@futurebird

Around 15 years ago, a small branch of the supermarket I normally shopped from (they did deliveries for online orders) opened a few minutes walk from my house. I was interested that things I routinely bought online started appearing in the shop. A good use of data, I thought: they know where I live and what I’m buying, making sure that those things are available to buy locally will increase their sales when I don’t need enough to do an online order.

About five years later I happened to be talking to the person who ran their loyalty-card scheme. I recounted this and told him I was impressed. It turns out, they didn’t do this at all. They just put things in new shops with a high churn early on and kept the things that people nearby bought. He had been trying to persuade them to do the thing I thought they were doing for many years but they didn’t really have enough data to make it work well (in particular, people often stockpile varying amounts of non-perishables, so it’s hard to build good correlations between ‘people in this area buy X online’ and ‘people in this area would buy X from a local shop if it were stocked’ (keeping in mind that the prices in the smaller branches were slightly higher because they had less inventory space).

So these companies are already unable to get useful data out of their recorded purchasing.

At that time, there was apparently internal disagreement between the people who wanted to just add a revenue source selling the data and the people who were concerned that selling the data about their customers’ purchases would empower the competition.