The Big Picture- Machine Learning
As my family decorates the Christmas tree following our late-day jaunt for string lights and seasonal sundry, I’m struck by one of the miracles of Christmas: everything we needed was there at a store at the right time and place, and in the right quantity – miraculous indeed if you consider all that happens to make this so. Merchandise availability is no accident, but rather the outcome of a finely coordinated system of analytics, computing, and communications technologies. In other words, this miracle of Christmas is touched with the science of supply chain.
Today, as more and more people use terms like machine learning, cognitive computing, and data lake (evoking promise, and sometimes fatigue), I thought it worth a reminder that the supply chain is perhaps the industrial epicenter of analytics innovation, and has historically been a major force in the advancement of all manner of predictive technology, including machine learning.
In this blog series, I’d like to offer some concrete examples of what new techniques various industries are bringing to analytics. I also hope to shed light on some new opportunities that leaders in supply chain will recognize. I hope you’ll stay tuned for each part.
Seeing Into the Future, Long Ago- Machine Learning
In the beginning, before the explosion of data or learning algorithms, before anyone heard of a computer – sometime around World War I – supply-chain innovators began by introducing new techniques for quantitative analysis, things like ANOVA, and seasonal decomposition. While these methods were invented and loosely defined in scientific circles, it was the thrust of business that turned them into consummate disciplines.
It might surprise some to know that the same time-series business statistics many rely on today (with vastly more horsepower) were being used nearly 100 years ago by appliance manufacturers and department stores to see into the future of demand and shed light on some of the most important product decisions of the time.
Although it wasn’t called “analytics” yet, it was predictive and it really worked – so well that time-series business statistics became the basis of university courses. As it’s been practiced over the ensuing decades, much has been added and improved in time-series forecasting techniques (even Vanguard Software has a secret sauce). But the biggest push of all came with our ability to capture our accumulated know-how on a humming, blinking server. That was when our efforts really gained efficiency and our quality of life rose ever further beyond subsistence.
Finding Data Correlations in the PC Era- Machine Learning
When the PC took over the retail point-of-sale in the 1980s and 90s, supply chain gained a vast new territory in which to practice the new art of data mining – pouring through millions of transactions to find valuable correlations in time, place, people, product, and much more. Note here that even the term warehousing suggests its origin in supply chain. In any case, data warehousing at that time meant: “we don’t really need these transactions anymore and they’re taking up a lot of disk space, but let’s keep them around in the hope that we’ll find something useful in them someday.”
We did, and still do. And while data mining is still a largely manual process, the techniques developed over time have truly delivered, from the detection of purchasing patterns to the optimization of product placements. To complement data mining we began creating and storing more and more of it.
The introduction of the loyalty card is a great example. Customer behavior could now be seen over time, within groups of similar shoppers with similar habits, teasing out important insights such as what kind of promotions were most effective at delivering long-term customers and quantifying that in dollars. Today, much of what was called data mining has matriculated into the more lofty realm of data science, but it was data mining first, and it was driven by supply chain.
And Much More- Machine Learning
Supply chain has driven countless sector-specific methods of optimization, prescriptive and otherwise, in transportation routing, fruit produce aging by temperature variation, RFID, and so much more.
With respect to machine learning algorithms and other, yet unknown forms of Artificial Intelligence (AI), the best is yet to come. So first, pat yourself on the back for being part of supply chain – a long-time catalyst for innovation by pioneering the use of algorithms and data to hit the target spot on, with every possible advantage. It may not be going to the moon, but darn close in scope and complexity, and may actually be more important to mankind.
Properly congratulated, we can now turn our attention to what’s new, and how we can utilize advances from other fields with the same spirit of “hey everyone, look what we just did with our data!” The next post in this series is roughly titled “How algorithms know more than we do,” an explainer piece on how machine learning works and what it can do. Stay tuned.
To learn more about machine learning, feel free to reach out to us today.