Applications of predictive analytics above the 85th parallel

Senior Content Marketing Manager, Communications Sector, IBM Analytics

Last year, I was lucky enough to interview the one and only Santa Claus during some deep diving research into holiday marketing and seasonal retail trends. It was a fascinating look into what had always been a mysterious, if not fabled, operation at the northern most reaches of our lovely little planet.

As I started to reflect on my exchange with jolly old St. Nick, I began to realize that Santa was primarily applying analytics for big data exploration and operations analysis. It seemed to me like something was missing. And so, it was time to reach out to him again. I knew I had to act quickly before he was ready to suit up, fly out and get his Christmas on. I put my letter in the mail (along with a list of presents I’m hoping he brings me) and soon after I was lucky enough to receive his reply.

Naturally, it arrived in an overnight Red-Ex envelope that was dropped down my chimney. Although I admire his interest in designing a culture and brand that is authentically one, I’m not sure how Fed-Ex’s CMO might feel about it

But, I digress. Please read on if you still believe:

When we last spoke, you were primarily looking at historical holiday data and real-time Christmas consumer, operations and sensor data. Has that changed in the past year?

OH HO HO, absolutely! Much like little Timmy Johnston’s behavior has gone from naughty to nice, my use of consumer data has gone from reactive to predictive. I’d like to get to the point that my present matching prediction engine, code named  YULELOG, is so accurate that the kids don’t even need to worry about submitting letters anymore. It’s not that I don’t like the letters, I do. However, receiving tens of millions of regular mail letters, emails and web forms adds to the complexity of our operations at the North Pole and it’s a business expense that I think we can eventually work to minimize.

 I would much rather have all the little boys and girls spend that time doing something nice for Mommy or Daddy which, of course, would decrease the ratings of their Negative Behavioral Sentiment Index or “Pout Scores,” as we call them.

By combining all available data, from both above and below the 85th parallel, and performing some highly accurate look-alike models, we can analyze what toys are trending and work to accurately match those toys to all the little girls and boys. This gives us a substantial head start on our annual work as most of the letters we receive are typically mailed or submitted within a few weeks or even days of Christmas. That creates a massive rush in our operation that we’re hoping to phase out in the years to come. However, for tradition's sake, we will always welcome the letters and, once our predictive models prove to be accurate enough, we might even save the letters to be opened and reviewed during the off-holiday seasons. Those letters are still a valuable source of data that can be utilized when the time is right!

In short, I’m working closely with my Chief Elven Officer to build a culture that infuses analytics everywhere. Considering that one night out of the year, I literally need to be everywhere, you can see how adopting a new approach to the onslaught of data, where we can make speed the differentiator, makes a lot of sense for my complex operations.

How do you apply analytics to forecast year-over-year demand for toys, movies, books and other presents?

That’s a good question, Graeme. I think you’ve earned that new two-stage snow-blower I predict you’ll want this winter! The answer is that this has become a new and fascinating part of our ability to leverage all available data so we can have a positive effect not only on our operations but also on the accuracy for the gift wrapped presents that we work so hard to match, personalize and deliver. In fact, what we do way up here is not too dissimilar from the way in which your organization helps movie studios predict the relationship between social signal and box office sales.

We start by gathering a vast amount of internal data from the igloos of actionable customer insights, annual product requests and naughty/nice behavioral data we’ve manually collected over hundreds of years and then we combine it with an even larger amount of external data from various social channels, 3rd party data sources, global retail partners and market research firms.

As you can imagine, this paints a pretty big picture.

At that point, the Elven Analysts in Santa’s Big Data Workshop can look at which variables are the most likely predictors. Once they know what to look for and how to look for it, they can accurately forecast demand for all sorts of items that are typically requested on Christmas. And those data models better be good, for goodness sake! HO HO HO!

This sort of analytical transformation, especially one that gets refined over time,gives us a tremendous head start on the season as we can finally be more right, more often. And we’re not only trying to guard against poor-decision making  and reducing inefficiencies throughout the Pole,  we’re also working to protect against security and privacy risks and, therefore, get  the risk-opportunity equation right.

Speaking of risks, Mrs. Claus is concerned that my eyesight is getting worse as I get older. She suggested we try out a new high intensity LED bulb on Rudolph’s red nose. Even being as old as I am, I’m not sure how I feel about that. Rudy doesn’t seem too excited. But, hey, maybe we need to teach that old reindeer some new tricks!


How are you working to build a smarter holiday merchandising and supply network?

As much as I want you to believe that everything we do is the result of the magic of Christmas, which is partly true, a lot of it boils down to building on recognized best practices in retail to deliver a consistent view across systems, including planning, merchandising, supply chain and delivery, and the ability to analyze the lowest level of information to uncover hidden data patterns and insights. As with my demand forecasting models, we also incorporate social data as a leading indicator of trends enabling my happy little elves to ingest data like I ingest milk and cookies: with pleasure!

In all seriousness, this heightened level of optimization covers several key scenarios that are, in many ways, compelling reasons for someone like me to look at how I can leverage big data to optimize processes that would deliver positive impact to the bottom line. This includes sizing analysis, Christmas list collection, inventory productivity, manufacturing process improvement, localized assortment and, most importantly, global delivery pattern optimization.

The inability to balance assortment effectively, for example, can be challenging and often times caused by lack of visibility to customer demand at the attribute level: size, color, style, brand, batteries, boys/girls, can it fit in a stocking or not, etc. If the elves find themselves not knowing what items they are running low on until it is too late, I can find myself sitting in an empty sleigh faster than you can jingle a bell and that, my friend, simply cannot happen.

It is our belief that preventive actions powered by predictive analytics and a whole lot of hot chocolate can really make the difference.

Finally, if you do bring me that two-stage snow blower I want, can you make sure it has an electric start?

Sure, Graeme, I don’t see why not. I had a well-informed feeling I was right about that. If you keep being a good boy, I’ll even make sure it has heated hand grips.