Emma Herrod investigates how data, digital technologies and machine learning are altering merchandising.
Customer insight, big data, machine learning and artificial intelligence (AI) have all been transforming retail jobs across the industry as retailers use every trick in the book to make the customer believe that they, as an individual, are understood. Brand messaging shows how the company knows what it feels like to finish the marathon, to discover the right outfit for an important occasion or to catch the biggest fish in the pond. Whatever the customer’s passion, fears, uncertainties or taste in music, there’s a retailer ready to match them, and provide inspiration or support through products and value-added services.
The retail experience has become more than simply putting product in front of a customer and hoping they will buy it, as was the practice during the early days of ecommerce. Data provides a view into each customer’s life in a way that’s reminiscent of the type of knowledge a local butcher or baker would have once had.
By 9 o’clock this morning, social media, search engine and 2 retailers had all been privy to my innermost thoughts and social life through posts and images viewed, searches made, items browsed and added to my grocery shopping list and the purchases actually made. They know which online news services I read, my credit card number and where I live – and they can surmise much more from people who behave like me, too.
Our digital data trails – freely given by each of us – offer a treasure trove of insight for those that can unlock them. Personalisation in the form of marketing material, discounts proffered, adverts displayed and how a website appears to each shopper are all common practice now. And AI, in the form of marketing automation and web personalisation systems, is continually learning and adapting to website visitor behaviour.
We’re all familiar with the theory of how machine learning and AI can help in merchandising and trading ecommerce sites, learning through customer data and how shopping journeys are presented to potential customers. But not all retailers are using AI yet.
A survey of the InternetRetailing Top500 discovered that about a third of respondents are not actively using AI marketing at this stage, while the remaining two thirds are working on it or actually using it. While this is encouraging, the spread of experience with AI marketing is huge.
The largest single segment of those interested in AI marketing are the 17% at the planning stage only, showing that it is for many still very early days. Those who are in the defined and executing and fully implemented stages add up to just over 20%, while 8% are expanding their existing processes. However, with 11% ticking ‘don’t know’ and 34% indicating that they are interested but not taking any action, there is still a long way to go before the majority of retailers are using AI.
Start-ups, though, as ever at the forefront of retail thinking and unhindered by legacy attitudes or systems, are getting the most out of AI. For example, US personalised fashion service Stitch Fix uses AI to match stylists to shoppers and to decide which 5 items to send out in each subscriber’s monthly parcel.
Attribute-based forecasting, machine learning and human experience will enable demand forecasting to be down to the individual customer level within 10 years
It is one thing to use customer insight or AI for promotional marketing and personalisation, placing the right product in front of the right customer at the right time, but how is technology, digital and AI transforming customer insight at either end of the product lifecycle and how is data helping merchandisers in the rest of their duties?
One aspect of merchandising is making decisions about what stock needs to go to which store locations, how much, and the product mix in terms of sizes and colours. By looking at historical data – as well as what customers in different postcode areas are buying online – patterns emerge in customer characteristics as well as in the attributes of each product and the overall mix.
But it’s difficult for a retailer to make decisions based on thousands of different bits of information, customer comments or by looking at individual purchases; the data has to be displayed in an actionable format so they can translate it into a buying decision or product merchandising decision.
Peter Charness, SVP Americas and global CMO at TXT Retail, believes that the tools available to merchandisers are getting better at providing this. He says: “A lot of systems were developed that supported the merchant understanding of what sold. The next generation, as people start to bring in more capabilities, is going to provide the basis for the merchant to understand why something sold.”
He explains that merchandisers will be able to identify the characteristics that they need to look for more of in the products that are selling well, rather than simply that x number of a specific red dress were sold each week. They will be able to look behind what sold to find out why something sold and how many people are likely to be interested in purchasing a product based on its characteristics.
He adds that although the role of AI in merchandising is still in its infancy outside of marketing and promotion, it is something that is becoming increasingly necessary for retailers.
