By Amir and Eli Schragenheim
Physical stores face a wicked problem that gets bigger and bigger. An increasing number of customers buy through online shops, offering lower prices and wider choice. Online shops also face a huge business problem, coming from fierce global competition and lack of a clear competitive edge other than price. This does not help the physical stores to re-invent themselves and offer a real alternative to the customers, being offered by online shops wider variety, getting everything shipped home and with lower prices.
Part of the reason that so many customers buy online is that e-commerce stores succeed to understand better the specific taste of customers and approach them with lucrative suggestions. This understanding is achieved by investing huge efforts to gather a lot of data from every user entering their site, and perform sophisticated analysis of that data. These are ongoing efforts so we can expect more improvements in manipulating customers to buy through the Internet.
How come physical stores do not take similar actions?
Physical stores have harder access to the relevant information. For instance, currently there is no good way to record customers who do not find the product they’re looking for. The behavior patterns of the customers are not recorded. There might be video cameras in a store, but their objective is security and all other behavioral aspects are not considered. So, it seems there is not much the physical stores can do to study better their customers.
This is a HUGE mistake. There is plenty of available data that can be processed into valuable information, and there are ways to access more data, which could be used to yield even more information.
Customers coming to big stores, certainly to supermarkets and drug-stores, buy, many times, more than one item. This is an opportunity to learn more about the possible relationships between different items, and the personal tendencies to brands, plus the role of the price in choosing from a variety of similar products.
The total purchase of different items purchased by a customer carries hidden information that testifies to the taste and economic level of the customer. To reveal the relevant information certain analysis has to be carried out, with the aid of statistics & machine learning (ML), to be able to come up with answers to key questions concerning the most important decisions every retail store has to make:
What new items to hold? What items should be eliminated? What is the relative importance of keeping perfect availability of an item? What items should be placed close together? What items could be used for promotions? What additional services, like an on-site bakery, should be added (or removed)?
Inquiring the total purchase of a customer reveals so much more than just looking at the sales of every item. The mix of items purchased at one time reveals wider needs, taste and economical behavior. When it is legally possible to identify the specific buyer, then the previous purchases of the buyer can be analyzed in order to define in greater detail the market segment that customer belongs to. Part of the value of maintaining customer loyalty clubs is the ability to link different purchases, at different times, to the same client. Thus, a client profile can be deduced.
The most obvious outcome is mapping the customers according to market segments, noting the different characteristics between the segments. Inquiring the purchases could highlight aspects regarding the family of the customer: spouse and children, their approximate age, their financial status and their preferences. These characteristics can be revealed through analysis of the purchases and the frequency of purchasing. Certain preferences, like smoking or being vegetarian, can be identified. Together the key characteristics are revealed in order to define several layers of the market segment.
Another important value that can be deduced from inquiry of purchases is the dependencies between different items: when item X is purchased then there is a good chance that item Y is purchased as well. These dependencies are sometimes intuitively deduced by some managers and they definitely impact decisions concerning maintaining the availability of both items, their placement within the store and even the possibility to sell them together as a package. Understanding the linkages between products helps to check changes of purchasing habits over time. For instance, when the economy goes down, how it impacts different segments, which is much more valuable than just watching the impact on individual products. We can expect a general shift to cheaper products, but which brands are replaced by cheaper ones, and which segment makes those changes more than the other, should be valuable in forecasting those changes before the actual change in the economy takes place.
The structure of typical purchases by different market segments would certainly initiate marketing moves that would capitalize on that understanding. Analytical knowledge, translated into operational policies, would impact the performance of the various branches of the retail chain, as the specific needs of the branch are recognized, but also some of the generic insights.
When every purchase of a specific client can be linked with the previous purchases of that client, then the frequency of buying could lead to initiatives to influence the content of a typical purchase of a specific market segment.
Developing the machine-learning (ML) module to categorize better the different segments the store serves, should enhance both the marketing and the logistics of every retail store. There are always dilemmas in holding slow movers given the amount of the logistical efforts required to keep that slow mover available. Being exposed to the right priorities, by understanding the full financial impact of the slow mover sales, would lead to better decisions about what items to hold in stock. The relative value of a slow mover includes its impact on the sales of other products. Understanding the relative importance of that particular item to a specific segment contributes to determine the slow mover impact on the desirability of the store from the viewpoint of the market segment.
Through ML the retailers can get better understanding of the customer loyalty to specific brands and items. When it is already established that a certain segment prefers item X to item Y, then by intentionally creating unavailability of X for one day, it is possible to discover whether most clients from that segment bought Y, or refrained from buying a replacement. It also answers the question whether buying the replacement would impact the brand loyalty. Promotions also cause people to buy the less preferable items, but it is of major interest to know whether it impacts the brand loyalty.
It is highly desirable to have access to the information on the availability of all items at the time a certain purchase was made. When item X happens to be unavailable, then it provides an opportunity to check whether the seemingly brand-loyal customers switch easily to the replacement.
Supermarkets and drug-stores are typical retailers where purchasing usually includes several different items. It seems absolutely necessary that such retail chains would invest efforts in ML to learn more about their customers habits and develop the process of coming up with superior decisions to capitalize on that knowledge.