Inside Information

Analyzing your company’s performance data provides insights into changes that humans miss, and helps you make smarter decisions on everything from stock control to store location. It can also help build a business model that will fight disruptive newcomers.

​

If you’re not already using predictive analytics as an FMCG (fast moving consumer goods) brand or retailer, you ought to be. The chances are that your competitors (current and emerging) are already profiting from advanced analytic capabilities, helping them, to take just one example, to cope with the growing need to make decisions in real time.

Predictive analytics – a form of business modelling – is about being able to spot what’s coming, what’s changing, or what’s subtly different from one context to the next, to a degree that exceeds human capability. With greater foresight and deeper insight, FMCG manufacturers and retailers are better able to adjust their stock and services, store layouts, and pricing to gain a competitive edge.

​

The potential applications are diverse. Furniture giant IKEA is experimenting with predictive analytics to cut checkout queues at peak times, using mobile signals to monitor customer movement inside buildings. In the U.S., specialist retailer American Eagle Outfitters employs a team of PhD-level data modelling specialists to closely track – and learn from –  consumer behaviour across the company’s physical and online channels, so it can deliver what customers want. 

​

What these and many other retailers have realized is that serious data science holds the key to growing revenues and profits without pushing up costs or sacrificing quality. It provides the basis for new differentiation, enabling brands and retailers to pre-empt consumers’ needs and desires, and launch product lines and services that enhance the customer experience.

​

Accessible To Anyone

​

Probability calculations can help organizations spot when consumers are likely to defect to a rival, or when their propensity to buy may be highest, so that timely, relevant, personalized messages and offers can be delivered.

​

“Predictive analytics has become pervasive,” says Bill Nowacki, Managing Director of Decision Science at KPMG in the UK. The rise of cloud-based analytics capabilities and services has, he says, made it much easier for organizations to develop a sophisticated analytics capability without having to become experts internally. “Platforms like Microsoft Azure make very sophisticated capabilities accessible to anyone, so it’s not only the big brands with deep pockets that have the advantage.”

​

Many other industries are exploiting predictive analytics. Sports teams crunch historical and current performance and injury data to improve results and maintain fitness with tailored training and therapy programs. Utilities and infrastructure companies use predictive analytics to target maintenance spending and keep services running.

​

For brands, predictive analytics offers a way of making sense of customer sales data and consumers’ social data – which provide a lot of clues about changing moods and early purchase-decision triggers. New digital opportunities are emerging. For example, brands are starting to assign digital properties to drink containers to create a new line of communication with consumers beyond the point of sale.

​

This early foray into the internet of things via ‘connected packaging’ could generate an exciting new data flow to improve brands’ understanding of customers’ behaviour.

Diageo’s whisky brand Johnnie Walker is innovating with its Blue Label product, using a sensor tag (an antenna and integrated circuit) printed on the bottle label. Once scanned by the consumer’s smartphone (using Near-Field Communication technology), this forms the basis for a unique interaction with a specific consumer. The initial aim is to push out relevant information to the consumer, but the ultimate goal is that data will flow both ways, providing the basis for advanced consumer behaviour analytics, fed into marketing and R&D teams.

​

In retail, predictive analytics can help brands and retailers stay ahead of emerging trends in an age when the potential influencers of consumer behaviour have become so numerous and complex that companies can no longer rely on human instinct and experience to spot subtle shifts, variances, and opportunities.

​

Predictive analytics uncovers these discreet insights by analyzing the reams of data that organizations are collecting about customer transactions and behaviour (alongside other data, such as stock information, location-based data, and so on). Armed with this insight, managers can start to do things differently.

​

“Take any chain of stores and typically they will all be stocked similarly, even though each trading location is very different,” Nowacki says. “Retailers talk a big game about tailoring their activities to each customer base, but there is a disconnect between these goals and the execution.”

​

Old habits can be hard to break, but until now it’s been hard for retailers to make the case for different models on a store-to-store basis. Defaulting to the flagship store model has seemed less risky when there has been only a store manager’s hunch or a one-dimensional set of sales figures to go on.

