Software is boring now. Or, at least, that’s part of the argument that Sam Lessin made which rocked the world of venture capital.
The world of software has become tame. Some of the largest companies in the world are software companies. Marc Andreessen’s prediction that “software will eat the world” has come true, and it is mostly digested.
We know the kind of metrics that demonstrate promise in terms of customer acquisition costs, growth, etc. We even have the truism of the core three factors in an early-stage software startup (it goes something like team, team, and team).
Deep tech is different
Hedge funds, private equity, and other players from traditional finance are pouring into software. However, deep tech has been more insulated. Part of the reason is obvious: “deep tech” requires more technical expertise. Although definitions differ, most people classify technologies like AI/robotics, synthetic biology, and advanced materials in deep tech.
However, there’s another factor in deep tech: it’s not all about the team.
Not all about the team
What do I mean by that? One of the core reasons why “move fast and break things” was such a common approach in software—with constant iterations and potentially numerous pivots—was because software was fast. Delivery of new versions was instant. Reaching customers was also instant, especially once digital advertising and promotion matured.
The friction between imagine, build, and deploy was extremely low, and it’s far better to get imperfect versions out there to hone in on the exact product-market fit through direct experience in the market. After all, all the research in the world will not do better than actual customer feedback and reactions.
This problem isn’t unique or new
For many software founders, it’s mind-boggling why devices and even software in healthcare is so clunky and unfriendly. In the “open field” of the consumer internet, or even B2B SaaS, such ugly, hard-to-use, and imperfect products would have been outcompeted and been made extinct (unless they also adapted).
For those who work in healthcare, the answer is obvious. There’s much more friction. Regulatory approvals or clearance, arcane and difficult requirements around data usage and privacy, and simply the need to work with critical, cannot-be-circumvented legacy systems—all of these contribute to a much slower development cycle.
However, more importantly, it also means that you have fewer shots to get it right. You can’t forever beta test with a hospital with constant (literal) life and death situations. HIPAA is not quite so forgiving about moving fast and breaking things. The FDA frowns upon randomly changing one’s approved device on the fly in significant ways.
You may only have one shot to get the broad outlines of your product out there. From there, you may only be able to do more minor iterations until the next major product you unveil. After all of that, the new major product might then require you to start from scratch in the adoption cycle, if it’s enough of a change.
This is, in fact, before our two to three decades of software mania, this is how things used to work.
Deep tech is not software
You can’t overnight change up a robot for an entirely different application (no matter what freshly-minted PhDs who have big ideas about six-axis arms think). You can’t retool a wet lab process to just suddenly “pivot” to a different market without significant cost. Even in AI, which is “closest” to traditional software, it’s difficult to take a dataset that is grounded in the physical world (healthcare, robotics, etc.) and suddenly change it to target something else.
Now, I’ll let you in on a dirty secret in software startups from someone who’s seen them firsthand on the ground: it isn’t that easy to pivot even there. Caffeine-fueled, overnight complete pivots for anything but the earliest-stage software startups (who seem to have no technical debt and can instantly rewrite everything) are mainly the province of dramatization that is about as realistic as the 1990s movie Hackers.
That being said, it is much easier than any area within deep tech with branches that reach into the world of atoms (electrical, mechanical, chemical, or even biological). And what is much, much easier and is a true, massive advantage is iteration and refinement. Good luck delivering iterations of new molecules to your customer as a synthetic biology startup three to five times every day.
Shoot your shot—with forethought
Given that a deep tech startup can’t constantly refine and iterate, you’ll need to put a lot more thought into your product-market fit, your business model, and everything else ahead of time. You’ll need to do enough work to really understand your market (and ideally partner with investors and advisors that also know your market).
In software, that level of theory-crafting is a waste of time. You might as well build, deploy, and see what happens. The data you’ll gather will be far more valuable, and if you iterate faster with contact with the real market, you’ll be able to out-evolve your competition.
In deep tech, you can’t do that. On the bright side, neither can your competitors. Will this mean that your product will ultimately be “imperfect” relative to some imaginary world where you could iterate like software? Yes, of course, but we have to deal with the world as it is, versus how we might wish it works.
That being said, one rewarding part of doing a deep tech startup—and one that typically deals with atoms vs bits—is when you do find the product-market fit, there is no real question about its impact. It’s always hard to perfectly quantify or visualize the impact of some SaaS tool or purely software product (even if the impact is huge). It’s not hard to understand how most deep tech startups can impact productivity since they mostly operate in the real world.