And yes 99.999% of them fail.
That's fine.
Let's say we are Edison, a long time ago, trying to invent the worlds first, working light bulb.
(Very, very approximately), you set up a glass test light bulb (which can be sealed and vacuumated, as necessary), which is ready, to take on lots of test filaments.
You get hold of something like 2,000 test samples. Cotton, carbon coated cotton, aluminium wire, gold wire, etc etc.
Most of them almost burn out straight away, and either give no light, or just light up for half a second and then go pop, and burn out.
Eventually, you come to carbon or Tungsten, 1999 samples later. You suddenly find it glows a really nice, semi-white, yellowish glow and lasts for ages and ages.
You then take the Tungsten (or carbon) filament, and refine/improve it, until it lasts for a very long time before burning out.
Finally, you go to the patent office, and say you have invented the worlds first electric light bulb, then manufacture and sell it. Then accept your fame in the worlds history books.
What you don't/shouldn't do, is try your first sample. E.g. Stone. It doesn't work, even slightly. Then stop, and tell everyone about your wonderful, stone filament light bulbs. Which glow, a really fetching shade of black/invisible light.
tl;dr
Fine, try the invention. If it doesn't work, don't tell people about it. Move on to the next thing. Because you may need to try 2, 10, 100 or even 2,000 things (such as lamp filaments, a long time ago), before you find something useful and sellable/marketable.