Hey Michael! I just wanted to write about this exact same thing, but a quick search popped up your article — really nicely written. One additional thing I’ve observed lately is that they also react and learn quite well (in a data driven way) if you don’t take the bait. For instance, my usual drink is a short cappucino — and most of my star challenges have been around short cappucinos. I got offered the bait to try a Teavana tea (much like the refresher) and I didn’t actually complete that challenge: so that product has never come back.

Another time, I was offered 200 stars for 7 purchases — since they were all short cappucinos : which I get anyway, I did complete that one. The next challenge the app offered was now 200 for 9 (one I did very poorly on, because it coincided with a week of breaking out of my routine). The backoff from there went to 2 purchases for 75 stars — but not on cappuccinos, but on purchases of spinach feta sandwiches — which I purchase occassionally: a $1.95 value to me for spending $8, to build up a habit of purchasing an occassional item more].

I’m fascinated by the algorithm at scale :) and yeah, it’s a brilliant strategy in shaping consumer behaviour. Would love to learn of more examples like this one — if you’ve noticed any.