Is your startup’s technology defensible? My test for technical defensibility in a startup

When I am working on the technical due diligence of a startup company before VC investment, one of my goals is determining in what ways the technology being built affects the company’s strategy: how (and if) it contributes to the product’s defensibility.

Technology is not the only way to create defensibility. Network effects, having better data or money than the competition, a strategic relationship with key players, etc. can all be important. In some cases it makes sense to invest in a company where the technology is only an afterthought, assuming that the engineering involved is at least adequate. However, innovative tech is a bit special. After all, it’s harder to bring true disruption with stuff that has been around for a very long time.

So, what does it mean for your startup to create technical defensibility? I like to explain it with a thought experiment. Let’s imagine that you, as a startup, are facing an obnoxiously rich adversary. As soon as your company starts making any waves they decide that whatever gizmo your startup is making is the best thing since sliced bread. However, they don’t want to buy anything from you. They are dead set on conquering (not merely winning) that market, no matter what. Uh-oh.

Think of a behemoth like Google, Facebook, or Amazon. You can disregard the usual objections about minimal market sizes, and even whether the whole thing would make commercial sense (after all, that is out of scope for technical due diligence). Imagine it’s personal, with a slightly desperate and arbitrary twist. I like this quote from a recent interview with an Amazonian to set the mood:

What’s an example of a division that AWS subsidizes particularly heavily? Prime Video, for one. Jeff loves Prime Video because it gives him access to the social scene in LA and New York. He’s newly divorced and the richest man in the world. Prime Video is a loss leader for Jeff’s sex life.

This is the kind of force I want you to imagine is trying to crush you.

Your startup’s tech defensibility is the measure of how much your technology would help it survive or win such a contest. What are the rules of engagement?

They have more money. They will happily set it on fire if they think it may give them an edge. Assume they will overspend you when it comes to hardware, computing power, marketing, everything. I said obnoxiously rich, right?

They have a better and bigger engineering team. This is a corollary to the above — it’s much easier to create a great engineering team if you have a generous budget for salaries and benefits. However, it merits some additional explanation. When it comes to generic job functions (Java/Rust/Haskell/Go programmers, data scientists, ML researchers, DevOps, scrum masters — you name it), assume they will get more experienced people, and more of them. You can count, however, on talented individuals that you already have on your team. So, if your co-founder invented the fancy gizmo algorithm — that’s something you can use.

They are better at reimplementing anything that’s available in the scientific literature. Implementing an algorithm from a scientific paper may be far from trivial, but it’s essentially a matter of craftsmanship, and they have more and better scientists. Not to mention that they can simply hire the authors of the paper (unless you have already hired them).

You have some head start. Let’s be generous and assume that they start competing in earnest only when you get your first serious media mentions. However, remember that they have a bigger, better team (am I repeating myself?), so they might be able to work faster. You’ve got a little bit more time, but time is not on your side.

The situation as I present it looks bleak. No one said it would be easy! Creating meaningful technical defensibility is extremely challenging. However, some options come to my mind (and there’s probably tons I didn’t think of):

Not all processes scale linearly with the number of people involved. For example, Google may have the best ML researchers in the world, and they have plenty of them. However, if they have to learn about a whole new domain, there’s no magic bullet, they still have to put in the time. You get additional points if the domain is esoteric, boring, or otherwise unattractive for most people.

Sometimes, there just isn’t a lot of prior art in some subjects. If you have discovered such a problem, and it’s a problem whose solution may be lucrative, that’s amazing. Usually, when I am researching some technology and cannot find enough relevant papers on Google Scholar, it means that I am searching wrong. In the few situations when that was not the case I tend to get very excited.

Data very often is the key to an advantage, especially when it comes to AI-based products. If you can have a monopoly on the relevant data, or you can create a virtuous circle in which a better product creates more data, which leads to an even better product, etc. then you are in business.

Some processes are hard to accelerate. For example, if you are developing a new drug and want to prove there are no long term side effects, waiting is your only option. Similarly, regulatory certification may be extremely difficult to expedite. If you are the first to get it, you can give you an advantage.

Many problems get exponentially harder as you make progress. Only when you think you are 90% of the way there, you realize that the remaining “10%” will take at least as much time as you have already invested. This kind of surprise may repeat, more than once. In my experience, this is especially common when someone is trying to transfer technology from an academic setting to an industrial one. What is crucial is that, almost by definition, this situation is unknowable at the start, so it’s almost impossible to budget for it.

As annoying as this may sound from the point of view of someone with an engineering mindset, who is trying to solve a problem, being in such a situation is a huge advantage for you. The fact that your rich competitor has more people may mean that they will realize the hurdle’s existence more quickly than you had, but not by an order of magnitude more quickly. The important question here is: how “deep” is your problem? Is it going to be completely solved after just a few “oh crap we are not there yet” cycles? Or are you going to keep making meaningful improvements indefinitely? From a defensibility point of view, working on a problem that keeps getting exponentially harder indefinitely is much better. By the time you get where you were when the rich competitor started, you will have already created some additional distance.

Defensibility is not a binary option (as in “defensible” or “not defensible”) when it comes to technology. Also, technology may be only one of many factors that are contributing to it. My thought experiment is a (negatively) idealized situation, with the scale tipped against you. It’s not that every startup gets in the way of a billionaire getting laid, after all.

Finally, in the real world, we don’t always care about “winning.” On occasions, all you need to do is make it a better option for the obnoxiously rich adversary to buy you out instead of competing. Sometimes it’s not about outrunning the bear, but about creating a situation in which the bear wants to acquire your company at a good valuation, and have you join his organization as a VP.

Aspiring to be a gentleman and a scholar. Ex-Googler. Trained to be a philosopher. Interested in Deep Learning, scalability and startups.