​

Bringing decision ‘science’ into the equation allows retailers to be bolder in their approach to outlying stores. With detailed, multidimensional data to call on, managers can compare and contrast local and national sales and take into account local demographics, the position of the store in the town, and the impact of local events on a particular day of the week (so that if nearby elderly residents are bussed into town every Wednesday, stock and promotions are adjusted accordingly).

​

“We use 6,500 different indicators to create a detailed picture of a retail environment and to predict customer behaviour,” Nowacki says. “These include the position of the building in relation to other stores and to the nearest bus stop, and the ratio of home ownership to renters in the area.”

​

All this information needs to be used efficiently. The key to a slick supply chain is ‘farm-to-fork’ visibility, so that as soon as a customer orders a ham sandwich, the next pig is being taken to market. Accurate forecasting isn’t just about looking at past or current sales; it’s about anticipating what customers will do next, making and adjusting decisions on an ongoing basis.

​

“What retailers haven’t figured out yet is how to weave machine-based analytics into the fabric of everyday decision-making,” Nowacki says. “It’s not a case of replacing human decisions, but creating and putting decision engines next to a company’s decision-makers.”

​

Companies need to assess the quality of data used as the basis for future projections, so that when the probability of certain outcomes is calculated, companies can have confidence in the actions they take. The more sophisticated and reliable the analytics, the smarter retailers can be.

​

“Every decision could be better with more insight,” Nowacki says. For instance, retailers could use predictive analytics to adjust their offers and maximize yield at different times of the day or week, as when it’s the end of the season and time to clear summer or winter stock.

​

“Instead of a single, all-encompassing decision being taken, it’s about interconnected decisions and being able to pre-empt the situation at 10 a.m., 4 p.m., or 8 p.m.”

​

Seeing The Bigger Picture

​

Staffing levels can be a delicate area for decision-making, but data science can resolve the debate by providing an irrefutable and objective case for why a particular store needs more or less people, and how that might vary at particular times.

​

Predictive analytics can help refine pricing variations between outlets. Nowacki points to the example of a fuel retailer whose pricing strategy was to sell petrol at 2-4 cents per gallon below its nearest market rival. “They had 1,000 stores and saw Shell and British Petroleum as their competitors,” he says. “But in many areas, this picture didn’t bear out – perhaps because rival garages were further away, or because local people preferred the personal service they got from the smaller guy.

​

Objective findings helped demonstrate the variation.” In such circumstances, KPMG provides ‘heat maps’ of the trading areas around stores, which are as unique as a thumbprint. Predictive analytics isn’t just about matching stocks, store layouts, and staff levels to consumer demand. It can help managers see the bigger picture. Owners of shopping malls need to ensure the optimum configurations and positioning of stores, to draw customers in and maximize footfall. Predictive analytics can drive initial planning and promotional activities.

​

One of IKEA’s location analytics pilots predicted queues at cash registers at its concept store in the Netherlands. It demonstrated up to 30 per cent reduction of queue times is feasible during peak times. Casinos use predictive analytics to keep customers spending money for as long as possible, for example by optimizing the position of slot machines, entertainment areas, food outlets, and bars. Arenas and stadia use analytics to model and predict crowd behaviour, to improve safety and reduce the risk of crush incidents.

Predictive accounting allows organizations to make more astute decisions based on customer behaviour recorded through loyalty cards, Air Miles schemes, and gift cards. With deeper analysis, brands and retailers can forecast, for instance, the proportion of gift cards that are likely to remain unredeemed.

​

If it emerges that 10 per cent are regularly unclaimed – a percentage that may rise for a certain kind of customer – the issuer can re-invest those savings or re-target promotional activity to address the missed additional sales (gift cards usually result in add-on spending).

​

Tighter targeting is a priority for most retailers and brands. Online, customers leave a strong data trail when they register and complete transactions on a site. The tools for analyzing those data trails are similarly advanced, now helping to sift out new, discrete categories of visitors, even down to the way they navigate web pages.

