How is AI and IoT Changing Retail Analytics for Physical Stores?

How is AI and IoT Changing Retail Analytics for Physical Stores?

Physical inventories and bricks-and-mortar stores are not immune to being affected by advances in retail analytics. IoT and AI in retail are opening up new approaches and ways of thinking about physical store space. In many ways these new approaches are converging with ideas of retail analytics previously only applicable to e-commerce. This trend is concerned with performing tests and gathering data in order to optimise customer experiences, offer accurate recommendations and generally be more experiment and data-driven. 

The Problem of Retail Architecture 

From the earliest studies in retail architecture in the 1950’s, there has been an acceptance that the way customers behave within a store space is not clear cut and is subject to many elements which are not always rational or readily and easily measured in retail analytics.  

The store is defined in a shopper’s mind, partly by its functional qualities and partly by an aura of psychological attributes. Whereas the retailer thinks of himself as a merchant concerned with value and quality, there is a wide range of intangibles which also play a critical role in the success or failure of his store.   

Further research has sought to quantify the retail space in a variety of ways and across different disciplines such as psychology, design and architecture. The aim has been to isolate the variables which affect a shopper’s experience in order to be able to tweak and optimise them. One study broke down the elements of retail success into: lighting, air conditioning, washrooms, in-store layout, aisle placement, fixtures as well as a visitor’s feeling of warmth acceptance and ease. Working out correlation with such a number of variables would be difficult even if some of them weren’t so intangible.  

The Streamlining of Retail 

The trajectory for physical retail spaces has been towards rationalisation, streamlining and a cutting of costs wherever possible. Whether logistics or customer service, excesses and inefficiencies have been gradually trimmed with the intention that the savings be passed on to customers with ever-cheaper prices for products. This has been seen in staff shifting their focus away from customer experience, automated checkouts and now to Amazon Go, a concept store which uses sensors to completely eliminate the need for cashiers. 

The Amazon approach is one way to use sensors to augment a modern store, but it’s a very particular approach.  This tendency towards cost-cutting and rationalisation clashes with the research that states that, for most customers, price isn’t the only factor in determining whether or not they have an overall positive retail experience. As we’ll discuss, sensors and machine vision open up a range of other possibilities for modern retail spaces and their associated data-driven retail analytics. 

an open sign hanging on the window to a conventional retail store which is augmented by ai and iot for retail analytics

What is The ZMET Approach in Retail Analytics?

The Zaltman Metaphor Elicitation Technique (ZMET) is a concept in marketing which accounts for some of these difficulties in quantifying intangibles. It is an understanding of consumer decision making that is a combination of rational and irrational, conscious and unconscious. A ZMET approach would involve both asking a consumer their reasoned feedback on a topic as well as getting them to take creative photography of a store, which would then be discussed in an interview. The idea is that this process draws out and captures some of the unconscious and non-rational reasons why customers react a certain way to a retail space.  

Although ZMET is a patented method available to retailers, it is a long and involved process that is not exactly feasible for everyone to implement. Qualitative feedback about physical retail spaces is effectively out of reach for the majority.  

Can E-commerce Influence Physical Store Analytics?

Due to the obvious difference in the operating environment, e-commerce stores have a lot more options available to them in how they present wares to customers and how they work towards optimising experiences. Because customer browsing habits and purchasing history are typically recorded, it is the type of data that can be analysed and used to direct evidence-based changes. 

Because of the availability of browsing data and purchase data, e-commerce store operators have two powerful tools at their disposal, dynamic product recommendations and A/B testing.  

A/B Testing 

A/B testing is a way for online retailer to use accessible, evidence-based methods to act on some of the intangibles that make customers more likely to make a purchase and be satisfied. By utilising high-iteration testing of page elements, products, prices, and other elements, e-commerce platforms can eventually zero in on what works best. In this case there is no need to know the ideal preferred state of affairs ahead of time. Rather, this process of testing and feedback guides the development of the store gradually in the right direction. 

Recommender Systems 

Recommendation systems, as we wrote about in a previous blog post, are a vital in helping reduce choice-induced anxiety for customers. These systems come out of a recognition that providing a large among of choices at low prices by itself is not sufficient in generating the best outcomes for consumers.  

Especially because of the trend in doing away with dedicated and experienced service staff, physical retailers haven’t had any kind of equivalent systems that they could potentially deploy. 

a person using a tablet to brose an e-commerce store with non physical ai retail analytics technology

How is Physical Store Analytics is Becoming More Like E-commerce? 

The development of affordable sensors and machine vision technology has allowed a convergence that had previously not been possible. Now it is possible for physical retail spaces to use some of the retail analytics methods previously only available to e-commerce stores.  

Previously traffic measurement in stores was very rudimentary, with rough counters at the exits. Another important signal is, of course, inventory and the volume at which it is depleted at different points. Once inside the store, there was no reliable information about how navigation occurred, where customers lingered, which products were frequently considered but rarely bought, which products were inexplicably missed.  

There is an almost endless amount of combinations in how products as well as the space between products can be arranged even if the retailer isn’t a supermarket which carries, on average, over 38 thousand products. With the ability to quantify all of this lost customer interaction through sensors and machine vision, A/B testing becomes effectively possible in the physical retail space. It is possible to set experiments and measure outcomes, only in physical stores rather than on a digital webpage.  

What's Next for Physical Store Analytics?

A swathe of retail analytics becomes available, such as heat maps, average dwell time by area and accurate store occupancy times plotted to each individual customer. At this point none of the data requires access to personal or individual information of any kind, it is simply the accurate measurement of bodies moving within spaces. Even extremely deindivisualised data presented in aggregate from the use of physical store IoT sensors can bring a wealth of insights without any potential privacy concerns that may stem from collecting data on individuals. 

Real-time Sentiment Analysis and Retail IoT

However, there is also increasingly the ability to bring hyper-targeted recommendation engines into physical space. Using Azure Cognitive Services it is now possible to reactively display product recommendations based on the demographic details of a person who approaches the display. Depending on customer opt-in and privacy consideration these recommendations can also be made based on purchase history as well. Rudimentary sentiment analysis can also be used to display customised content based on estimates as to the customer's mood. 

Real-time sentiment analysis is the latest practical development derived from a field known as "behavioural economics" which posits that significant amounts of human economic decision making is driven by emotion rather than rational decision making. Gallup estimates that the breakdown is 70% emotion to 30% rational and that companies that incorporate the principles of behavioural economics into their sales strategy can outperform their peers up to around 85% in sales growth.

Marketing has always been concerned with sentiment, driving demand and desire through tapping into the unconscious and libidinal. With the ability to roughly quantify sentiment at the point of purchase and having the control group be all customers rather than focus groups, this process can become a lot more scientific and make marketing and store outfitting budgets a lot more impactful.

Sound like science fiction? Hardly. These are all solutions that are very possible with technology that is available today. We have already delivered similar solutions to clients. If you want to know more about these real-world applications, watch the free webinar on AI for Retail hosted by BizData: