Eleven years ago, I wrote a post about coffee shop loyalty schemes, and the way that they deviate from the psychological evidence base.

The thing that links all of the common schemes is they are effectively fixed ratio reinforcement schedules. That is, they entice customers to buy more coffee by promising a freebie every X visits. But a wealth of literature from psychology reveals that this isn’t really very effective in getting people to form habits, not least because their motivation to consume drops off immediately after claiming free coffee Y.

A far more effective method of getting people to establish habits is to build a variable ratio reinforcement schedule. As with gambling, this means that the punter / customer never knows when the win / free product is going to materialise. This keeps motivation consistently high.

In practice, what I’m suggesting is that the ratio of visits to free coffees is kept the same (X+1:1), but that the free coffees are dispensed at random.

I still think about this a lot. Eleven years ago, most coffee shops used paper stamp cards, which didn’t really lend themselves to this sort of thing. I proposed using scratch cards as an alternative.

These days, these schemes use apps, so the switch should be much simpler. Indeed, some coffee shops use variable ratio reinforcement within the apps for special promotions: Caffè Nero’s app sometimes has a fairground spinner, and it has crackers at Christmas—both operate on the variable ratio reinforcement principle. I see no good technical reason why the randomised approach couldn’t be used for the core loyalty scheme.

And yet, the fixed ratio ‘stamp card’ approach is ubiquitous. Costa, which had a different system when I wrote that post, has come into line. Even M&S has a version now.

I don’t doubt that there’s a good reason for the lack of adoption of a variable ratio approach, but I’m not sure what it is: I struggle to imagine that large corporations are squeamish about using gambling-like mechanics to increase sales, so I expect that it’s to do with effectiveness. I’d love to see a trial, not least as it might help illuminate the limits of effectiveness of variable ratio approaches.

The image at the top of this post was generated by DALL·E 3.