The good project manager will still do all the same planning up front, to include creating cost estimates and PERT charts, but everyone involved has to accept that discovering unknown unknowns can—no, WILL—radically change the estimates during project execution. And rarely are those surprises in the direction of requiring less time or money.
Consider a historical example where schedule was the overriding priority: the Manhattan Project to build the atomic bomb.
Planners knew early on they needed a certain amount of fissile material to make bombs. But they didn’t know what technical and engineering approach had the fewest unknown unknown problems. So they considered no less than five approaches and launched at least four major industrial efforts to produce that material:
1. Uranium isotope separation/enrichment through particle accelerators. The Calutrons at Oak Ridge created the highly enriched uranium for the Hiroshima bomb. But after the war they eventually shut them down because they weren’t as cost-efficient as other mechanisms.
2. Uranium isotope separation/enrichment through barrier separation. These also contributed to the highly enriched uranium for the Hiroshima bomb and kept running for many years, but eventually this approach was abandoned as also not cost-effective as other mechanisms.
3. Uranium isotope separation/enrichment through mechanical centrifuges. Early planners discarded this approach because they calculated the development time was not competitive and therefore the resources should be devoted to the other approaches. But after the war the centrifuges were developed and that’s how most uranium enrichment is performed today for both civilian and military programs.
4. Breeding plutonium in graphite-moderated reactors. The B reactor at Hanford created the plutonium for the Nagasaki bomb and this process turned out to be relatively cost effective so it continued through the Cold War.
5. Breeding plutonium in heavy water-moderated reactors. The CANDU reactors in Canada evolved into power generators, but you don’t hear much about them these days.
You don’t hear much about rogue states trying to build Calutrons, barrier separation plants, or CANDU reactors. That’s because—in hindsight—everyone knows the graphite-moderated reactors and centrifuges are the most cost effective ways of making fissile material.
But with a war on, the planners justified the massive industrial investments in four different approaches just so they could get one of them done to stop the killing. (Did you know that the Manhattan Project built more floor space during the war than the entire US automobile industry had in their factories?)
80 years from now everyone will have learned what approaches work and don’t work (at all or cost effectively) for self driving cars. But in the meantime, Tesla is obviously not willing to throw that kind of resources at multiple parallel approaches just to save a few months or years.