Big Data Knows the Customer’s Shopping Trail
Customer analytics making use of Big Data can significantly increase the business operations of companies.
Customer analytics making use of Big Data can significantly increase the business operations of companies. The N4S Program has also launched projects for studying this topic. They create pilot models for collecting and analysing in-depth customer data.
Using Big Data may become a central part of a company’s operations and it can generate new business operations. On an international level, Big Data and analytics are used in banking, retail and entertainment. Also many other sectors, such as the health sector, contain vast possibilities. In Finland, VTT Technical Research Centre of Finland is involved in a European project that is developing tools for earlier diagnosis of memory illnesses and identifying people in the risk group. The tools being developed are based on processing background information masses from different sources, and they are used in image analysis and machine learning methods.
At what point do people quit and never acquire the products? This is what we can find out with collecting and analysing data, says Joonas Lyytinen, leader of the Deep Customer Insight work package of the N4S Program.
“Data science is a new field that is very much talked about. The basic principle is that a large number of raw data can be used to build different models with the help of IT solutions and statistical methods in order to improve products and services,” says Joonas Lyytinen from Reaktor. Lyytinen is the leader of the N4S Program’s research area that aims to develop models for collecting and analysing in-depth customer data.
Big Data can be collected from mobile phones, internet browsers, customer databases and other digital sources.
“Nearly all business operations, and in the future also the Internet of Things (IoT), will produce data. These include even factory machines for industrial companies. Enormous production machines have a large amount of meters and sensors, providing indicators for how the machine operates. With the help of the Big Data produced by the machine, predictions about possible malfunctions, breakdowns and other problems can be made,” says Lyytinen.
Lyytinen started with Reaktor in 2004, which was a software development company at the time, engaged in coding for its clients’ software development projects. With the clients’ needs, the company service package quickly extended to designing user interfaces and graphic layout as well as concept and service design, and nowadays there are more than fifty Reaktor employees working in this sector. In the past few years, e.g. training services as well as data analysis and modelling have become parallel operations.
“There are approximately 300 people working at Reaktor, and we have our own department of experts who perform Big Data mining, statistical modelling and analysis. This is clearly an area with a lot of demand,” says Lyytinen.
Better service for customers
Joonas Lyytinen on Big Data
According to Lyytinen, the most visible services utilising Big Data are in retail and online services. Big Data is studied in terms of customer behaviour. “Which products do the customers look at, which products do they not look at? What do they place in the trash bin, what do they end up buying? What pages do they visit, what pages do they not visit? The intention is to collect data regarding customer behaviour and produce better service by analysing it, as well as information regarding what the customers want.”
“With in-depth customer data, we know how customers behave in a certain situation. Here we can generalise trends regarding what kinds of products they are most likely to buy and what kind of services they appreciate. Compared to traditional marketing studies, for example, this information in concrete, based on real customer behaviour. The aim is to be able to make nearly real-time product and service suggestions for customers through the data analysis,” says Lyytinen.
He elaborates that the customers are modelled based on certain variables, such as purchase history for products, and different segments are made of their profiles. New customers are placed in one of these segments, and products and services according to the customer group in question are offered to them.
“The movie and game rental company Netflix, for example, makes efficient use of this. The company suggests new products based on the customer’s purchase history and recommendation rating. The company has been even more successful in getting the customers to buy new products when the people in the same customer segment have recommended similar products,” says Lyytinen.
”With qualitative measures, it is possible to obtain information for why customers quit when faced with a certain function. We believe that cooperation with the research institutes in the N4S Program will bring new methods of analysis.”
Results with the right questions
Raw data is generated in different sources with great speed, and the tools necessary for processing it depend on how automated or unique the model is supposed to be.
“You rarely have to code algorithms and tools yourself. You could divide the tools roughly into three categories: graphic interfaces, script writing, and software libraries. The tools for graphic interfaces are limited, and therefore, when conducting progressive or rare modelling, the properties run out in the very beginning. The best tools for analysis are e.g. the open source code analysis environment R and some Python libraries. However, they require some programming skills. Open source code software, such as Apache Hadoop, is useful if there are several terabytes of data,” says Lyytinen.
Data collection and modelling are based on carefully set up questions. Lyytinen says that the correct questions are such where finding the answers helps the company in its business operations.
“The answer must be such that the person posing the question can understand it. All other questions are redundant, and therefore analysis is very much a question of the professional skills of the person posing the questions. The person performing the analysis, on the other hand, must know how to pose the question in a numeric form and make sure that the response is reliable in a statistical sense, and in general.”
Qualitative methods to support analysis
According to Lyytinen, statistical analysis produces probabilities, based on which the phases of the customer’s shopping process can be detected.
“For example, at what point do people quit and never acquire the products. This is what we can find out with collecting and analysing data,” says Lyytinen.
He explains that the reasons for why customers quit buying at a certain phase remain somewhat unclear. Indeed, Reaktor is interested in combining research with a qualitative point of view.
“With qualitative measures, it is possible to obtain information for why customers quit when faced with a certain function. We believe that cooperation with the research institutes in the N4S Program will bring new methods of analysis.”
One of Reaktor’s cooperation partners is the Department of Computer Science and Engineering at the Aalto University School of Science, where research has been conducted for 15 years regarding how to gain in-depth insight on customer behaviour and needs, especially hidden needs, which can be accessed by observing the customers in their operating environment. This issue has been studied extensively e.g. in terms of the usability of services and products. Furthermore, long-term research in the field of requirement specification and recent research in the field of service design bring in-depth customer insight along with the statistical methods.
