Comment by cperciva | Hacker News Reader

Comment by cperciva | Hacker News Reader

PG claims that a shift towards smaller companies results in people being judged on their performance rather than based on factors which do rather poor jobs of predicting said performance; my experience with tarsnap indicates otherwise.

I have a small number of customers -- somewhere between 10% and 20% -- who are using tarsnap based on a recommendation from another tarsnap user; but the vast majority of tarsnap users signed up based on some other factor which is at best predictive of the quality of tarsnap. Reasons I've heard include:

  * I am the FreeBSD Security Officer,
  * They find my blog to be interesting and well-written,
  * I have a doctorate from Oxford,
  * They like the fact that I made the tarsnap client source code available, and
  * I beat them on a mathematics competition 10 years ago.

There are a great many instances of companies succeeding with inferior products as a result of having superior marketing; while I am fortunate that people are interested in using tarsnap for reasons which are quite unrelated to the quality of the tarsnap code or service, the fact remains that the market isn't (in PG's words) "[taking] every organization and [keeping] just the good ones" -- rather, the market takes every organization and keeps just the ones which rank best on a nebulous combination of quality, marketing, and luck.

Yes, credentials are an imperfect way of predicting quality; but are they any worse than the marketing and luck which the market introduces?

Initially the users of a new product will have random motivations. They can't all come from recommendations when you've only just launched. But a year from now the number of users you have will depend mainly on how good tarsnap is.

Certainly, once you get some momentum your startup will survive or die based on how good it is. But if I hear cperciva correctly, had it not been for his credentials lots of users might not have even tried tarsnap to begin with, and I assume that would also translate into far less users a year from now, even though the product would be the same.

I assume that would also translate into far less users a year from now

That's the mistake. The number of users you get long term depends more on the "interest rate" than the size of the initial seed. E.g. Google.

I have to dispute the notion that the "initial seed" is irrelevant. Given a fixed "interest rate", the number of users at any time will be proportional to the initial number of users; and given that the initial number of users can easily vary by an order of magnitude based on external factors, I'd say this is pretty significant. (Of course, it's quite possible that the growth rate is not independent of the initial number of users.)

There's also a more dangerous factor to consider: The number of users of a startup is far more visible than its growth rate, so it's entirely possible that great startups end up shutting down due to a disappointing apparent lack of traction when -- if they had stayed around for longer -- their inherent advantages would have produced enough growth to make them wildly successful. Even if we suppose that the ultimate potential of a business is entirely dependent on the inherent quality of the product or service it produces -- which would be rather stretching credulity -- then the external factors of marketing and luck still have a significant impact on whether companies remain in existence long enough to reach their potential.

I didn't say the size of the initial user base is irrelevant, just that in practice the growth rate matters more.

And while growth rate is not in itself visible, what drives it is something that's usually even more visible than the size of the user base: the quality of the product.

while growth rate is not in itself visible, what drives it is something that's usually even more visible than the size of the user base: the quality of the product.

I'm not convinced. Most startup founders think that their products are the greatest thing since sliced bread, but they're often also aware that they are biassed. How many startup founders haven't at some point been disappointed by the size of their user base and said to themselves "gee, I think what I'm doing is really cool, but maybe the market for this isn't as big as I thought it was"?

I'm sympathetic to this argument, but the "long term" could easily be quite long on the timescale of a startup's solvency. FogBugz, for example, might well not have survived without the initial user base provided by the readers of Joel on Software. Having that promotion channel could easily have meant a factor of ten in the initial customer base compared to a "normal" startup. And that factor of ten represents more than three years of growth even if you double your users each year. Of course, with growth like that presumably you could raise money to increase your runway, but that adds the complexity of dealing with investors and adds more time for things to go wrong.

i'm having difficulty imagining a bug tracking? software having viral growth ... so i guess the 'interest rate' is steadily low; therefore, initial capital is important

Bug tracking software, like other tools for programmers, often travels as a virus with team members. You hire some guy, he says "why are you using this horrible crap we used fogbugz at my last job and it wuz not teh suck", and so you try it.

Maybe we can call this resume-viral: products that are passed around by virtue of become recognizable bullet items on resumes.

or as Keynes said, "on the long term we're all dead"

I think that in case of tarsnap it is relatively hard to make the decision if you don't have basic knowledge in cryptography, at least. That's why the information assymetry occurs and thus signalling has much more impact on it.

Reply on news.ycombinator.com