Tesco supermarket

Big Data giving retailers greater customer insights

From food to furniture, store owners are trying to profit from looking deeper into customer behaviour.

Retailers have eagerly embraced the idea of what they call the ‘omni-channel’ strategy: their attempts to sell to people not just in physical stores but through the online channels. Being able to offer customers a homogenous shopping experience relies heavily not only on the systems and IT architecture but on the data that lies beneath.This approach demands seamless integration of systems at both the front end and behind the scenes and has seen huge investment from retailers. Product availability, information accessibility, promotional relevance, ease of checkout and effective fulfilment are all key components of a competitive and omni-channel strategy.

A recent survey of 4,000 people by Mindtree underscored these points. The survey showed that shoppers expect a seamless experience across channels – 70 per cent say they combine online and in-store purchases. In addition 90 per cent will share personal information for a better experience and 17 per cent will spend more if the overall experience exceeds expectations.

Data is key

But key to facilitate this is having the right data and being able to leverage it. Andrew Fowkes, retail solutions director EMEA at SAS, says: “The joined-up customer experience is about much more than an outbound conversation with the customer at the point where the transaction is made. Things have moved on and it is about the touch points leading up to the transaction, the advice available, the possibility of returns, the variety of fulfilment options etc. All these experiences need to be joined up and the loop can then be closed. This requires data.”

Retailers need to profile their customers and reach them more effectively across multiple mobile, online and offline channels. This means that data needs to be held in a single place in a single format. Retailers know this and are investing accordingly. Argos, Tesco, Matalan, Walmart and Macy’s are just some of the companies heading down this road.

Pepjin Richter, director, ERP product marketing at Microsoft, comments: “Customers tend to go into the store much later in the sales process than ever before – if a sales rep can see a copy of what they have been looking at online, what preferences they have, what is on their wish list, then the service provision has so much more depth. It’s a counter strategy to pure online stores and also helps to promote brand loyalty though better customer service.”

Having this data in a single format also makes for better analysis to inform better targeted offers and discounts. Feu Vert, a French tyre company, used SAS’s Master Data management module to align customer data that had previously been held in various different databases, such as loyalty schemes, marketing lists and payment data. Collating the data meant that the company ended up with an accurate and properly referenced set of customer data that could then be held in a central format to be built on and used with other systems.

According to Marie-Pierre Dussoilet-Barthod, customer relationship management administrator at Feu Vert, one of the biggest elements was recovering and collating postal addresses. Within three weeks, 25 million had become 9.2 million, making for considerably better-targeted communications.

Customer data includes log-in details, purchase history and payment methods. For over-the-counter trade it extends to debit and credit card data recorded at the point-of-sale. The challenge then lies in finding the gold nuggets in this mountain of raw data. At the same time retailers must abidy by strict legislation to protect customers’ personal data and guard against security breaches.

As with Feu Vert, to effectively use the data captured, retailers need to be able to correlate it and have a proper referencing hierarchy so that groups of customer profiles can be created using like with like data and special offers can be made more relevant. One famous example of this is the American retailer Walmart which offers Friday evening specials of six-packs of beer and nappies. It was by combining data from its loyalty card system with that from its point of sale systems that Walmart was able to spot a correlation between nappy and beer sales and act accordingly.

Analysing billions of records captured over years that are held in the Teradata data-warehouse system, Tesco has uncovered a number of patterns that help the company stock the trucks that go to each store with products it expects will sell better that day. The patterns uncovered by the regression model built in MatLab are as much driven by weather as by date. Not surprisingly, hot weather leads to big increases in the sales of barbecue meat but hidden within the data are more subtle effects, such as the ‘first hot weekend effect’. Compared across two consecutive hot weekends, salad and barbecue sales will be much higher when the weather first warms up.

Good timing

The data that stores receive every day drives many other decisions. Fowkes explains that today, discounting is very much about being able to operate in real time. “Although offers can be pre-prepared they are useless if the timing is not right, that is, someone buys something at full price only to receive a discount offer a week later. This is about using known behaviour to subsequently influence buying behaviour and make sure that the retailer gets the sale.”

