Over the past 25-30 years, I've been working "in tech" on various projects at different companies. Several of the companies failed. I could tell, within a matter of weeks of working at these places, that the companies were doomed. My assessments were 100% right over time. I am not really sure off the top of my head what characteristics of the doomed companies tipped me off.
An individual who knows what he is doing acts a certain way. An individual who doesn't have a clue acts a different way. In the corporate or political context, it involves lying and telling stories and gaslighting people that they aren't working hard enough or whatever. The crux of the problem at the failed companies was they tried to do "too much", that is, they embarked on a big project without really knowing the problems involved. It's like building a house from the roof down.
When people don't know what they're doing that approach seems plausible, because they know certain aspects of the structure of the problem and imagine they can fill in the gaps later because they can hire an expert, or whatever. That doesn't work for fairly obvious reasons.
I briefly worked as a contractor at a company that made radioactive source screening equipment for scrap smelters. That company went from a back of the envelope concept to a full scale system--that didn't work. I got hired "to help out". I started to delve into the code to try to understand it, and started putting together some models of systems that might work but I was let go, because the people running the project thought it only had cosmetic problems and didn't want a mere contractor to work on their "design". A few months later they shut the project down and the company is gone.
Another company I worked at outsourced their core intellectual property and then stupidly agreed to onerous licensing terms. That decision was inexplicable and obvious. Rather than bring the core technology in-house, they hired another contractor to implement the motion controller. At the time, in the late 1990s, there was actually published open-source software for a motion controller system from NIST. The task to make that work was straightforward. A few people, including me, argued to do that. The company declined, then went out of business a few years later.
When I do my own tech projects, I find it's fairly common to run into a gap in my expertise and have to pause. For example, I implemented some drone mapping software a few years ago. It relied on a photographic pattern matching technique that works in open areas, but fails when there are trees. I worked on several algorithms that attempted to work around that issue, but they all failed. In recent years, people came up with brute force algorithms that might potentially work, plus people developed some alternative sensors that work, but they're too expensive for me to try.
Anyway, as an individual, I can keep iterating on projects, or trying things that might have a niche application. Tech companies could do that too, but when they rely on investor money, ironically, they get pushed onto a nonsensical path more often than not. That is, they have to make unrealistic promises and follow that roof-down model. The investors know even less about what they're doing than the managers of these companies. For them it's just a bet that will pay off or not. They rely on a theory that a 1/10 success rate will make up for losses. All that waste chewed up the purchasing power of the dollar.
It's pretty fatiguing to be involved with projects that have no chance of success because there's no organic way for engineers to decide what to do. When there's key missing parts, all the work is arbitrary lists of nonsense made by dumb asses.
Anyway, the way to do things is obvious and natural. Get the basics working, delve into the aspects of a project that aren't understood to determine if something is really feasible. All that effort might take a year with key people prototyping the difficult aspects of a problem. At that point, the potential expenses to develop some new tech gizmo or system would actually be known reasonably well.