The best data begins from the null hypothesis. We assume, "there is no difference" or "our intervention had no effect." We then look for evidence that we can reject that hypothesis. If you "look for evidence in the data" you are not using data but you are cherry picking data.