Building a car probably costs more in production than in design, so you could afford to have several design projects. But building software is nearly nothing but design... I couldn't imagine building 3 DB backend with 1/3 the programmers workforce each and still making a profitable product.

Sometimes not deciding anything is good though. If instead of basing an app strongly on a specific library/platform/etc., you program something fairly generic and get to decide later, the more power to you.