Giuseppe Rapisarda, Director, Oracle Retail agrees, saying: “Within 10 years, demand forecasting will be down to the individual customer level.” He adds that this will be enabled by a combination of attribute-based forecasting, machine learning and human experience.
Even once an item is sold it can continue to offer insight. There are the reviews from customers, for example, which have built trust for others when making purchasing decisions as well as informing merchandisers and development teams about issues with products and why they may not be selling as well as predicted.
“If somebody is complaining that something was cheap, didn’t last or didn’t fit properly, that’s feedback that can go into the merchandising office and something can be done about it in terms of better quality control or changing the size specs on the next thing that’s manufactured,” Charness says. “The translation of what the customer is saying out of social media into the next merchandising decision is still a difficult translation to make.”
It is from ratings and reviews, along with unstructured, ‘dark data channels’, that some of the most valuable forms of data can be gleaned.
Some retailers have tried to bring social to their own sites by adding forums and ways for customers to share information and post their own images. This user-generated content provides insight as shoppers happily share what they have bought or post images of themselves wearing or using the item. This social proof provides plenty of information for retailers able to utilise the insight as well as increasing the trust other shoppers have in them.
Luxury watch brand Bell & Ross, for example, collects photos and videos of shoppers from social media and incorporates them into its website to enhance engagement with customers, boosting its CTR and in turn increasing conversion. Images can also be made ‘shoppable’ enabling website visitors to purchase by clicking on them.
“For a luxury brand, we have always been close to our fans and customers,” explains Antonin van Niel, Head of Digital, Bell & Ross. “Our online community is large and has become more and more active. We are seeing huge growth in our customers sharing their experiences with our hashtags [#BellRossCommunity - #WatchBeyond]. This content drives a strong sense of engagement in our website, creating a more social shoppable environment. With Photoslurp we found exactly the platform we were looking for to increase conversion using social proof: an easy-to-use, flexible, data-driven platform that enables us to optimise our digital and social strategy effectively.”
US outdoor retailer LL Bean is taking things one step further and exploring how performance and use information can be collected from clothing once it has been sold. It is using technology from Loomia which collects data from the clothing and, with the person’s permission, shares it securely with the brand. Security is provided by blockchain. It is hoped that the solution will provide valuable insight into the use rate of products returned under LL Bean’s 100% satisfaction guarantee.
A collaboration between Google and Levi’s has created a denim jacket for cycling that has sport-specific adaptations as well as conductive threads woven into the fabric. These threads connect with and control the Jacquard platform which connects the wearer’s phone controls with touch input on the jacket.
There have been experiments with connected clothing – where the wearer is connected to the internet – for some time. As Ian Jindal reported after seeing the Levi’s jacket at NRF, subject to the necessary user permissions, the firm can now have a relationship with the customer based on how and where they use the garment. It will have access to their commute, the music they listen to, their frequently used commands and apps. This offers it a much more relevant conversation opportunity than the usual “you bought this a year ago so please buy it again”. It’s conceivable that the jacket’s data could link with Google Home data as well.
All of this information about consumers and products streaming back to trading teams, merchandisers and buyers has gone beyond the capabilities of spreadsheet analysis. AI and its ability continually to learn, adapt and discard old ways of thinking means that retailers will be able to keep pace with changing consumer thoughts, behaviour and shopping patterns, and optimise planning, pricing and clearance to maximise margin; all the while keeping their decision making in line with business rules and merchandiser expertise and instinct.
Where once – and in some cases still – the butcher knew the optimum price to sell the most lamb shanks and the Rigby & Peller dresser knew how much support the Queen needed, personal, intimate information about every product, garment and customer nuance in large retailers and fast fashion chains can be known by computers with data shared and insight gleaned to assist in the development of the next product range, pricing decision or promotional activity.
The jury is still out on how far machines will take over different roles in retailing, but for now their assistance is taking away monotonous tasks, real-time processing and big data crunching.