​

Sander Klous, a professor in Big Data Ecosystems at the University of Amsterdam, and Managing Director of Big Data Analytics at KPMG, explains how this works in the context of a Dutch media company. “They have looked at visitor patterns for the purpose of selling advertising space to clients,” he says. “Beyond identifying whether someone is male, female, an adult, or a child, they’re applying machine learning to discover more about the behaviour patterns and different user categories.

​

“For example, a Category A viewer is characterized by rapid mouse action and a tendency to look at the top corner of the screen. In some contexts, that can be more powerful than knowing the user’s age or sex. The media company can then use this knowledge to improve click-through rates and negotiate higher prices. Similar machine-learning algorithms could be applied to improve the performance of retailers’ online stores, by influencing what goes where on its pages.”

​

This is where the boundaries between predictive and prescriptive analytics start to blur. “From a technology perspective, prescriptive analytics involves more mature use of machine learning and cognitive computing,” Klous says. “In the future, we can expect to see this kind of technology being used – in customer support, for example – to build up a more complete picture of the customer’s situation and history before he or she is put through to a live agent, so they don’t waste time proposing solutions that aren’t appropriate. Using machine learning, the system will have deduced what kind of customer you are and the type of response that will work best for you.”

​

Prolific Experimenters

​

Retailers should experiment continuously to explore such opportunities. “If you want to be at the forefront of being data driven, you need a different attitude,” Klous says.

​

“Booking.com, the online travel agency, experiments prolifically on a tiny percentage of clients, and performs dozens of experiments a day. If 90 per cent of these fail, it doesn’t matter because at this scale they’re learning a lot about what does and doesn’t work.”

Bricks-and-mortar retailers need to adopt a similar mindset, he adds. “Experimentation isn’t a privilege reserved for online businesses,” he says. “The cycle times may be longer offline, but the same needs and opportunities exist. Willingness to experiment is really more of a cultural thing. IKEA has put a lot of effort into experimentation.

​

“Standing still is far riskier,” Klous warns. “You don’t need a crystal ball to see what’s coming. Just look at the music, travel, and taxi industries. If you don’t embrace analytics and the new opportunities they highlight, plenty of new entrants will.”

​

The good news is that FMCG companies don’t need to get bogged down in the technology or the complex mathematics of data science. “Companies should aim to be technology light,” says Nowacki. “Think of it like a bank making credit risk decisions when they have a client in front of them, asking for a loan. In just a few moments, the clerk can get a decision based on the information given, yet the credit risk analysis and quote is being done in the cloud. Retailers have an opportunity to use cloud-based services in the same way to make data-driven decisions.

​

“When it comes down to it, they need to consider what will happen if they do nothing. Whether it’s Amazon’s drones delivering internet orders within an hour or something else, companies need to ask themselves how much longer they’ll maintain brand affinity when that’s what they’re up against. Disruption is coming, not only from small, new players but from the big behemoths, so it’s only a matter of time before they’ll feel the squeeze.”

Key Learnings

• Embrace change – it’s coming anyway. As brands and retailers focus on growth, intelligent analytics will create new business opportunities and help to keep competitors at bay.

• Review the range (and quality) of data available to you, which could help you to predict future behaviour (for example, past sales information combined with location-based activity monitoring, or online social behaviour).

• Develop a culture of analytics, and of experimentation. Get buy-in from the top and create processes to foster and enable this.

• Encourage staff – and managers – to be more analytical in their decision-making and to move away from a ‘gut feeling’ culture, by ensuring the data is at the disposal of decision-makers.

• Draw on external skills and technology-based services rather than bringing everything in-house.

 

To learn more on how data science can help businesses prepare for disruptive innovators, globalization, and digital technology download KPMG’s ConsumerCurrents 21, visit https://home.kpmg.com/ca/en/home/insights/2016/06/consumer-currents-21.html or contact Willy Kruh, Global Chair, Consumer Markets, KPMG at wkruh@kpmg.ca or 416-777-8710