If you want to use data science, don't learn "data science"

Many companies and marketers understand the importance of data science and have begun to incorporate it into their work, but are struggling to see results.

The main cause of this is that the objectives and problems of data science are not properly defined. As a result, the right data cannot be analyzed in the right way, and the data analysis is likely to be wasted.

This time we spoke with Professor Takahiro Hoshino of Keio University, who is an expert in statistics, behavioral economics, and marketing, and is also known as one of Japan's top data scientists. Professor Hoshino said,To master data science that can be used in the business world, it is essential to first understand "business science" such as economics, management, and marketing science." he says.

Professor, Faculty of Economics, Keio University
Takahiro Hoshino

In March 2004, he completed his doctoral course at the University of Tokyo Graduate School of Arts and Sciences. He holds a PhD in Economics. After working at the Institute of Statistical Mathematics, Information and Systems Research Organization, and the Graduate School of Economics, Nagoya University, he is now a professor at Keio University Faculty of Economics. He has served as a visiting researcher at the University of Chicago and the Kellogg School of Management, Northwestern University, among others. He is Vice President of the Society for Behavioral Economics and a Director of the Marketing Science Society. He also serves as the team leader of a team working to utilize AI in the field of economics and management at the RIKEN AIP Center. In 3, he received the Japan Society for the Promotion of Science Prize, the most prestigious award given by the government to researchers under the age of 2017. He has also received many other awards, including the Japan Statistical Society Research Achievement Award. He is involved in the development of evidence-based government policy decision-making as a committee member for the Cabinet Secretariat, Ministry of Internal Affairs and Communications, Ministry of Economy, Trade and Industry, and Ministry of Education, Culture, Sports, Science and Technology, and provides academic technology as a technical advisor to CyberAgent, Money Forward, Yahoo! Research, and others. In addition, he provides practical consulting to many companies on areas such as marketing and human resource allocation, and in 45 he founded Economics Design, a company that provides consulting based on economics knowledge, together with Professor Toyoki Sakai of Keio University and Associate Professor Yosuke Yasuda of Osaka University.

The first thing to learn is "business science"

── Professor Hoshino, what made you decide to pursue a career in data science?

I was very interested in how people make decisions and how they should make decisions.

The former is deeply related to psychology and behavioral economics, while the latter is deeply related to economics, statistics, machine learning, etc. In fact, there is no academic field called "data science" either in Japan or overseas, and I have been conducting research across these fields.

At first, I was interested in individual decision-making, but as I continued my research, I became more interested in decision-making in companies and other organizations. I think one of the factors that strengthened my awareness of organizations was the many opportunities I had to conduct collaborative research with companies.

Decision-making bodies are extremely diverse, including governments, local governments, companies, and individuals. I wanted to be involved in evidence-based policy making (EBPM) for the government and local governments, so I sought a place where I could study decision-making without being limited to a specific field, and decided to become a researcher.

── More and more companies are turning their attention to "data science" as a means of decision-making. However, in actual business situations, it seems that data science is not being used effectively in many cases. What do you think is the reason for this?

Both management (executives and marketers) and data scientistsThere is a significant lack of understanding of "business science" (in this article, this refers to academic fields that are deeply related to business, such as economics, management, and marketing science).But I think that's a big reason why data science doesn't work. In other words, it has nothing to do with the data.

To achieve results such as increasing corporate value and profits"What kind of decisions should we make?" "What should we optimize?" Business science is the academic discipline that provides the foundation for thinking about these issues, and is a treasure trove of knowledge that can be applied to the business world.Overseas, there is a huge amount of research results accumulated over many years that are actively utilized in corporate management, but the reality is that they are hardly utilized at all in Japan.

If management is unable to utilize knowledge of business science, they will be unable to set goals and issues, or devise measures that take into account the overall picture of the business. All measures will be ad hoc, and there is a high possibility that they will fall into a situation where they never achieve results.

