A failure is not just a failure; its price depends entirely on when it happens. If a compound is abandoned during early laboratory screening, the loss is measured in weeks of work and modest material costs. If the same compound is abandoned after it has already been tested in human volunteers, the loss includes years of development time, the full cost of manufacturing and running trials, and the wasted participation of patients who took a treatment that did not work.
The comparison is stark in both dimensions. On cost, late failure can be orders of magnitude more expensive than early failure, because every earlier stage has already been paid for. On time, a late failure can set a program back by years, delaying any replacement candidate and postponing help for patients. There is also a human dimension that does not appear on a budget sheet: volunteers in human trials accept real risk, and a failure discovered late means that risk was taken without benefit.
This is why the value of early detection is not merely financial. Moving a no-go decision earlier converts an expensive, slow, and ethically uncomfortable event into a cheap and quick one. The earlier the signal, the more of the funnel's cost is avoided.