How to Use Machine Learning to Increase Your Loyalty ROI

Jun 15, 2023

When anyone brings Artificial Intelligence into the conversation, it is often met with eye rolls, annoyed grumbling, and hyperbole that ends with the world in the grips of a hostile machine takeover straight out of I, Robot.  While this has absolutely been a hot topic all over the world, it does what most topic du jours do and brings with it a wealth of anxiety and misinformation.  It is this misinformation that does not allow many to utilize AI to its fullest potential.

The most pressing question for many is, “What impact will AI have on my job in the future?”  It is true that the widespread use of AI is leading to unprecedented changes in the workforce. The anxiety is real, especially in the realm of those who are not looking into all the things it is currently a part of, and where it is heading. In so many industries, AI is already utilized as a means of making tasks and projects more efficient. While some jobs are being automated and replaced by machines, AI is also creating new jobs that require human skills such as critical thinking, problem-solving, and creativity.  Steve Flaming, VP of Data & Analytics at Kobie, recognizes the importance of that human element and says,

“we have more tools, we have more data than ever before, but the human remains the critical element in how we’re going to bring it together, how we’re going to orchestrate it, and how we’re going to use it to drive measurable value.”

That measurable value is the critical component of what partnerships like Kobie bring to the table.  Having data scientists and strategists that know your needs and can use the right machine learning models to analyze, interpret and utilize your data and information for your company’s benefit is a game changer.  The best insight comes when Flaming reminds everyone that, while machine learning can do all of these things, “it’s just an algorithm that’s going to give us numbers back, and without [domain expertise] and our strategy team there’s nothing that we’re going to be able to do with it.” On its own, AI and machine learning can tell you the basics.  It can give you the high-level definitions and run the numbers as they sit on the spreadsheet.  What it’s missing is the nuance of understanding causal relationships and the reasoning ability associated with human intelligence.  In other words, without the human element, all you have is data with no relationship.  That relationship is where the magic is.

Where are you on the curve?

When it comes to latching onto new technology, especially with respect to business, you absolutely see a good majority of companies sit back on their heels and wait to see whether this technology is going to be worth their time, and more importantly their money.  Flaming references the Diffusion of Innovation Curve to illustrate this idea and encourages participants to truly think about where they fall when it comes to their brand.  He asserts that those in the Early Majority and Early Adopter sections will be the winners due to the fast-moving nature of this sector. Anywhere else and you could spend a lot of time and money trying to catch up.  It takes more than adoption to win in this race, however.  In this game of technology and numbers your team must include expertise that knows how to layer science, loyalty expertise, and domain expertise together into a winning strategy.  Without that combination, again, all you have are numbers.

Churn, Baby, Churn

In this discussion, the examples of models used are analyzing customer churn with the necessary data points for your program that can make exponential increases in revenue just by looking at the data in the right way.  Machine learning can take in and process vast amounts of data in a more efficient way, allowing for more precise and in-depth answers to complex problems without the lag time.  Flaming pulled data from clients and utilized an algorithm Kobie uses daily that goes into the major touch points within a loyalty program and shows the incremental value from improving them.  With that algorithm, he was able to find the following statistics connected to customer churn:

  • 5 to 10% of customers in these programs are 20 to 40% of their sales.
  • 25 to 40% of members in these programs who have purchased in the past six months will not purchase again.
  • 10 to 20% of VIPs in these programs will either completely churn or they will significantly decline in the next six to nine months.

With all that data, Kobie’s algorithm can generate an output that tells you exactly what will impact the most increase in incremental revenue for your program.  What it is missing is the strategy.  That is always the next piece to the puzzle. The pitfalls many groups fall into is that they look at the data as the only answer.  When that happens, the nuance gets lost.  Understanding that customers tend to churn at 17 months might make you think that 17 months is your timeline.  True expertise lets you know that you have indicators of churn that you have control over happening 5-7 months before that.  If you wait until that 17 month window hits, they are already gone.  Having a model that doesn’t predict churn, but rather the drivers of churn, allows the strategy to come into play that changes everything.

How much bias is too much?

One of the biggest arguments against AI has been the idea of bias.  What is interesting is that bias happens in all scientific endeavors, whether we know it or not.  Flaming said that

“bias is inherent in all statistical processes and our goal as practitioners is to learn the major sources, how much bias we’re comfortable with, and how to minimize it to an acceptable level.”

Bias occurs when your statistical analysis or data collection consistently differs from the truth.  The types of bias Flaming asserts we must consider in loyalty are:

  • Self-Selection Bias
  • Measurement Bias
  • Omitted Variable Bias

We can overcome the biases with three things:

  • Your loyalty data
  • Non-loyalty data
  • The right algorithm

According to Flaming, using this information, the right combination of algorithms, and then the addition of the human element can help you “understand whether you’re within your tolerance and what that means if you actually go too far.”

“Knowledge is knowing that a tomato is a fruit. Wisdom is knowing that it doesn’t belong in a fruit salad.”

Technology and its applications are doing innovative and amazing things in so many different sectors, and it is only going to keep moving forward.  The most important aspect to remember is that what technology lacks are the elements that only humans can bring to the table.  Strategy, reasoning, humanity, and the ability and knowledge to make a great fruit salad.

To view the full session, check out the video below.