere. The "other" world of big data/time series/graphs/nosql is hard to get right, and so is worth more, as if you crack it someone else copying you is decidedly non-trivial, meaning that it alone can form the core of a successful business.

This is a bit like what the web people are trying to do in mobile, where if HTML5 was magically the best cross platform mobile deployment option when we wake up tomorrow the value of mobile developers will collapse, and the web people will then cease to be remotely excited about mobile.

Oh, so it's a barrier-to-entry argument? That makes more sense. The barrier to doing something with postgres is pretty low, so competitors can more easily copy your ideas unless there are more barriers somewhere else.

An interesting point. More broadly: if what you're working on is not hard (and awkward), then others can copy your idea easily. Technology like postgres makes a new class of problems easy, and thus you need to find new problems to solve if you want a sustainable business.