From scratch. Is it worth using a new framework for automation?

Having worked in three companies as QA automation engineer and communicating with colleagues from other teams and companies, I always meet people who are facing a dilemma - is it worth switching to a new automation framework, because the old framework is difficult to use, expand or support? What is more expensive - to spend resources on refactoring and training employees to the old framework or to use a new one? When will the new framework bring the first results? 

In this talk, we will try to consider the main points that should be taken into account in the decision-making process, calculate risks and possible solutions.