Just in time: how predictive analytics could transform in-store retail
As a store owner, imagine the following: in the never-ending search for ways to drive down costs and boost efficiency, in the quest for the ultimate in responsive, personalised customer experience, you decide to turn to data.
You gather together all of your transaction data for the past two, five, 10 years - the long lists containing every detail about which products sold and when. To this you add all the customer data you can find, giving you insight into who has bought what, when they bought it, what factors seemed to have influenced each purchase, the details of each customer's broader relationship with your company, participation in loyalty schemes and so on.
You also bring in all the data you can gather from marketing, merchandising, inventory, and supply chain, knowing that hidden amongst all the numbers there is invaluable information about how factors like branding, promotional campaigns, stock selection, availability and where products were located in store correlated with transaction patterns. All you need to do is ask the right questions and make the right connections.
What, for example, if you could cross-reference transaction, footfall and customer demographic data not just to tell you who bought what on any given day, but how patterns of product sales correlate to when and why different customer groups come into the store? What if you could piece together, season-by-season, month-by-month, week-by-week, and even day-by-day, who was coming into your store, what their motivations were and what they most liked to buy?
What if then, armed with very precise insight into the relationships between customer, category, time and place, you could start to arrange your merchandising to suit who was most likely to come into your store? What if you could ensure the products certain people were most likely to buy were always in stock when they arrived, located in a prominent position where they were particularly likely to browse, backed up with a promotion designed to appeal to their preferences?
The possibilities go on and on. It sounds like trying to see the future. But, what we are actually describing is the robust science of predictive analytics applied to retail, the discipline of extracting patterns of probability from (often very large) data sets in order to inform future decision making.
Predicting supply and demand
The kind of scenario described above - guessing when a particular customer is most likely to want to buy a particular product so everything can be done to maximise the chance of a sale - is far from the stuff of science fiction. It sounds like mind-reading, but it really boils down to sophisticated mathematical analyses of supply and demand.
In industry, manufacturers already use predictive analytics to streamline their operations for ultimate efficiency. Not only is production strictly governed by anticipated demand, so is the supply of materials to keep cost-to-output ratios at an optimum. They call this 'just-in-time' manufacturing, where waste is kept to an absolute minimum.
In ecommerce, the likes of Amazon and Chinese marketplace giants Alibaba and JD.com have applied predictive analytics to re-write the rules on everything from online product discovery to fulfilment, multichannel availability to personalised shopping journeys. Drawing on the incredible wealth of user data available through digital platforms, these sites have started to master the art of putting the right product in front of the right people on the right channel at the right time.
To back up this advanced form of personalisation, Amazon also uses predictive analytics in its logistics management, moving stock between depots in response to real-time demand predictions to guarantee rapid fulfilment.
But what about bricks-and-mortar retailers - do they have access to the depth of data required to create fully personalised shopping journeys the way the ecommerce pioneers have? Do they have the insight and control over supply chains and merchandising to achieve just-in-time agility, ensuring every customer gets what they want while minimising the risk of products sitting on shelves unsold?
How predictive analytics is used in retail
Physical store operators are no strangers to predictive analytics and there is no shortage of data for them to draw their insights from. Major retailers already employ 'Big Data' specialists to model the potential impact of marketing campaigns, pricing and service changes to prevent churn, and to assess which customers afford the highest CLV (Customer Lifetime Value), informing decisions on where to focus their relationship-building efforts.
Brands use analytics to scour social media and other online conversations, using sentiment analysis techniques to reveal assumptions and attitudes hidden in the emotional tone of a post or blog. This adds a great deal of depth to understanding of brand image and reputation, and provides early warning of when negative opinions about their products or services are starting to escalate.
Another established use of predictive analytics in retail is product propensity, which combines insight into consumer behaviour and sentiment with transaction patterns to help brands and retailers triangulate the relationship between customer, product, and channel. This is precisely where the opportunity to achieve a more efficient, demand-focused, personalised approach to in-store merchandising reveals itself, by optimising both channels and products to target the demographic with the highest chance of converting.
The key is having the ability to interrogate multiple datasets in concert so you can build up a meaningful, lifelike and three-dimensional picture of consumer behaviour on which to base your predictions. As this article in Forbes outlines, most physical stores can now make use of data from the following categories:
- Transaction data from point of sale
- Retail pricing strategies
- Loyalty scheme and shopper insights
- Footfall and customer demographics
- Store size, department sizes and product arrangement/merchandising
- Store navigation and traffic flow
- Competitor insights
- Contextual factors (seasonality, weather, major events etc.)
Plus, of course, the majority of stores now run websites, ecommerce platforms, social media accounts and so on, adding all of that abundant online data into the mix. So, the issue is not a lack of raw data for stores to work with. It is how these masses of unprocessed information are interrogated to reveal meaningful insights that counts.
Analytics, automation, agility
How likely are we to see predictive analytics have a transformative impact on bricks and mortar retail? According to McKinsey, it will be critical to stores becoming "nimble and adaptive" in a faster-paced, more complex market environment, not least in automating many of the admin-heavy processes involved in merchandising.
We are starting to see large retailers adopt technology specifically to improve their data-gathering capabilities. In-store video analysis of CCTV, for example, helps to reveal the relationships between product placement and purchasing decisions and also explores the sentiments behind decision-making at the shelf through body language analysis.
Perhaps the online innovators will show the way for how predictive analytics and Big Data science can best be applied in store. In China, JD.com, the country's second biggest ecommerce company, has expanded aggressively into bricks-and-mortar and is pioneering sensor technology to track and analyse shopping journeys and product selection. JD.com is very clear that its aim is to meet customer expectations with the same precision it does online.
Amazon has also built its checkout-less Go store concept around sensor technology. Based around an app which controls access to the store, tallies product selection from on-shelf sensors and processes payment, Amazon can identify every item each Go customer purchases on every visit.
Sensor technology already plays an important role in retail logistics and can supply stores with data on which to drive more agile, efficient supply chain management. Data gathered from shipping tags can be used to build up a detailed picture of supply lines, identifying where delays and wastage typically accumulate over time. Similarly, real-time analysis using machine learning techniques can create smart ordering systems to automatically adjust order volumes to both accommodate anticipated spikes in demand and adjust for shrinkage, helping to make sure you always get what you need on shelf.
We can also imagine a future where the other key ingredient in the Amazon Go concept, the mobile app, combines with predictive analytics to remove friction from the in-store experience. We have already mentioned how ecommerce operators use customer behaviour data to ensure customers see the products they are most likely to buy. One technique used is predictive search, using queues from the shopping journey so far to decide which landing pages a customer should arrive on.
The same principle could be applied to guide customers around a store via a mobile app, perhaps using AR-activated store apps to highlight where certain categories and promotions are based on previous purchasing history, online search data, social conversations and so on.
Giving the customer what they want is one of the great old maxims of commerce. The trouble is knowing what the customer wants in the first place. As tested and demonstrated using the vast data resources available via digital channels, predictive analytics can reveal information about customer behaviour and preference in a level of granular detail previously unimaginable. The old arts of customer modelling and profiling are giving way to a precise science based on probability, driving smarter decision making and vast efficiency gains. This is a world now open to in-store retailers, and is one that could reinvigorate both operational margins and customer experience.
Predictive Analytics is just one of the digital trends we explore in our Futures 2019 trend report.