Blind Date

Background

At work, we discussed the various brands of coke and how they differ in taste. We wondered whether they in fact all taste similar and it's just the branding that influences how we experience our beverages. Being scientists, we decided to run a (non-representative and not-strictly-scientific) double blind trial to confirm or disproof our hypothesis.

Implementation

I hacked together a Flask app (live version, source code), where one can enter a list of participants and a list of objects to test (the different brands of coke, in our case). For each participant, the app then generated a different shuffled list of coke brands.

The app instructed the test technicians (colleagues that didn't want to drink that amount of coffeine) in which sequence they should fill the consecutively numbered cups. We participants then were presented the cups and had to evalute each sample on a feedback sheet (automatic processing of evaluations is not yet included in the app). Finally, after all participants had rated each sample, the app revealed which cup contained which beverage.

Results

We found that big-brand names tended to be rated superiour to discounter products (although participants who disclosed to normally drink one kind of discounter coke positively rated that coke). 'Light' products were close to their regular equivalent, 'zero' products received worse ratings.

In an additional trial, all participants confidently identified which sample was Coca Cola and which was Pepsi Cola. 50% of them got it right.

Lots of small samples still lead to a high dosage of coffeine which is noticeable throughout the day.

Flask is great for quick development of tools.