Most NCEs fail during drug development, either because they have unacceptable toxicity (>25%) or because they simply do not have the intended efficacy (>50%). In general, this leads to the very low succes rate of only ~10% of drugs that enter the clinical development stage, to reach the market. This high attrition rate of ~90% is a huge risk factor for pharmaceutical companies. Drug development is therefore not only an industry of high costs, but also one of high risks. A succesful drug therefore does not only need to cover its own costs, but also needs to make up for all the failures of those that did not reach the market successfully (not even mentioning the chance a drug is withdrawn from the market after approval in Phase IV). 

It becomes therefore of crucial importance to obtain as early as possible the crucial information to know whether a drug can be succesful or not. Failing a drug in phase I clinical stages is significantly less costly than failing in late phase III, once hundreds of millions extra have been spend. Phase I in clinical drug development therefore becomes more important to predict the chances for the other clinical phases, for example with data-rich studies that cover many biomarkers and various clinical endpoints. All to know whether to "kill your darling as early as possible".