What we found
Mid-cost missions tend to return excellent scientific impact per dollar. In three of the four science divisions with enough missions to compare, missions costing between $150 million and $1 billion (e.g., Discovery, Explorer, and New Frontiers classes) matched or beat both cheaper and more expensive missions in science per dollar.
Cheaper does not always mean faster science. Large flagship missions take a long time to construct, but many very low-cost missions produce no high-impact science at all, and those that do can take as long as mid-cost missions to get there.
Cheap missions fail far more often. Of missions below $150 million in our study, 30% failed. Among all more expensive missions, 8% did. For the small-mission class as a whole to be as efficient as larger missions, each surviving small mission would need to be about a third more productive on average. They are not.
Competed missions also tend to return excellent scientific impact per dollar. Among mid-cost missions, the most scientifically productive ones have tended to be led by independent scientists who proposed the mission in an open NASA competition, rather than directed from within NASA.
Commercial lunar landers are trailing in scientific publications. The four CLPS missions flown so far have each produced fewer papers than comparable NASA-led science missions as measured from the start of operations.
What our study does not say
Nothing here argues for fully abandoning either small or large missions. But if the goal is to ensure high-impact science more quickly, top-down small missions have historically not provided that outcome.
Small missions remain very much worth doing for other reasons. They train people who go on to lead in the space industry, they prove new technology, and some of them return excellent science. Flagships are worth doing too. They make discoveries once thought impossible to achieve, and they sustain entire communities of scientists for decades. Both ends of the range, and all missions in between, depend on one another.