When customers are ready to buy, perhaps encouraged by a discount voucher, they want to be sure they can head to where the product is in stock. Without being able to see stock levels and location any retailer is going to struggle with an omni-channel fulfilment strategy.

John Lewis was one of the front runners in updating its back-end systems to meet front-end demands. One of the key gains from this was that its product classification was simplified and this gave both customers and the retailer itself greater visibility. Julian Burnett, head of IT architecture at John Lewis, comments: “By creating a unified and consistent view of product availability and inventory data, for example, we are better placed to meet customer needs in a way that is most convenient for them.”

Sarah Taylor, senior industry director at Oracle Retail, adds: “Investing in planning applications and inventory management is essential. For example, although click-and-collect is great for the customer, it requires sophisticated inventory management and fulfilment.”

Mike Sackman, IT director at Argos, says there is a need for channels to be underpinned by a single platform which offers one central understanding of stock position and availability, which can be presented to the customer irrespective of the channel they use.

Holding data centrally and in real time also means that retailers can move more quickly to get product to where it needs to be if it is selling fast and adjust pricing. It also means they can introduce new ranges and react to local behaviour in a timely manner.

Fowkes says: “Forecasting is made so much harder if data points are not all joined up. It is about translating joined up data sets into being able to make omni-channel work operationally behind the scenes.

“Accurate forecasting means knowing how your own promotions will impact the wider category and how that affects inventory and storage at the warehouse: it is being able to accurately forecast how many units of product are needed. This means less waste and mark-?downs and it also means that the product can be in the right place or channel at the right time.”

Shelf life

Forecasting and waste reduction is clearly important in the food sector where products have a limited shelf life. This is further complicated when the ability to order in food to be picked up in store comes into play. Should those orders be fulfilled from a store’s existing stock or be brought in from a warehouse? This is something that Marks & Spencer is looking at and is very reliant on being able to accurately forecast demand and predict where promotions on one product will positively or negatively impact related products and adjust stock levels accordingly.

The same principles work in the non-food world, too. For example, Carrefour recently installed new software to make forecasting decisions on non-food items on a store-by-store basis. The technology is supported by three statisticians and one data manager working to ensure that the software is asking the right questions of the underlying data and can present that in a readable format.

Pierre Emmanuel Berneau, statistics and data management manager, comments: “We wanted to be able to work with the data and provide accurate forecasts, identify the right promotions and the resulting uplifts and effects on other products. The system allows us to transform our data accurately and to build in other components such as stock management.

“We have looked to bring together data from previous promotions and built up a profile of sales by geography, size, seasonality, etc. This gives us profiling built on accurate data and from where we can build and extend our modelling capabilities,” he says.

Fowkes adds: “The demand has gone from just having time series data to historical data to now being able to add in any number of variables that can go down to a very granular level. This is important for example when it comes to express stores where storage space is limited. Here getting the inventory right is important.”

Having the right forecasting data also impacts pricing. This is not just about a retailer’s own pricing strategy but also being able to compare and contrast with competitors. Fowkes comments: “Pricing is coming to the fore. Being able to accurately price, mark down, discount relies so much on being able to know what other retailers are doing. Up until now this has been done very much manually and the metrics to be able to compare and contrast are lacking, as is the ability to firstly get that data and then make it comparable and therefore able to be analysed in the context of in-house data.”

Finally, data can be used dynamically to influence overall strategy and operational decisions. This is where the role of the data scientist comes in, by being able to ask the right questions of data and then present it to various departments and decision makers who need to take the results of the analysis and use it to make better strategic decisions.

Analytics drive

Carphone Warehouse, for example, has been working with Telefonica to use data analytics to inform its store location strategy. The aim was to be able to better identify footfall and dwell times in any given high street and from that to be able to decide where to open stores.

As a result, data and the tools to analyse it are must-haves for retailers. Data records are no longer static things to be stored but live, dynamic assets.

Wise retailers know this. With trillions of bytes of information available about customers, suppliers and operations, it is apparent that those brands that invest in big data now are likely to reap rewards in the future.

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them