Readers: this blog is set in the future (sometime after the year 2020). Each entry assumes there has been a 5th revolution in the US — the Revenge Revolution. More about the Revenge Revolution and author, How the 5th US Revolution Begins and About the Author. Many entries are formatted as conversations. Characters appear in a number of entries, with many entries building on previous conversations.
Occasionally I do a “sense check” about the likelihood of a Revenge Revolution. Entry #318 is the most recent “sense check.” One more note — sometimes I write about another topic that does not quite fit the theme of the blog. Those comments are available on the page titled “JRD Thoughts and Comments.”
This series is about the coming “Technology Tsunami.” The series addresses what might be involved and some suggestions for mitigating and even capitalizing on the opportunity. Entries #321 and #322 are intended to describe AI in more understandable terms, using personal experiences as examples.
A widespread use of AI today is what is called “profiling.” Ever notice after you’ve searched something on Google, an ad appears for the product? How does the computer know?
This entry discusses how AI was used to create “profiles” and how those profiles were used in a commercial application. The examples in this entry and the next (Entry #322) are “early stage” and intentionally selected to demonstrate: (i) applications that are easy to understand: (ii) AI-based applications have been around for a number of years; (iii) how AI can be used to increase the effectiveness of “gut-feel” profiling.
Of course, not everyone in the flock, or profile group, behaves the same way. But to the user, profiling is not about individuals but about probabilities of member in the group. What percentage of the people in the profile group will behave a certain way? The goal is to create a group, or profile, where there is a high likelihood that members will have a specific desired behavior.
Profiling is not a new idea. Profiles existed for eons before being formalized with computer programs. Further, virtually everyone creates profiles. Most all of us put strangers into categories based on such factors as geographic location, appearance – skin color, hair color, hair style, clothing, etc. – age, education and a host of other criteria. Think back to someone you met, then after you got to know the person much better, said to yourself, “Gee, that person is a lot different from I first imagined.”
As for this entry, the first example seems rather crude by today’s standards. At the time the profiling technique described was considered “state-of-the-art.” Remember an abacus was considered state-of-the-art when introduced.
The time period for this entry is the mid-1980’s, at Buick Motor Division of General Motors, where I’m director of marketing. As described previously, Buick has used AI-programs to improve the accuracy of its sales forecast and to start allowing dealers more discretion when ordering cars. (Reading Entry #320 will provide more context.)
The next logical step to try to continue building market share was helping dealers refine how to order the appropriate number and mix/models of cars. For example, dealers in the Northeast knew smaller cars were preferred, but which ones were likely to sell more rapidly in a dealer’s particular sales area? Same problem with dealers throughout the country.
I do not remember who or how the introduction was made – could have been one of the “crazy phone calls” the staff often accused me of taking – but Buick was introduced to a company called Claritas. At the time Claritas had combined zip code and general demographic data. The results were “clustered” into 40 groups, or profiles. Each profile had general buying information for products ranging from food to wine to vehicles to many other items. Claritas also assigned a descriptive and memorable to each group. Some examples of names of group – “Pools & Patios,” “Furs & Station Wagons,” “Hard Scrabble,” “Down-Home Gentry,” “Blue Blood Estates,” etc.
As I recall Buick was the first auto company to use the Claritas profiling. We introduced the concept at the annual dealer announcement meeting. And then not much happened for several months. Finally I got a call from a dealer who purchased a store in Florida that had gone bankrupt and was in the process of converting to a Buick store.
The call went something like this, “You remember that program you told us about at the announcement meeting? I’ve forgotten the name of the program but do you think it might help me order the cars more effectively for this new store?” We asked for the zip codes he thought most likely to consider shopping buy at his store. Based on the zip codes we suggested a mix of cars he should consider ordering.
About six months later, my wife and I were hosts on an incentive trip for dealers. During cocktail hour one night, the dealer said, “I owe you a drink. You’ve made me a ton of money.” As he told the story, the profiling program had been a major contributor to helping him turn what had been an unprofitable dealership into one that was very profitable. And, yes, I let him buy me a drink…even though drinks were already paid for.
He told his success story to many other Buick dealers and the use of the program became more widespread. What seems like standard marketing procedure now was anything but standard then.
Within a couple of years starting to work with Claritas, Buick developed a variation of an existing car that designed to appeal to a very small audience. Because introducing such a “niche” car in the traditional way would be too expensive – major national tv and print campaign that could eat all the potential profits – we decided on a targeted campaign using the information from Claritas. What was the result of promoting the niche car to selected profiles – “Pools & Patios,” “Blue Blood Estates” and a few others? A very successful, and profitable, introduction.
What’s the status of profiling today? Profiling has migrated from projecting buying patterns based on zip code (5 digit) to neighborhood profiles (9-digit zip codes) to profiles by families to profiles for individuals within the same household. The same philosophy applies – birds of a feather flock together. However, the flock is no longer defined by geography but by attitudes and behavior gathered from information on search engines, websites, on-line purchases and social media platforms. Profiling is still about probabilities – and not individuals – even though the clusters can include specific information about individuals within the group.
What does this migration portend for the future? One of the unintended consequences of profiling seems to be the diminished value of small geographic social groups. When one had more face-to-face interaction with neighbors, it was difficult to simply walk away from people with different opinions. While you might not always agree with your neighbor, one at least tried to be civil because that neighbor would be there the next morning and you might need to rely on him or her for something. Amazon, on-line buying, delivery services, etc. have reduced the reliance for many activities. No longer does longer does one even need to talk face-to-face with neighbors. One can replace face-to-face chats by going on-line and finding a chat room of like-minded people, thereby avoiding having to listen to the neighbor with whom one might disagree.
In the future are we going to continue only to seek others in our profile and therefore become more isolated? Maybe for a few more years…then I’m hopeful the tide will turn. The underlying premise of my blog (www.usrevolution5.com): the US is headed for a 5th revolution sometime after 2020. I’ve labeled revolution as the Revenge Revolution. One societal change that I think will result is a return to neighborhoods. Some groups and communities have maintained active neighborhoods, but far too few. What I’m hoping evolves from the Revenge Revolution is a sense of cohesion among neighbors.
Yes, post Revenge Revolution you’ll still be able to use your smart phone and order on-line. At the same time, people will become more aware of and concerned about others, especially those in their neighborhoods. Maybe naively, this awareness will help neighborhoods begin to have the feel more like the 1950’s – not quite like Wally and the Beaver but a lot closer than today. And, no, in my view fences do not make good neighbors.
(In the next entry, a discussion about how AI-developed personality profiles can be extremely useful in dealing with others. Of course, women have used this approach for centuries. Men are still in the learning phase.)