Big Data & Machine Learning in Retail

Auteur
Marco ter Horst
Datum

The retail sector is in full swing . The retail winners of tomorrow are rapidly implementing technological innovations to offer a personalized customer shopping experience. Big Data and Machine Learning are two such innovations that can provide an important contribution to this new digital experience.

Big Data for retail is bringing together the many different data sources and streams that are generated by existing and new innovative systems. Big Data is not an unknown concept in the retail sector. Financial data has been collected and analyzed for centuries. New to retail is the data generated by the ever-expanding set of technological innovations which made their way into the retail sector in recent years. But also existing systems are increasingly sophisticated. POS systems are no longer just digital cash registers, but interfaces in which data is collected and displayed. Collecting, combining and analyzing these different data streams generated by these systems, offers a whole new set of insights and opportunities.

Machine Learning is letting an algorithm independently learn to generate useful and correct output based on both structured and unstructured input. Machine Learning was made widely available in 2015 through cloud providers like Amazon and Microsoft. For the retail sector Machine Learning is potentially a real game changer. Where analysis of Big Data in retail can provide insights into customer behavior in the past, Machine Learning can make predictions about customer behavior in the future. These predictions will have a big impact on how retailers do business in the future. Retailers can tailor the customer experience based on these predictions, but they can e.g. also tune orders with suppliers based on such predictions.

The Cloud provides us with technical capabilities out-of-the-box and instantly available to everyone. Standard methods to easily collect, store, process and analyze large volumes of different data, brings new opportunities to retailers. Today Machine Learning also available as a service in the cloud, making it much easier to be used for different applications. This makes the cloud essential for the use of Big Data and Machine Learning for the retail sector. Technologies that were previously only available after months of planning and design. These features make the cloud eminently well suited to quickly and relatively cheaply make new innovations reality.

The retail sector makes great strides in integrating new technologies and innovations into their online and offline shops. For the online shops things like customer profiling and analytics are becoming more common and mainstream. This is not yet true for offline shops. However the latest innovations in the field of customer motion tracking, via eg. Microsoft Kinect or Ultra HD cameras and item tracking via eg. RFID chips, do allow for the same level of profiling and analytics. Currently these technologies are only used sporadically. And when they are used, the generated data is hardly used at all. A current use case for a Microsoft Kinect in a store may be: Automated recognition of customer movement in a store. Signaling a sales employee when a customer stands in front of a shelf for a certain prolonged period.

A very powerful sales tool, but the real added value lies in the large volumes of data generated by these technologies. And especially in the analysis of the combined data of several different technologies. And then eventually even the integration of online and offline customer profiling and journey to enable a true omni-channel customer experience.

Use Case – Fashion – Big Data

A Fashion Brand shop has an Ultra HD camera system installed throughout the store. This system can recognize people, their gender, assess their age and follow customers through the shop. Additionally, all items in the shop are fitted with RFID chips. At all entrances to the shop, at each dressing room, as well as on each separate shelf RFID antennas are installed in order to record the movement of articles. Also, the customer reward card of this Fashion Brand has an RFID chip.

The POS system registers which product is sold. And by using a customer reward card, a customer is eventually personally identifiable. These three systems generate large volumes of structured and unstructured data. This data is sent to the Microsoft Azure cloud. Via the data stream service Azure Stream Analytics. The data is edited and stored in a usable format in real-time. Making this data available for analysis immediately. Both per data stream, as well as in combinations. By combining data, it is possible not only to analyze how customers and individual articles move through a store separate from each other, but also to analyze how customers and items move together through a store.

Combining the data makes it possible to say something about which type of customers, men or women, in any age group, show interest in a particular item. Simply because it’s taken out of a shelf and into the dressing room. This type of behavior analysis can provide valuable insights on eg. the popularity of certain products and product groups. Where financial data only shows how often a product is sold. Behavioral analysis will give insights into how popular a product compared to another article.

Finally customers and products pass the POS system. Here a customer is personally identified by a customer reward card or credit/debit card. Completing the customers profile, including his offline shopping behavior. This profile is very valuable to the omni-channel proposition of the retailer. The data which is collected and analyzed offline, can be used for a personalized online customer experience.

Use Case – Fashion – Machine Learning

In the above Big Data use case we collect and analyze data to make the omni-channel customer experience better and more personalized, based on the past. Machine Learning is used to make the customer experience even better and more personal, by making predictions about the behavior of the customer in the future.

Following a broad rollout of Ultra HD Cameras and RFID systems over 100 shops of Fashion Brand and several months collecting data, the Microsoft Azure cloud has created a large set of relevant data. Azure Machine Learning uses this data to make precise shopping suggestions for the online customers. The Fashion Brand entices customers to shop online, using a personal or social account. Thus, all the necessary characteristics such as gender and age, but also behavioral patterns of the past, readily available. Azure Machine Learning combines this data with personal data and of similar groups, to generate real time product suggestions relevant to the customer. The Fashion Brand knows exactly what a customer wants to buy, even before the customer knows it himself!

Tags

Machine Learning