In an age of online-only retailers like Amazon, Kogan and Alibaba, it’s easy to think that digital is the antithesis to bricks-and-mortar stores. But in reality, digital technology, such as machine learning and data analytics, can go a long way to actually help physical stores.
Take for example knowing where to put one. When opening a new store or branch, where should it go; where the most foot traffic is? On the high street or in a shopping centre? Where there’s a gap on the map?
Most advice starts with consulting a map but this is a mistake. Technology should be used to find the optimal retail location, starting not with the address, but with the customer.
Opening a new premise is an expensive business. Getting the location wrong can represent a substantial risk; at worst spelling financial disaster for the brand itself or at best, being a waste of time and potential revenue.
Using geospatial analytic insights a business can not only identify targeted opportunities to open a new branch, but also understand how to develop underperforming stores, move the average by identifying what makes the best stores the best, dominate in attractive areas and know what bad markets or locations to leave.
Our work with clients has found that starting with geography does not bring the best results. We developed an approach that instead starts with the customer.
Using artificial intelligence and machine learning allows a company to more efficiently segment its existing customers and to identify priority customer groups by their unique attributes and behaviors. Traditional location analysis would typically only consider a handful of metrics, but the more sophisticated machine learning and AI modelling techniques are often now analysing thousands of metrics describing consumer behaviours, preferences, wants and desires – all at the same time.
For example, opening a new self-service, low-touch bank branch would be best in an area where customers are early adopters, younger in age and who often shop online. Other priority segments could be around high-value customers, homeowners, high net worth customers or those that prefer face-to-face shopping.
This allows a business to know the kind of customers that they are trying to service within a particular store. It also provides a quick way to rule out areas that don’t represent these priority customer segments.
The second part of the process involves adding external layers of data to the customer model that will allow a business to understand the market. Here, socio-economic and cultural data can be pulled in from various sources. ABS census data, surveys and polling by companies such as Roy Morgan or Neilson, employment data and any other sources available can then be used to add to the picture.
Knowing the types of customers being targeted from the customer modelling in step one, more of these type of people can then be identified with machine learning. So for example, again back to the banking example, knowing which locations align with a greater amount of homeowners (and, knowing where your customers are, being able to tell which areas have more mortgages not with your bank, aka potential conversions).
These customer segments and markets will help you to define the number of areas you are narrowing down to, and which locations should be evaluated further. Another layer here could be to add in where the competition is.
By leveraging your existing data to model the market and customer segments – such as store revenue, transaction data, or employee performance and the like, a business can then see further layers of detail in the story that is starting to take shape. This allows them to identify where a new store will best fit on the map precisely, but also, has the added benefit of giving critical information into potential problems with existing locations.
Knowing which areas are doing well with sales alongside knowing if they are in areas of high competition or not removes ambiguity. If your high net worth customer segments are doing well in areas with low competition, but badly in areas with high competition, this could point to a product or services problem that can, with the above data, be better understood and addressed. On the other hand, if most branches are doing well in shopping centres in areas of similar socio-economic clustering, with similar product lines, but one specific store isn’t? It could be an employee or cultural problem solved with training.
Essentially, data modelling can narrow down the factors until you know exactly what situation is being faced in what areas.
The benefit of this approach is that unlike modelling solely on geography or market data, a business can make sure it’s honing in on exactly where its existing and potential customers are – and know what they want. In short, this method gives business more than just maps.
With the insights garnered, marketing efforts can also be optimised – for example, targeted online advertising to those early adopters in the above example. Or, you may discover that a particular store is doing well because you are able to stock the products likely to appeal to a customer segments that frequently live, work or play in those locations.
These examples are obviously quite simplistic for the sake of making the point, but it’s important to note that with AI modelling the insights that can be derived can be extremely detailed and endlessly useful.
So when opening that new branch or launching a new store, success may not be guaranteed, but if you’ve based the decision on how best to meet your customers’ needs, it will be a lot more likely.
PwC’s Insight Analytics team provides detailed advice and analytic modelling for retail location optimisation. For further information please contact Phil Bolton.
© 2017 - 2021 PwC. All rights reserved. PwC refers to the PwC network and/or one or more of its member firms, each of which is a separate legal entity. Please see www.pwc.com/structure for further details. Liability limited by a scheme approved under Professional Standards Legislation.