There are many sources of errors that people have described in other comments. The proper way to deal with this is a blinded experiment. Preparation: 1. Get t

There are many sources of errors that people have described in other comments. The proper way to deal with this is a blinded experiment. Preparation: 1. Get two identical phones, one that you use, and one that has a dead battery. 2. Fix a set of product-categories. Experiment: Every week, 1. Label the phones A and B. Use a coinflip to decide whether the working one is A or B. Record which it was. 2. Hand the phone to someone else. They exchange the labels for 1 and 2. If heads A=1, if tails A=2, and record the coinflip. 3. Get the phone labelled 1 back. Neither of you know whether it is working or not. 4. Randomly pick a topic from the pre-fixed list and talk about the topic near the phone. 5. Let the other person remove the labels. 6. Pick out the working phone and go about your day. 7. Write down whether you see ads relating to the topic, yes or no. Analysis: Join the records, to see which weeks you used a working phone, and if those corresponded to seeing ads.