It appears to me that the OP is making a very serious, fundamental mistake: He keeps talking about "most" startups.
Alas, in shockingly stark terms, as a VC he is quite definitely, necessarily, in very strong terms not looking for "most" startups.
Instead, he is looking for exceptional startups. How exceptional? A few or so each decade.
What is true for most startups is no doubt close to irrelevant for the exceptional startups he is looking for.
Sure, due diligence is not important for lemonade stands, but it was just crucial for, say, the Lockheed SR-71. And for any startup that is using new, advanced, original, powerful, valuable technology, proprietary intellectual property, barrier to entry, technological advantage, due diligence stands to be crucial for, say, estimating the exit value of the company. Sure, a lemonade stand may have traction growing rapidly right away, but it will never have much exit value.
Moreover, don't do due diligence on just the code or the architecture but also on the crucial core technology, e.g., some original applied math. We're talking theorems and proofs here, guys, maybe with some pure math grad school prerequisites missing among nearly all chaired full profs of computer science. Did I mention technological barrier to entry?
Uh, again, we are looking for the exceptional, and that may look different from the "most".
Does anyone have the data to do the math and see what the difference in ROI is between startups in general and startups which a) should have technical due diligence done and b) which have or would have passed tech due diligence? Maybe we could call these something like "tech startups" to distinguish them from the others. :)
It would be good to see how much delta exists between these two different pools of startups. This would be a fairly easy filter for VCs to apply.
In the meanwhile, pretend you are the Johns Hopkins University Applied Physics Lab (JHU/APL) working for the US Navy, and they have explained that for their nuclear armed, submarine launched intercontinental ballistic missiles they have one heck of a navigation problem. They tried inertial navigation but are not pleased with it.
So, some physics guys thought about some satellites sending out signals, the Doppler shift, etc. Some back of the envelope physics was some convincing due diligence. The system was built and deployed as planned, and a receiver on the roof of the lab routinely navigated itself within one foot.
No question. No doubt. Worked great the first time.
Disclosure: Have to track the satellites quite carefully and update the data on their orbits frequently, and the JHU/APL had a group doing that. At one time I was a programmer, on the fast Fourier transform with passive sonar data, in that group and heard the stories.
There are many more examples where some such applied math/physics, etc. worked great. The due diligence was carefully done, and the batting average for the projects was terrific.
Instead of empirical pattern matching, just do due diligence on such projects one at a time. That's what the US DoD has long done.
E.g., look at that truck with some long tubes on the back the US just deployed to South Korea. Each tube has a missile that can get up to altitude like right now and hit an incoming warhead. So, given several warheads and several missiles, how to assign missiles to the warheads? There's some nice applied math there. And don't have to wait for a pattern of nuclear wars to know that the math will work.
Maybe the lack of enough data for pattern matching is the bad news, but, then, the good news is that same lack of data -- we're talking greenfield here, guys!
To heck with the pattern matching: Instead pick a problem, one that with the first good or a much better solution enough customer/users will care enough about to yield spectacular revenue and earnings. Then, use applied math, physics, etc. to get the crucial core of such a good solution. Write the software. Go live. Get publicity. Please the target audience. Smile on the way to the bank. The core tech permits better or first good solutions to challenging problems, gives crucial technological advantage, barrier to entry, proprietary intellectual property, etc.
But, as for the JHU/APL project for the US Navy and many more US DoD projects, have to do the due diligence one project at a time. Sorry 'bout that.
Alas, in shockingly stark terms, as a VC he is quite definitely, necessarily, in very strong terms not looking for "most" startups.
Instead, he is looking for exceptional startups. How exceptional? A few or so each decade.
What is true for most startups is no doubt close to irrelevant for the exceptional startups he is looking for.
Sure, due diligence is not important for lemonade stands, but it was just crucial for, say, the Lockheed SR-71. And for any startup that is using new, advanced, original, powerful, valuable technology, proprietary intellectual property, barrier to entry, technological advantage, due diligence stands to be crucial for, say, estimating the exit value of the company. Sure, a lemonade stand may have traction growing rapidly right away, but it will never have much exit value.
Moreover, don't do due diligence on just the code or the architecture but also on the crucial core technology, e.g., some original applied math. We're talking theorems and proofs here, guys, maybe with some pure math grad school prerequisites missing among nearly all chaired full profs of computer science. Did I mention technological barrier to entry?
Uh, again, we are looking for the exceptional, and that may look different from the "most".