I believe we are in the midst of a radical transformation, not just in retailing, but also in how offers are designed, produced and delivered. In Part I of this series, I discussed why retailers, manufacturers and distributors are struggling to adjust to the new demands of the omni-channel consumer. In Part II, I described key activities that comprise the end-to-end Delight process. I mentioned that this ideal process may seem like science fiction to most companies. In this final post of the series, I highlight three reasons why I believe learning technologies will make this future not only possible, but also imminent: First, there are breakthrough innovations in the underlying learning algorithms. Second, there is an abundance of data that is the fuel that these engines need to run. And third, the Cloud is paving the road to travel quickly from insights to action.
Machine Learning technology has finally come of age
Just a decade ago, experts believed that tasks such as navigating a car through traffic were impossible for computers to ever be able to do. [i] But since then, machine learning has not only given us self-driving cars, but also practical speech recognition, a Jeopardy! champion computer, and a vastly improved understanding of the human genome. [ii]
Practical applications of machine learning are already all around us. We can train our phones to recognize our voices and our photo management software to recognize faces within our images. Facebook’s DeepFace research project is able to recognize faces almost as well as humans. [iii] Researchers at Stanford are predicting that computers may soon be able to characterize our personality better than our spouses. [iv]
Machine learning has clearly arrived in the mainstream. Applications in supply chain and retail are only now beginning to be explored. One familiar example is Netflix’s algorithm to identify what characteristics of movies will appeal to you. [v] There are tremendous opportunities to understand behavior patterns and clues to anticipate shopper needs, understand personal preferences, predict purchase paths, forecast demand, estimate returns, anticipate supply disruptions and much more.
Learning will improve with the abundance of data
An interesting fact about supervised machine learning algorithms is that the quality of decisions they make improves rapidly with the size of the training dataset. Even simpler algorithms can make reasonably good decisions if they have been trained with sufficiently large datasets. On the other hand, unsupervised learning comes from testing hypotheses, experimentation and discerning patterns. Again, performance improves with experience. The good news is that there is an abundant and growing bounty of content (Big Data) on practically anything and everyone.
In 2002, the University of Michigan gave Google’s Larry Page and Sergey Brin an estimate of 1,000 years for scanning the library’s seven million volumes. But in only a few years since then, Google’s Books Library project has digitized and scanned books from many more libraries and sources. We are also creating new content at an ever more accelerating pace. It is widely believed that 90 percent of the world’s data was created in only the last two years. [vi] Cisco estimates that some 25 billion devices will be connected in the Internet of Things (IoT) by 2015, and 50 billion by 2020. [vii] According to Statista [viii] the number of monthly active social network users worldwide reached 1 billion in Q2 2014 just for Facebook alone – that is approximately one-seventh of the world’s population. Furthermore, content old and new is now searchable and discoverable online. Indeed, it is the combination of Machine Learning and Big Data that made IBM’s Watson even possible.
Cloud will enable timely insights and action
Cloud Computing will eliminate latency and make it possible for learned insights and actions to be timely and profitable. Everything and everyone in the commerce network is getting connected via the Cloud. Cloud offers a completely different and better way of doing business. Beyond the obvious cost of ownership benefits, Cloud also enables delivery of reduced process latency, social intelligence, predictive analytics, crowd sourcing, community knowledge and collaborative problem solving. [ix] Entire industries are evolving to leverage the Cloud beyond support services to more strategic use cases. [x]
The commerce network of tomorrow will be smart and low touch. Orders will propagate across the value chain instantly and without touch. Shipments will identify themselves and tell you where they are at all times. Shipment loads will be assembled and refined automatically. You will be able to tell where an order is, what orders are in a load, what loads are on which truck, when certain orders will be unloaded or loaded in a warehouse … and so on.
The problem of “unknown knowns” will be eliminated. Any insights about opportunities or obstacles anywhere in the value chain will be processed and profitable responses will be initiated immediately wherever in the value chain it is needed. Stakeholders will be able to collaborate with anyone in the value chain instantly and their collective decisions will be immediately propagated for execution. A key to achieving agility and profitability at the same time will be to move planning and execution from periodic and sequential processes to continuous and simultaneous activities. [xi]
Timely insights from Big Data are already being used to make smarter offers implemented via Cloud. For example, the Tesco supply chain analytics team has built algorithms that create discount offers for food items nearing the end of their shelf life. These discounts are transmitted to handheld devices in-store, avoiding about 30 million Euros of wasted stock per year. [xii]
And since consumers are fickle, any models for predicting needs and for understanding paths to purchase will have to be self-learning–able to learn faster than the speed of change in order to stay current and valid. Again, machine learning working with Big Data will make that possible.
I do not mean to paint a dry picture where computer algorithms replace most decision making across the supply chain. Instead, I concur with Brynjolfsson and McAfee [xiii], to predict a future of systems that combine human and machine capabilities for far superior performance. Observing a chess contest in 2005, Gary Kasparov, the world champion chess master who famously lost to IBM’s Deep Blue computer in 1997, noted that teams of humans plus machines consistently beat even the most powerful computers acting alone.
I see computers providing intelligent, timely and actionable insights that will help people provide far superior service than either people or computers would be able to do on their own. Computers and smart algorithms will help marketers do sophisticated customer segmentation; [xiv] merchandizers use segmentation and insights to develop optimized assortments, sales and operations planners to manage profitable mixes of offers, and fulfillment people execute profitably. JDA research and development is working to make this vision a reality. Stay tuned for more details and specifics in future posts.
[i] Frank Levy and Richard J. Murnane, The New Division of Labor: How Computers Are Creating the Next Job Market (Princeton, NJ: Princeton University Press, 2004)
[x] Gartner Report: “Industries Aim to Evolve Cloud Computing Beyond Support Functions to More Strategic Uses.” (G00234673)
[xiii] Erik Brynjolfsson and Andrew McAfee, The Second Machine Age: Work, Progress and Prosperity in a time of brilliant technologies (New York, NY: W.W. Norton& Company Ltd.)