The Lindy Effect on startup potential

The first 90 percent of the code accounts for the first 10 percent of the development time. The remaining 10 percent of the code accounts for the other 90 percent of the development time.
— Tom Cargill, Bell Labs

However long the project took to get “close to done,” it will probably take that much time again to really be done. It’s funny because it’s true.

Why is it true? Here’s a novel way to frame it: When you’re exploring something new, where the terminus is unknown, you never know how far along the path you are. On average, however, you’re halfway there. This is due to the very definition of “average” — you’ll spend half your time before the half-way point, and half after.


The general rule is called the Lindy Effect: For certain non-perishable things (like technology, companies, and ideas), the expected lifespan increases according to the length of its current age.

A simple way to rewrite the adage is: However far you are, you can probably double it.

Where else does this “doubling” rule apply?

Consider the question: How large can your company get?

If you’ve gotten 100 customers you can probably get another 100 in a similar way. Will you ever get 2000? I hope so, but most companies that do get 100 never get 2000. Putting it another way, doubling the size of the company always sounds plausible, because you’ve proved you can do it once, we’d all agree you can probably do it again, even faster this time. But 10x is not as clear, and 20x is quite unlikely.

One way to understand why 2x is plausible but 20x requires innovation, is to observe that the actions that got you your first 10 customers are probably not sufficient for generating 100, even though they’re probably sufficient for getting another 10. You might have scratched and clawed inside your social network to get the first 10, but that doesn’t scale to 100 (though it might scale to another 10), and just because you were successful convincing customers one at a time to convert through hour-long Skype calls doesn’t mean you can convert 100 customers self-serve via Facebook ads.

Or you might have gotten to 1000 customers through one marketing channel, so although surely that same channel can produce another 1000, it’s unlikely that there is 10x the inventory inside that one channel to get you to 10,000 customers. Thus you’ll need to make other channels work, which is not a gimme, as anyone who’s gone through that challenge can attest. In fact, to achieve 10x you’ll need to make multiple other channels work. Yikes! Possible? Sure. Likely? Not really, certainly not with the confidence you have in 2x.

So the “startup growth” version of this rule is: You can probably double your size, doing roughly what you did to get to this point, but 10x will require innovation.

Or, instead of innovation, time and luck. Specifically, waiting a long time for the existing mechanism to keep working, and lucky that nothing stops that slow but predictable growth: the channel doesn’t saturate and get worse, new competitors don’t arise, market conditions don’t make the product less desirable, the economy doesn’t slow, and so on. Again, possible, but in the fast-moving world of tech, unlikely.

Indeed, at some point 10x is strictly impossible. At the high end, you hit market size limits (Facebook has 1.3B users; there aren’t 13B humans) or run out of marketing channels for acquisition (GoDaddy’s customer growth rate is 13%/yr since 2009; at that rate it will take 18 years for them to 10x, assuming market size and conditions would even allow them to sustain that). At the low end, maybe there really isn’t a market, or the market really doesn’t want that product, at a profitable price, so you can squeeze out some early sales but it can’t get substantially bigger.

You get 2x by assumption but 10x you have to prove. If you’re in the business of raising money, you actually have to prove. But how do you prove something that we just agreed was unlikely?

The solution — the antidote to the Lindy Effect — is extremely rapid growth. If you’re doubling every six months, you clearly have line-of-sight for more growth than just 2x. Your trajectory proves intense market demand is getting coupled with an ability to find and service it. If the underlying market is large and/or growing, you have a good case that 10x is already within reach, and innovation could potentially get you 100x or 1000x. And indeed, the companies who have shown that sort of growth at interesting size have indeed shown 100x or 1000x size thereafter. (“Interesting size” doesn’t mean going from 10 customers to 20 in six months and then being proud of your “rapid growth.” Going from 1000 to 2000 in six months is more like it.)

This is why investors (and founders) wishing to build enormous companies are so fixated on hyper-growth. It’s the only way to have even the potential of building something enormous.

