PRINTABLE VERSION HERE

Omnichannel department stores have two big advantages over pure players, including Amazon: they know their products, and they have stores. Department stores need to capitalise on these by using their human potential to the full with product knowledge, and by tailoring stores to the behaviour of digital customers.

Know the product

We have all experienced recommendations by Amazon for books which can be, to say the least, puzzling. They seem to bear no connection to previous purchases, interests or even searches. Ben Evans makes the provocative point that Amazon does not really know what it sells. The key for Evans is that Amazon operates with a logistics-based business model. That is, it deals with relatively uniform packages on an infinitely long shelf. The system can describe the products, but usually using only information from the suppliers, and incorporate that limited information in its presentation and description, and in its recommendations. All categories of goods are treated in the same way as long as they can be fitted into a box on an infinitely long shelf. “Products are a number, a size and a weight” (see Ben Evans). And anyway 60% of what Amazon sells is not sold by Amazon.

Recommendations use a model based on the history of the user and the behaviour of similar users. Even on sites which are personalised to each customer, AI is able to provide recommendations which are increasingly subtle and feel close to serendipity from the customer perspective.

a)    Content-based filtering relies on using as much customer data as possible. A recommendation like “products similar to this” is an example.

b)    Collaborative filtering collects information from many other users to derive suggestions for a particular customer. A typical recommendation in this case may suggest items which go well with one another.

c)    Knowledge-based systems are more appropriate when there exists less prior knowledge because, for example, items are infrequently purchased, and data is lacking.

The challenges faced by recommendation systems are sparsity of data; latent associations (when labelling or description is imperfect); and scalability (when data sets widen and become overwhelmed by the multiplicity of products and clients). An interesting example is Netflix which integrates the history of what members have watched with product tags by employees who understand the content, and finally pulls these parts together with a proprietary machine learning algorithm.

Now it is clear that while Amazon has no sparsity of data, it does nevertheless have problems with product knowledge (and perhaps also scalability). This accounts for its sometimes bizarre, recommendations which appear to have no reason other than a word in common in the suppliers’ descriptions.

Convenience and automation vs experience and interaction

The Amazon model privileges price, speed, convenience and automation, and managed from the outset to dominate a category, books, known for content, browsing, experience and human interaction. In so doing, it changed the category itself. The large impersonal chains such as Barnes & Noble or Waterstones which had put so many small local bookshops out of business, found themselves in their turn struggling against the apparently unstoppable might of Amazon.

However, customers fought back by patronising what was left of the small independents, and the chains used the best features of the independents to transform themselves and win back market share from Amazon. Book retailing is thus polarising between an efficient commodity system on the one hand, and a local community experience on the other. Both apparently growing (for example, independents are now growing again from a low of 1650 across the US in 2009 to 2470 in 2018).

According to the Financial Times, James Daunt, originally founder of Daunt Books of Marylebone High Street in London, has successfully rescued the Waterstone’s chain in the UK from the Amazon juggernaut by transforming it into a chain of local stores. Since then, Waterstones has acquired the US giant Barnes & Noble, and Daunt has the (daunting) task of doing the same in the US with analysts and investors betting against him.

The ideas developed in his own shop including catering to local customers in a club-like atmosphere rather than developing a single and replicable store matrix, and recommending books which he and his staff had actually read rather than two-for-one offers, was extended to a vast chain. The market was with him since the current trend appears, for example, to be favouring books as objects against the e-book fashion which has plateaued at 15% in the US.

The model has also been surprisingly viable since the returns to publishers fell from 25% to 5%. And the publishers appear happy to hand over a degree of control to intelligent booksellers. The model relies on engaged staff but as Daunt puts it, “you need fewer people, but the compact is you invest a decent amount in others. You get into a virtuous cycle, and you end up with a well-run, happy bookshop”.

Thus, the advantage of the Daunt strategy over Amazon relies on catering to local customers, using that knowledge to offer a better deal to suppliers (publishers), and, importantly, knowing the product sufficiently well to marry the two. Arguably these qualities are at least in part at the centre of the very DNA of department stores which have historically tailored their offer to their customers, provided useful conditions to brands and suppliers, and once again importantly curated and studied their assortment so as to be able to offer advice and recommendations.

Adapt the stores

These elements rely to a large extent on having a physical presence as well as an online one since the stores provide the local touch and the human product knowledge. This is the second advantage which department stores have over pure online players.

Recent circumstances have transformed retail, in particular regarding online business: in 2019, less than a third of US retailers had implemented a digital strategy. The transformation wrought by the covid pandemic saw retailers implementing omnichannel experiments such as kerbside pickup, same day home delivery, or buy online pick up in store. For example, the share of retailers offering kerbside pickup had jumped to 44% by the end of summer 2020. Some of these were, of course, catering to the convenience market, customers who did not want to shop physically but who appreciated the speed and return advantages of a physical pick-up point. But others were more complex.

This buy online pick up in store (BOPIS) service is the one chosen by Ketzenberg and Akturk in a recent HBR article. They found in a large-scale survey that retailers offering BOPIS actually stole business from their competitors who did not. They were not only increasing their store traffic but also gaining in sales of the higher-priced, more profitable items since customer fear was alleviated by the possibility of effortless return in store.

BOPIS also meant that customers avoided the shipping costs and had visibility into the availability of the product, thus avoiding a wasted trip. While on that pick-up trip, some 85% of customers made additional unplanned in-store purchases. It should be remembered that BOPIS is one of the least costly and more profitable omnichannel services for retailers.

This approach clearly argues against abandoning the advantages of in-store shopping. However, how many stores are catering specifically to their online customers in store? What would it mean to organise a store around the idea that the customer is a “digital” customer, who has perhaps made a purchase online which they are picking up in store? Would the store look different from the current one designed around store-based purchase if it were catering to the online shopper primarily and as a pick-up point?

Staff and stores, past and future

The omnichannel challenge is not a new one. Sears of the US started life as a catalogue retailer in the 19th century and diversified into stores from 1925 until the store business overtook the catalogues. Sears is often compared to Amazon because of the breadth of its assortment. However, its mail order business model was quite different in spite of it also being “distance selling”. Indeed, Sears deliveries rested on the US Postal Service initiating rural free delivery in 1896 and parcel post in 1913, rather than on its own delivery capacity. Both the catalogue and the stores demonstrated considerable product knowledge with the early Chicago stores selling hunting goods, farm tractors and tombstones. The Sears buying offices, already large in Chicago, expanded their fashion competences in New York City when the stores opened. Also Sears developed its own brands (such as Kenmore and Craftsman) which have been a major asset of the group and have enjoyed a long and profitable life.

Thus, from an early date, omnichannel department stores (even those coming out of the mail order world) have had significant knowledge of the products they sell. This constitutes an advantage over some other retail formats today in that they are product specialists, or what are sometimes called “merchants”. Significantly, Amazon has always disputed its label as a retailer and described itself from the beginning as a tech company: “Amazon is a tech company, we just happen to do retail… really well.”

Department stores are notable also for having landmark stores which can provide additional service and experience. In spite of, as Kate Ancketill puts it, e-commerce having “deleveraged” the physical store, forward-looking retailers are opening stores, albeit newly designed stores, intended to act as support to the online business.

In conclusion, therefore, to exploit these two advantages to the full today would mean rekindling and significantly developing the product knowledge offer through staff, both selling and buying, as well as thinking of the store as a support to the online business rather than as a traditional selling machine.

Credits: IADS (Dr Christopher Knee)