The general elections are a lot like an A/B test. But…if life were an AI-driven self-optimizing campaign, everyone would end up happy. (The country as a whole? Not so much.)
Maybe I’m overly obsessed with marketing, but I can’t shake the thought that the presidential elections are a lot like an A/B test. This is an A/B test in which the population is trying to choose a president, and either A or B is going to be the winner. As we have seen during the past two weeks, in this system even a small tilt in favor of A or B compels the complete population to concede to the winner.
General Election: Aggregated Results
As we know, Trump didn’t win all the states – think about California, for example. California is a Democratic state, and the people of California choose Clinton as their winner, but their choice is subsumed in the general election. If we drill deeper to the county level, in some states certain counties differ from the generalized outcome. For instance, although Florida choose Trump, there are some areas in Florida who would have preferred Clinton as president. So the price of generalization is being paid also on a more granular level.
In an AI-based marketing world, we’d want every county to have its own winner – some would get candidate A as president, and others would get candidate B. Imagine how delighted everyone would be if all the A (red) areas had their choice as president, and all the B (blue) areas had theirs. But as the elections go, there is only one winner for the entire country, and this winner is chosen according to the aggregated results from all the different states.
AI-Based Consumer Marketing
Letting every county – and even every citizen – have its own winner is, essentially, what AI-based customer marketing does. AI-based customer marketing provides each micro-segment of customers with the offer they prefer out of all the available offers. It is built exactly to avoid paying the price of generalization.
Of course, this might not be the best solution for Western democracies. But it does wonders for communicating with customers. One of the biggest problems marketers face is designating the optimal offers for selected customer groups (called segments or micro-segments).
One From Column A…
How can a marketer know which offer will elicit a more favorable response? This is what A/B tests are for – but these tests are labor intensive and require ongoing measurements, recalibrations, more tests, more measurements, ad infinitum. For marketers deploying numerous concurrent campaigns for tens and sometimes hundreds of customer segments, this is simply impractical.
Consider that the US population would have been faced not with a choice between Clinton and Trump, but with two different marketing offers, sent weekly. Let’s call these offers “Trump” and “Clinton.” Based on the election results, a marketer wouldn’t want to send a Trump offer to Californians, or a Clinton offer to the people of Florida. However, within these states there are some segments who would be glad to receive such an offer.
Self-optimizing marketing campaign: Every subgroup gets exactly what it wants.
Algorithms that continuously scan the preferences of micro segments and the way they react to previous offers can adjust the offers on an ongoing basis, making sure that every sub-segment receives the most appealing offer, which will generate the optimal satisfaction or reaction.
This is exactly what self-optimizing campaigns do. Based on collected data they are able to identify the wants and needs of customer subgroups, even when those change overtime, and match them with the most relevant offer.
If only politics could be that simple.