Retail at the dawn of the twentieth century was predominantly corner stores selling basics to the local communities. Customers and shopkeepers knew each other. Even though choice was limited, shoppers mostly found what they needed because the shopkeeper knew what his shoppers wanted. Retail started with personalization and collaboration as core tenets.
Fast forward a few decades. The 1940’s saw automobiles and refrigeration become mainstream enabling shoppers to go beyond their neighborhood to buy and store goods longer. This saw the rise of department stores that carried multiple products and assortments under one roof, which turned shopping into an experience. Individual interaction with customers started declining. And the next 50 years saw an explosion of retail dominated by suburban malls, big box stores, and wholesalers. Aided in part by manufacturers’ mass production strategies, retailers were able to offer a wide range of assortments across all categories under one roof. They decided what products and services to offer, and marketing, through catalogs and mass advertising, enabled retailers to reach far and wide without having to actually know their customer base. Personalization fell by the wayside.
This dynamic started to shift in the 1990’s as the digital channel emerged and customers were easily able to discover, compare and contrast the retail offerings through the web. This trend accelerated in the early 2000’s with smart phones and mobile devices giving shoppers unprecedented access to information. With the ability to influence assortments through social networking sites, product reviews, and direct feedback, today’s shoppers have the power of commerce in their hands and have fundamentally changed the retail business model, as substantiated by this Gartner report which posits that “by 2017, at least eight of the world’s largest multichannel retailers will have incorporated customer collaboration “crowdstorming” platforms for innovation into their business model.” In other words, personalization and customization are once again driving retail planning. The question haunting retailers though is “how?” How can they possibly know their customers in today’s age and create personalized offers?
The answer ironically comes from the customers themselves. The very technology that has given them immense shopping power is also creating an incredible amount of customer behavior data. From socio-demographic details and loyalty programs, to online activity and shopping journeys, this information tells the retailers exactly what they need to create personalized offers. But this data is large, complex and relentless, and traditional approaches to understanding this data are inadequate. This is the world of Big Data!
The term “Big Data” rightfully conjures up images of unwieldy volumes of data. While important, there are three other “V’s” that need considering when dealing with big data. The first is Velocity: the speed with which data is generated: every second there is a new transaction or a new tweet. The second is Variety: in the past, just transaction data was enough to plan assortments. Now social media data, beacon generated data, etc. need to be also used. The third is Veracity: determining which data is actually useful and not just interesting. Clearly gleaning information from this big data is terribly hard, but retailers have no choice: they have to find that “insight needle” in this “big data haystack” to stay relevant to their customers. But how?
As IDC analyst Greg Girard in a recent perspective suggests, machine learning algorithms could provide an answer. These algorithms explore data patterns, learn from them, predict the future and in a properly architected solution, can evolve as new data arrives. Says Girard, “machine learning surpasses descriptive analytics (about what has happened) and extends predictive analytics (about what is likely to happen), with a third, discovery analytics (the “why” of what happened). Machine learning brings new capabilities to prescriptive analytics (what action should happen).”
To be fair, machine learning algorithms are not new and have been part of research in artificial intelligence for decades. While their potential was always understood, only in recent times has their promise been realized, thanks to cloud computing. Combining platform as a service offering with next-generation programming languages and paradigms, machine learning algorithms contribute to solutions that experiment continuously, are “always on” and deliver fast results. So what process must the retailers follow to leverage these cutting edge techniques and deliver personalization?
The process must start with the building of customer personas that represent the majority of the shopping experience as evidenced in the POS data. A k-means clustering algorithm, using similar purchasing behavior as a criterion, may help with this step. Assortments and targeted offers must then be planned for these personas, where perhaps an artificial neural network algorithm rates how attractive these assortments are likely to be. An association rule learning algorithm could determine product affinities, while genetic algorithms might recommend new products that are likely to perform well. Plans turn into action truly when a customer’s shopping journey begins and the retailer has to map that shopper to a persona to ensure they have a personalized experience. For example, Jane opts into a store’s wireless network; beacon technology identifies her and realizes that she has searched for a specific item on the retailer’s website recently. Jane’s smart phone receives an alert that the item is on Aisle 7 and knowing that she loves a sale, the system offers Jane a one-time promotion. This example can be played over and over again, across all customer interactions. And when customer responses vary from plan, these machine learning based technologies can prescribe a corrective action, in real time. Thanks to these new technologies, retailers can now ensure that their customers and their needs are continuously at the center of both their planning and execution processes.
Retailers must feel a sense of déjà vu. A century on, the once personalized conversation between retailer and customer is back again by means of cloud computing, machine learning and smart devices. So yes, this is a “back to the future” moment for retailers: only those who grab it will ensure a loyal customer base and thus secure a profitable future.