If you follow a large group of candidate compounds from the start of testing, the group shrinks at every stage. Many fail because they turn out to be toxic, or because the body breaks them down too quickly, or because they simply do not help patients in the trials that matter.
The result is a funnel: a very wide opening of starting candidates and a very narrow exit of approved medicines. This is called attrition, and it is the single most important economic fact about drug discovery. It means the cost of the few successes has to cover the cost of all the failures, which is a large part of why a single approved cancer drug can carry a price tag measured in the billions.
Attrition is not a sign that researchers are careless. It reflects genuine uncertainty: at the moment a compound is chosen, no one can know for certain whether it will work in a human body, because human biology is far more complex than any laboratory model. Reducing that uncertainty earlier — failing faster and more cheaply — is exactly the kind of problem where better use of data becomes valuable.