Observe, predict, test.
Nobody says *how* you should go about doing the "predict" part, so long as you test carefully afterwards. It happens that in certain fields of science, especially fundamental physics, prediction has become very difficult. And that's why we need a bunch of theoreticians developing wild new theories which we, the experimentalists, will then test. But as I said, this process isn't easy and takes some time. It takes time to develop the theory far enough to make concrete, testable predictions, and it takes time to develop the technology to carry out those tests.
Take string theory. Nobody in particle physics is under any illusion that this has been through the complete scientific process. String theory is not a scientific result, it is merely an intermediary phase in the discovery of something more complete than the existing Standard Model.