Of course not everyone cares about building something that services a million customers. Even so, applying the Lindy Effect is helpful in strategizing what to do next to build your business.

1,221 responses to “The Lindy Effect on startup potential”

  1. Small correction: “every” at the beginning of the last paragraph should be “everybody”.

    Since you’re starting out with the Tom Cargill quote on development time and then move on to the topic of growing a company, here’s a mathematician’s observation on these two processes. When growing a company, what you’re going for and often achieve is exponential growth, which will take you to saturation rather quickly, as is the case with facebook and similar companies. When developing and perfecting a software product, the very best you can hope for, and you rarely even achieve that, is the inverse of exponential growth: it’s logarithmic growth. You get to something presentable and usable fairly quickly. After that, the amount of improvement you get out of the same amount of effort decreases at a rate of 1/t. If you’re lucky, you’ll end up with something stable that can sustain your business for a good while. If not, you’ll end up with something like a monster of a desktop product in a world where everything’s on the web.

    There is, of course, a high-level explanation for that: the growth of a business tends to feed itself. The growth of complex system tends to inhibit itself.

    • Fixed the typo, thanks. It’s an interesting observation that business growth feeds growth, whereas project growth feeds pain. Both can get caught in its own momentum, however, with software creating legacy code-bases, and companies creating legacy reputations, cultures, and product-lines. Both ensure some sort of longevity, but both devolve into a lack of innovation. Those who can break that pattern and continue innovation, can truly build something special.

      • Ha! I knew there was some sort of commonality between business development and software development, despite their radically different growth patterns. I just couldn’t see it. Thanks for the insight. The enemy to fight, on all fronts, is the decay of the innovative spirit.

  2. It’s like you’re reading my mind. :-)

    This is focused on scaling (and exhausting) marketing channels for customer acquisition, but if what you’re trying to do is 10x $growth_metric and that metric is revenue, couldn’t you take advantage of multiplication and say 10x growth = 2x user base and 5x contract values? Or 3x user base and 3x contract values (OK that’s 9 but you get the point).

    Tom Tunguz had an (as per usual) awesome post about Veeva’s S1 filing where he talks about their ACV being $750k, with a total market of less than 200 companies! In that case, obsessing over channels not as important as obsessing over more value creation for existing base? Any expanded thoughts on navigating that obstacle course?


    • Mathematically you can get the concept of 10x (clearly the point isn’t 10x versus 9x or 11x, but just that it’s significantly bigger than 2x) by increasing two things by 3x that multiply together, like ARPU and conversion %. However, that’s highly unusual. Normally if you 3x prices, you don’t also 3x conversion; in fact usually conversion rates go down, even if the product is higher. (There are fun, notable exceptions, where startups have 2x’ed prices with no change in signups, which is always great!)
      So yes, of course it’s always good to 2x your prices or 2x your conversion rate, and if you’re VERY far away from optimal along several dimensions, there might even be 10x there. Hopefully so! Just not usually so.
      The Tunguz article you’re referring to supports the argument above — i.e. that Veeva has huge ARPU but very small number of customers, and no it doesn’t make sense that Veeva could 3x their conversion rate (or their ARPU!). That’s a special case which isn’t relevant to the article above, because it’s such a highly specialized sales process. There aren’t 10,000 potential customers — in fact perhaps not even 1000 — and there won’t be 10 different, interesting channels, because the customers are so specialized. So that’s an unusual business; interesting, but not the sort that this article applies to.

  3. This is a really interesting post. It feels like the mathematical extension of the “getting your first customer is harder than you think; it is easier than you think” sort of problem. When you start out and have no customers, its so hard to get that first one precisely because you have none. Once you have 10 customers, and you look back at that first customer, its almost comic how easy it was to find and convert that person.

    I love it.

  4. Hey Jason,

    Just re-reading some of your archive. This is probably my favorite of your posts. I mentioned this to you on Twitter maybe a year ago, but if you ever want to package up your blog into a book, standing offer to work on it with you.

    Thanks for taking the time to share :)

Sign up to receive 1-2 articles per month: