Everyone knows analytics, machine learning, and artificial intelligence are the future of decision making. But how many of us really know what that means?
People throw around buzzwords like neural networks, big data, and cloud computing without knowing the power these tools bring to an organization. But most importantly, companies have millions of data points, even billions, and have no clue what to do with it. In fact, every company in 2019 has millions of potential data points that could be utilized, but frequently, they don’t make an effort to do so.
I’m here to change that, by telling you a couple stories about how powerful big data, machine learning, and all those other buzzwords can be to your bottom line. After all, what good is a tool if you don’t know how to use it or what it’s for?
In 2013, the University of Cambridge and Stanford University researchers performed a study to see if they could use Facebook likes to feed a computer model and test its ability to judge personality traits better than other humans.
After gathering more than 85,000 volunteers, each participant had to take a 100-question personality quiz. (These participants also had to hand over their Facebook data). Then the researchers got to work having friends and family fill out a 10-question survey judging the volunteer’s personality.
The next move: Those 10 questions were compared to the data on FB, and when the results came back, even the researchers were surprised with how well the model performed: The computer outperformed the average human’s ability to judge a friend or family member when it was fed just 100 Facebooks. What’s more, it only took 10 likes for the algorithm to know the preferences of a persona better than their coworkers. With the average number of likes for the members of the study at 227, and the algorithm only requiring 150 likes to outperform a family member, they proved that the algorithm out-predicts the average person. Cool right?
Now, imagine all of the likes on random topics on Facebook, and what that could mean for marketers. As we collect more and more data points about customers — bringing together data from social, email, and marketing campaigns — we could ultimately predict customer desires better than they could.
So if marketing comes down to putting the right products or services in front of a customer at the right time, this algorithm could mean MAJOR changes for our industry. Not only can data help us get to know our customers’ profiles better, but we can use similar strategies to improve the effectiveness of the systems we have for reaching our customers.
Anyone familiar with IBM’s AI, Watson?
Watson has been used to improve a variety of issues, but the example I love is this one, and you should watch it: That one time Watson made a commercial about a mechanic who fixes an elevator issue before the elevator actually breaks.
What’s the takeaway? Watson just sat behind the scenes, building out patterns of behavior in the data, until one day, it noticed a hitch in the system that was outside of a statistically significant range. What did they do with this information? They sent a mechanic to fix the problem that was going to be an issue in a few days.
Watson recognized a variation in the normal behavior of the elevator and prescribed a solution before it became a problem. Why is that important? Because solving a problem after it occurs is a reactive solution and causes waste.
The applications of this in marketing is quite obvious. Right now, most of us are just pulling together report level data and seeing that this email has x number of opens, y number of clicks, and z number of opt-outs. We may look at this every week or month and won’t notice that something is wrong until it’s probably too late and the next couple of stages of the campaign are already out.
But if we can methodically track that data and notice the changes in behavior, marketing teams can put together prescriptive solutions, too. We could use the performance of all our past campaigns from all of our customers to know when the campaign is performing on track or outside of statistically expected bounds. Then with enough data, we could have the system come up with a prescriptive solution to do something like change the number of days between the email sends or maybe even change the subject line (or hopefully one day, have the understanding to change the body of the message, too). The applications of prescriptive analytics in today’s world with the level of processing are — quite literally — endless.
What makes algorithms so much better than humans? We’ve been running around solving problems for thousands of years right, so what’s the issue?
The problem is this: Humans can only connect three or four data points together to make decisions. Back in the days when we had smaller communities, people only connected deeply with a couple of people a day, so solving problems by connecting three or four data points worked. But now, we have hundreds of interactions a day: people, social media, Google Search, Alexa, DVRs, email, texts, and more. We’re quite literally being overloaded with data to sift through, and there’s just no way for us to do it all on our own efficiently.
Instead of trying to connect millions of customer stories together, we can have the computers do that. That gives us the ability to see the overarching combinations, series, and interactions so that we learn more about customers, and ultimately deliver better offers to them. By using analytics, we can find the real story behind customer behavior and the interactions that matter. And the only thing we have to do is use the raw data that’s out there.