Let me be upfront: I think A/B testing is a waste of time. I thought A/B testing had been a waste of time for a long time. But here is my reasoning…

For most  A/B tests, I see results come back where B performs over A so we chose B, but the difference is that 60% of responders chose B and 40% chose A. That means choosing B over A is slightly better than a coin toss. Naturally, there is an AI solution to everything, that’s how fashion goes.

However, in this case, I think there is an AI solution. AI allows us to generate A B C D E F G H offers easily and with very little brain work. So it helps break the nexus between offer precision and our human capacity to come up with enough different offers. We are now using AI to generate a multitude of offers. But this presents two key challenges that we have solved:  

1. How do you ensure the offers are not subtle variations of the same thing? Our solution to this has been to use behavioural tools to create offers that appeal to different psychological triggers, such as fear of missing out, social proof, commitment principle, fast thinking, etc. We have another 54 of them.

2. Once you have created all these offers, how do you decide which to send to whom to optimise the response rate? First, you randomise (because you have no idea what an individual customer’s best psychological trigger will be), then you track (recording which ones every individual customer responds to), and finally, you optimise (giving them more of what they like).

Now you no longer have to test 2 offers and blast out the marginally best one. You can create a large number of offers with a different psychological trigger and then match them to your customers. There is a moral and ethical debate to be had here for sure, so stay tuned.