But it is inevitable to have a lot of unsuccessful research projects before one breakthrough. Similar to startups where most fail. Those "failures" add up to the cost pretty quickly, and you cannot avoid them.