Welcome to the first installment of our three-part blog on this topic, which we get asked about frequently!
You’ve made up your mind. As a retailer, you know you want to improve your pricing and promotion capabilities. You’ve been reading about machine-learning science, predictive analytics, price elasticity, demand forecasting and all the other buzzwords out there, including the once taboo “Optimization”. And if you are like most people, there is a good chance you are stuck and can’t tell the true difference between a lot of the various technology claims out there. Don’t feel bad. Marketers have become really good at positioning their products in the best light. And in an era of Google (sorry, Yahoo and Bing) and search engine optimization, everyone wants to be found providing the latest buzzword. They use the same terminology and flood their material with the keywords. So now they are found, but as a buyer, you’re lost in the noise. Even once-trusted analyst reports don’t help. They seem to just list everyone so you think there is no differentiation, or maybe they just key in on the billion-dollar generalist tech companies.
This leaves you in a quandary. Do you simply pick one, going with the lowest cost just to find out that what you purchased wasn’t what you thought or hire some consultants to help you dig under the message? If you focus solely on cost and not overall value and return, be careful, for rarely will you get more than what you pay for. Those who have gone on this journey have learned that not all vendors are the same, hiring consultants and doing nothing else is an adrenaline shot that coming down from can be worse than the high, and the wrong selection wastes valuable capital, energy, and time.
So what do you watch out for?
Don’t be fooled by armchair quarterbacks and classroom theorists. There is a big difference between having science and successfully using science.
Machine-learning, price elasticity algorithms, analytics are nothing new. Those deeply educated in consumer behavioral science, like our Chief Science Officer Jeff Moore, remind me that Machine Learning and other terms date back to the late 50’s. Alan Turing predated the term by about a decade in his 1950 paper, “Computer Machinery and Intelligence”. [http://www.loebner.net/Prizef/TuringArticle.html]. Drawing it up in the classroom is easy, but proving its value in practice is tough.
In a classroom, everything is assumed to be there in perfect order with all variables known. In reality, this is rare. Retail is loaded with data, but it can be incomplete and messy. What’s worse is most don’t even know where the holes are. Algorithms, science, and processes have to be able to account for this. Yes, better data equals better results. We don’t live in a perfect world AND you can’t wait until it is, because it never will be!
Below is a very simple example of something that can trip up science that hasn’t been put to the test.
It’s a typical hockey puck sales curve showing an item’s sales from the first moment it was sold. The question is, what will happen next? Well, a number of things could, depending on the type of item.
The science has to be able to recognize this before it happens so it can fully recommend pricing going forward. Mistaking a technology item for a grill could have serious implications on your outcomes. You don’t want to act on those predictions only to find out afterward that you got it wrong. The issue many novice retail data scientists have is a belief that the solution will work until it’s proven wrong. But YOU don’t want to be the guinea pig who has to learn the hard way where and how it’s wrong. You need science that has been tested, battered, bruised AND strengthened, improved, and hardened. There are plenty of vendors who have either not proven their science or who continue to sell it without investing in it to keep it current, relevant and state-of-the-art. You need the comfort of knowing you’re working with a vendor whose science has been proven in real-world conditions, year after year and continues to innovate.
Check back soon for part two of our three-part discussion.