Ed Porter, VP – Customer Insights and Product Strategy, Pristine Infotech, Inc.
You’ve heard all the buzzwords. Machine learning. Deep Learning. Artificial intelligence. Prediction machines. Prescriptive analytics. Recommendation engines. You keep hearing that Amazon and other Silicon Valley companies are using these tools to steal all your customers. You can even see it in action when you look at your smart phone and see a hotel ad for a city you plan to visit three months from now.
But you can’t wrap your head around what all this is, or what you would do with it even if you did. So here is a simple way to think about “Machine Learning AI”, why it’s important, and some ways you can use it.
There are many definitions of Machine Learning AI. One that we like is this: they are computer applications that teach themselves rules to mimic human insight and tactical thinking. Humans learn through experience, by recognizing patterns and discerning the implications of those patterns. In this case, machines sift through very large data sets in a very comprehensive manner, looking at every combination and every possibility. They are able to detect many more patterns than humans can, and they are able to quantify the implications of those patterns with much more accuracy.
With their knowledge of these patterns, these machines are able to probabilistically predict outcomes. And like humans, as they gain more experience, their predictions become more accurate. And when given decision guidelines and constraints, they can generate recommendations to take suitable action.
These recommendations can be simple, such as “discontinue these 100 redundant and slower moving items”. Or they can be much more complex, such as recommending the best mix of everyday prices, advertised promotions, and personalized offers for a given week or quarter.
Why is this an important topic? The reason is your customers. They now expect to get whatever they want, when they want it, and how they want it. They want to do business with you on their terms, not yours. To meet this challenge, you now need to tailor your product choice, price, and convenience to fit the needs of each of your very diverse customers. And this needs to be done in near real time. Traditional analytic tools can’t help you do this. There are just too many possible decisions to be made. And not enough time to complete the analysis, debate the implications, and then decide on the best actions to take.
Here’s a practical example of how machine learning AI can be used to complete a very complex task that many retailers do every week; create a weekly promotion plan. First, consider all the decisions that have to be made to complete this task; what items to put on sale, what price or type of promotion to run for each, what to put on display, and when (in what week or weeks) to run the promotion. Also think about all the considerations you have; sales and margin budgets to achieve, competition, holiday opportunities, and vendor funding just to name a few. If you run 200 promotions each week, carry 2,000 promotable line groups, and wish to plan for 13 promotion weeks during an upcoming quarter, the number of possibilities is infinite. Not to mention the additional complexity of balancing those financial and competitive considerations.
Machine Learning AI can help you manage this task. It will work within all your guidelines and constraints. It will consider your financial goals and your competitive guidelines. It will make sure holiday opportunities and “must have” items are included. It will adjust quickly to changing market conditions. And it will make recommendations that probabilistically maximize your financial goals.
Machine learning AI can make you smarter, faster, and leaner. It can be applied to any merchandising or marketing activity where guidelines and financial goals can be defined, and will assure that your decisions are “always on strategy”. It combines the business knowledge, the goals, and the tactical thinking of your people, with the super-computing capacity that only a machine can possess.
+MORE
