Predictive analytics sounds futuristic, but some retailers are already crunching the numbers to improve efficiency. Lindsay Clarke charts what Tesco and German retailer Otto are doing.

Tesco’s analytics project uses historical data to predict demand for products using  weather forecasts

Big data is big news. The term has appeared everywhere from the Harvard Business Review to the Daily Mail, and is hailed as the future for retailers. Sixty four per cent of organisations worldwide say they are now investing in, or plan to invest in, big data projects, according to research firm Gartner.

For those who missed the memo, the term big data encompasses retailers’ use of the exponential growth in data from a variety of sources, including mobile phones, social media and internet searches. This is combined with data from internal sources such as transaction data, and - hopefully - used to uncover hidden patterns to help gain an edge on the competition.

Making data understandable

But for some retailers, big data has been part of their business for many years. Earlier this year, Retail Week reported that a Tesco project, which would now be classed as big data, has been bearing fruit for more than five years. It is
saving £100m a year through a reduction in wasted stock, compared with before the project started.

One strand of the scheme combines data from weather records with detailed sales data, broken down by store and products, to build computer models that predict future demand for product lines according to weather forecasts. A more accurate picture of demand means the retailer can avoid holding too much stock, or running out of stock all together.

The programme relies on a 100 terabyte data warehouse and a suite of sophisticated analytics and data
modelling software. But Tesco supply chain development programme manager Duncan Apthorp said at the Teradata Universe conference in Copenhagen earlier this year that communicating the message from a data project can be as important as the advanced technology producing it.

“One of the most important things is boiling down something very complicated to show how it affects a product, in a store, on a particular day. It’s about having [the message] so that a store manager can understand it quite simply. He will know nothing about big data or analytics, but knows an awful lot about retail.

“It’s all about putting what you do into someone else’s language and climbing into their shoes,” Apthorp said.

This ethos is reflected in how Tesco recruits for the project. While analysts and consultancy groups talk of a forthcoming shortage of data scientists, Apthorp’s team does not think in these terms. Instead the retailer hires graduates with the potential to understand the statistical and business problems.

“We’re hiring clever graduates: engineers, physicists and computer scientists. We train them in retail. We teach them to become data scientists, we don’t expect them to be that when they arrive,” he said.

Demand forecasts

The grocer does hire numerical specialists, but also recruits good communicators with some understanding of data and statistics, Apthorp said.

The approach appears to be working - it is running several cost-cutting projects. The inclusion of weather forecasts and factoring these into its demand planning alone saves £6m each year.

Another feature of the work is helping Tesco’s understanding of demand patterns caused by special offers. It has cut the number of instances of products on promotion being out of stock by 30%, said Apthorp.

In addition, managers are better equipped when it comes to discounting food towards the end of its shelf life. This was once left up to managers, but now algorithms built by the supply chain analytics team create discounts that are transmitted to handheld devices in store, avoiding about £30m of wasted stock a year.

Analytics have also been used to improve stock levels at distribution depots, saving about £50m annually. Apthorp says this proven track record has helped the team gain influence in the business and has raised awareness of the value data projects can provide.

Other retailers investing early in big data are also seeing an increased awareness of the importance of predictive analytics across the business.

Clear success

German online and catalogue retail giant Otto Group, which owns UK etailer Freemans and is a partner of Next, started using analytics software from software provider Blue Yonder about six years ago, collecting data from a wide variety of sources. And improvements in its demand forecasting have led to annual savings of tens of millions of Euros.

Since this success, business managers across the company have become more interested in the use of predictive
analytics, says Michael Sinn, Otto Group vice-president of category management support.

“A lot more departments and managers in Otto are very interested in big data and business intelligence,” he says.

Over the past year, Otto has used predictive analytics to slash rates of return on key fashion items, saving about e10m to e15m.

It has also improved gross profitability on men’s fashion items by introducing dynamic pricing using software from Blue Yonder, in which Otto owns a 50% stake. The technology means prices can be changed based on demand, and it can forecast what prices customers will accept on a particular day.

The analytics activity has been led by the supply chain team, but will now be supported by a central business intelligence team as data-driven decision-making becomes more embedded in all areas of the business. “It is a cultural change,” Sinn says.

But cultural change isn’t easy, and buyers and merchandisers may find it hard to accept computer predictions into their working life, he says.

“Our colleagues are buyers who may not have the education [in data] and they specialise in garments and fabrics.

They have to recognise that data is much more important than selecting the right yellow blouse,” says Sinn.

Dynamic pricing started in menswear because the department was not as focused on high fashion as womenswear and the team was keen to start a data project.

“Now all their colleagues ring me and send emails asking when can we start with dynamic pricing.

The numbers are the most convincing factor,” Sinn says. A survey of 51 decision-makers in the UK retail industry by research firm Loudhouse, on behalf of analytics software provider SAP, found that 46% said predictive analytics gave a competitive edge, while 48% said it improved customer satisfaction.

Alan Taylor, retail industry principal with SAP, says: “What I see working with a number of larger retailers is there is a thirst for predictive analytics. It is coming from the top down. There is, in the last 12 months, an understanding that it is an essential part of doing your job.”

As the volume of data in retail grows, it can be hard to see the woods for the trees - retailers such as Otto prove how much can be gained.

How analytics are saving Otto money

European online and catalogue retailer Otto is using predictive analytics technology from software provider Blue Yonder to forecast the return rates on women’s fashion, saving between e10m (£8.4m) and e15m (£12.6m) a year.

Category management support vice-president Michael Sinn led early analytics projects at Otto to help forecast demand. The system is now building forecasts of return rates per product by analysing data on customer behaviour, product characteristics, delivery times, pricing and costs going back to 2007.

Otto’s analytics team ran through 10 different hypotheses for different situations in which shoppers might return clothes. For example, the project found there is an relationship between deliveries being late and products being returned. As a result, Otto has worked with its logistic providers to tighten delivery times.

“The project has led to a reduction of return rates by at least 1% or 2%, which could be e10m to e15m for a company like Otto, through avoiding the cost of returns and overstocking,” Sinn says.