"Attempt to control" is not the same thing as proving it. You yourself admitted "you cannot control for any psychological factor". Well, maybe you can, but definitely it is not easy.
Now, you get a lot of data on lots of people running in non-laboratory setting. The people tested actually bought the shoes and we know, if they paid for them, they have invested in them and the idea that the shoes represent.
The idea behind this pair of shoes is "extreme efficiency even at high cost". What you get is you get people that may have been biased by the act of selection based on the idea and then you compare them to other people who chose to buy other pair of shoes.
Let's test this idea. Let's say you are comparing Lamborghini to Fiat with regards to whether the car is safe. Instead of doing laboratory tests, you decide to mine data of all miles ever driven in all cars ever produced. Now, you see that, on average, Lamborghini last less between crashes and then are more frequently totaled.
Is your conclusion that Lamborghinis are less safe than Fiats or maybe your conclusion should be that it is actually difficult to say given the data because there might be other factors at work (for example, Lamborghini buyers may be biased to be drivers that tend to drive faster and less safe?) It might also be that the act of sitting behind wheel of Lambo may cause you to drive less safely.