Take a pendulum, for instance: suspend a weight from a string, give it a push and measure the time it takes to swing from one end to the other – you will soon find out that its period is constant. You can then formulate a theory on what this may depend on – the mass or size of the weight, the length of the string, etc.; do controlled experiments, changing just one variable at a time, while keeping the others constant; and eventually you will find that the period depends on the length of the string only. There’s a consensus that scientific experiments should be done like that, under controlled conditions – and in physics, it’s relatively easy to do that – even complicated experiments such as the LHC are performed according to these principles.