I guess, typical scientific experiments, or their output, can roughly be divided into three different classes. When you look at the results of your experiment you can either…
… see that there is a correlation between what you did (altered) and what happened (what / how much changed in your sample). Thus you can show that, say, A influences B in this and that way. Very nice. If no one ever showed that before, you can write a paper about it and/or fly to a conference somewhere and tell your excited colleagues all about it. (If somebody already showed the same thing, you might still write a paper about it. Unfortunately, people do that all the time.)
… see that there is no correlation between what you did and what happened. That’s okay as well, you write a paper about it and so on, and in case you’re a PhD student you’ve got some more content for your thesis. Of course, usually you’d prefer the outcome mentioned above. Just because “Hey, I had a great idea about how things might work; so to check this I set up an experiment in this and that way; if you want to do the same experiment, you’ve got to take special care about this and that; now for sample preparation I did this and that, and some other things as well; than I indeed did measure the sample property B in question while carefully changing A in a controlled and complicated manner… and see! I can clearly show that B… umm, well, it doesn’t give the slightest damn about A.” is sort of anticlimactic, isn’t it? But still, it’s a result, and you learned something about how things work.
… or you can look at your measurement and realize that all you can see is clearly nothing. You get no idea if fiddling with A does anything to B or not, because the g#!§$%! sample just does as it pleases anyway. Or it just doesn’t work at all, even so it did yesterday and might do so tomorrow, or never again.
Someone said the third option happens 99 out of 100 times in science.
He must have been an overoptimistic lunatic.
