News and Insights
INNOVATE. INSPIRE. DELIVER.
How Retail Can Reap the Rewards of Data Science
Data science is a hot topic. And, as with all hot topics, there’s plenty of disinformation out there. REPL’s CEO, Mike Callender, and Rafik Salama, head of data science, took time out to discuss:
- the elements that make up data science
- where data science is headed
- how data science can improve retail
What is Data Science? Correcting Conflicting Information
Big data. Forecasting. Deep learning, machine learning, artificial intelligence. These are just some of the phrases bandied around when talking about data science.
But what is data science and how does it work? Rafik explains that data science is about more than crunching big data sets. It’s a combination of smart analytics, machine learning and artificial intelligence that “bring data to life in a hypothesis-driven approach”.
To unpick what this means you need to understand the different elements that combine to create data science:
- Data processing – involves tying together the sporadic data that organisations have and matching it up in an initial data cycle. This is where the quality of the data emerges.
- Data quality assessment – steps are taken to enhance or improve the existing data by finding patterns that reveal what’s missing and that can be fixed using machine learning.
- Problem exploration – the data is married up to the outcomes that need to be supported.
- Hypothesising – theories are proposed to start to answer the problem. Data analytics brings the information to life and gives insight into the decision-making process.
- The future is predicted – machine learning starts to learn from what’s happened before and begins to predict what will happen next. By understanding the drivers of historical patterns decisions can be taken to affect the movement of what will happen enabling businesses to become more prescriptive.
- Problem solving begins – the systems learn optimisation and decision making by testing and simulating. This gives organisations a handle on what will occur in terms of revenue if a certain decision is made now.
- Forecasting becomes more structured – data science brings a systematic approach and proves its effectiveness with irrefutable results.
- Machines impact the future – artificial intelligence computes the outcomes of various scenarios and recommends the optimum choice.
The margin for error with this kind of approach is excellent at just 3% to 4%. And it’s improving all the time as the machines continue to learn using collective intelligence and automated decision making.
However, as Mike notes: “We can only hit those error margins if we’re able to pull all the relevant technology in. So ML is doing the learning but AI is used to influence the future. You need to use both.”
How Data Science Can Improve Retail
Data has long been used in retail to inform decision making. From forecasting stock, category trends and loyalty promotions to sending customers tailored vouchers aligned to their shopping habits. And we’re also seeing the emergence of data science being used in the WFM space to great effect.
Currently, retailers are working at the macro level in their organisations. For instance, customers are sent discount vouchers on the basis of group data analysis and stock orders are placed using stock categories for a group of stores. “It’s a broad brush approach”, says Rafik. “Because, in the past, retailers didn’t have the computer power for anything more.”
However, technology has advanced and is now available to enable retailers to undertake data analysis at an individual level and leverage data-driven decision making at the micro level. As Mike sums up: “At the moment, supermarkets target a group of people based on predictions. Data science will target people as a group of one and learn what they do with the vouchers to get them to spend more.”
With the power of data science, retailers are now able to get down to much finer margins. For example, 15-minute level predictions for labour and half daily or daily predictions for stock. What’s more, this could be achieved at the lowest level: individual stores rather than groups or regions.
Without the right tech in place, this level of analysis is not possible as it’s too labour intensive. With advanced computer intelligence and deep learning, not only can huge sets of data be analysed but the systems will use machine learning to understand what to do with the data.
What does this mean in real terms? Find out in the next section.
5 Big Data Science Use Cases in Retail
IBM reports that 62% of retailers find using information and analytics creates a competitive advantage for their organisations. Here are five use cases that demonstrate its flexibility and power:
If a customer is looking for a particular ingredient for a recipe and the store has run out this is a loss. Multiply the issue up across a large number of products and locations and the loss becomes much more costly, making the problem of inventory management well worth solving.
There are alternatives to using data science, as Mike explains: “Retailers could increase their stock on hand value but their working capital would be very high.” Instead of operating the just-in-time model common to most retailers, it would be necessary to have lots of stock in store. Which is problematic for perishable stock and cash flow.
Data science solves the issue with highly accurate predictive analysis which reduces the amount of stock required and the high levels of category management work needed to predict the best stock by store. Mike notes: “Retailers will have more shop floor space to use for additional trading or it could be leased to other retailers or used to add other services, like a cafe or restaurant, to entice people in.”
Data science doesn’t stop at stock control. Getting items into the warehouse and onto the shelves requires labour, which needs to be forecast. “Without matched labour we end up with the same symptom: not enough stock on the shelves”, says Rafik.
With thousands of products retailers are faced with a big problem. They have to replicate their planning across every item and plan out labour on a day-to-day or hour-to-hour basis. Without data science, retailers can end up with a lost opportunity: they either run out of stock or items remain on the shelf and go out of life.
But add detailed, automated, large scale data analysis to the operational mix and leaders will understand exactly what’s happening and prescribe a data-driven solution. One example is to use data science to understand which employees work well together, based not on manager observations but by running scenarios that predict which pairings are most productive and offer the best levels of customer service. The system then uses this insight to create optimal schedules that control cost while maximising productivity and customer support.
Supporting the Customer Journey
In the fight against declining footfall, the Mall of America in Minnesota used voice recognition and data synthesis to create a holographic Christmas shopping helper in the form of an elf. All underpinned by data science.
Customers could speak to the elf who asked them a series of questions to identify who they were shopping for and the kind of gifts they would be interested in. The customers answers were used to recommend a range of suitable gifts for the recipient before showing the customer how to find the relevant store to make the purchase.
Yes, this is a little gimmicky, but it demonstrates the versatility of the tech that data science supports.
Increasing Online Sales
A significant 35% of Amazon’s sales come from proposals made by its recommendation engine. This classic use of data science analyses the consumer’s current basket, purchase history, items they have liked in the past and other customers’ purchases to generate product suggestions.
With over a third resulting in purchases, it’s fair to say that using data science in this way leads to accurate and effective recommendations.
The Pursuit of Profit
Data science is also being used to solve broad problems like profitability. By understanding which operational activities contribute to surpluses, it’s possible to build a performance management framework so store managers can focus on these areas.
This is accomplished by reviewing the entire makeup and impact of a decision and interpreting historical data, like sales, profit and revenue. With deep learning and data analysis, data science is able to anticipate the outcomes of different decisions.
When the optimal decision is identified – one that will help maximise revenue and minimise cost – the result is presented to leaders. With the evidence to support the recommendation, data-driven decisions can be taken with confidence.
Where Next for Data Science?
“Retailers already have the systems they need to use data science but they might not be at the right level,” says Mike. “At the moment we have lots of siloed intelligence systems, which is good but only half the plan. To get the true benefit you need intelligent integration which will take thinking to the next level with an AI intelligent net for retail.”
Data science has outgrown the lab and now informs real-life applications and decision making. This presents an exciting opportunity for retailers as the power of data science becomes increasingly apparent. With the development of integrated systems that will enhance collective thinking, it’s anyone’s guess where this powerful new technology could take us in the future. One thing you can rely on is that REPL will remain at the cutting edge of data science, using the latest technology to solve retailer’s biggest challenges.
Put your business on the front foot by leveraging data science. Talk to one of our experts today on +44 (0) 808 200 7375 or email us at firstname.lastname@example.org.