This applies to organizations and strategies as well, but here I want to focus specifically on the KPIs that data science is trying to optimize.

As a company, its primary goal should be to maximize the benefits of stakeholders (involved in corporate activities), even though the importance placed on whether the scope is short-term or long-term, whether to prioritize shareholders or employees, and whether to give back profits to society may differ from company to company.

Identify your company's issues from the perspective of generating profits, break them down into specific measures, and set KPIs for each measure. By achieving those KPIs, you can optimize profits. This is the way things should be, but in many companies, Management has neglected the individual optimization of individual KPIs for each department that "someone somewhere said was important," and control is not being carried out with the ultimate goal of maximizing profits. As a result, no matter how much advanced analysis is done to partially optimize KPIs, it ends up costing time and money and generating no profits, which is disappointing.

KPIs are merely milestones for monitoring measures. Of course, it depends on the individual business, but in principle, the vast knowledge of business science can tell us what measures contribute to profits and how. First of all, we should stand on the shoulders of giants of business science.

On the other hand, it must be said that the majority of data scientists have not even studied the basics of business science, such as economics, management, and marketing science.

What decisions need to be made to achieve business results? What measures can be taken as a result of data analysis? Furthermore, we often hear of cases where people do not understand the big picture of the business, and therefore are unable to produce output that contributes to the business, even if they perform very advanced data analysis.

What is "Data science that can be used in business"?

── Professor Hoshino has advocated the importance of "data science that can be used in the business world."

What is "data science that can be used in business?"Data science is carried out by working backwards from "what decisions should be made to achieve results"This is the flip side of the reason why data science is not working well in the business world of Japanese companies.

Standing on the shoulders of giants of business science with vast prior knowledge, utilizing the standard techniques for correct decision-making,The first step to "usable data science" is to properly define the problem: "What should be optimized and by what means?"And this is the most important point.

Thinking about data is the next step. Think about what data is needed to solve the problem you have set, and what kind of analysis is appropriate in relation to the company's strategy. Data science is, after all, a means to make the right decisions.

I don't think there is any need for marketers to learn the specific methodologies of data science themselves, and I think it would be better to leave it to the experts. Rather, it is much more important to understand the thinking and basics of business science.

── If a marketer wants to learn data science, where should they start?

Again, it is important to first learn business science and understand how to make the right decisions and set goals.

If I add one more thing to that,Knowing the biases to be aware of in order to interpret data correctlyI think that is mentioned.

The data available to us is actually subject to various biases, and problems arise when we make decisions based only on the data in front of us without taking these biases into account.

For example, the basic policy for TV commercials for consumer goods is to air large quantities of commercials at times when sales are likely to increase, such as summer for beer and early spring for mobile phones. If you simply compare the amount of commercials aired and sales, it may seem that "TV commercials contribute greatly to sales, and other advertising is unnecessary," but since commercials are aired at times when sales are likely to increase, it is only natural that sales will increase.

Even after removing the mechanisms of advertising, TV commercials still have a certain level of effectiveness, but it is not as simple as it appears, and it is necessary to combine various media other than TV commercials. If you study business science, you will be taught concepts such as endogeneity bias, where factors thought to affect sales are actually determined by other factors, and reverse causation, where a relationship between advertising volume and sales can be seen because advertising was conducted during a period of high sales.

Do not blindly accept the data in front of you, but correctly understand what biases are present and calmly assess whether the analysis you are about to carry out is in danger of drawing the wrong conclusion.Is important.

Better decision-making creates a society that makes the most of the individual

── Professor Hoshino is not only focusing on research into data science itself, but also on cultivating talent in data science.

While working on joint research with companies and consulting as an advisor, I have repeatedly witnessed situations where companies are stuck on local optimization, stuck on the KPIs that someone somewhere said were important, as I mentioned earlier. I felt frustrated by this and my desire to improve the productivity of Japanese companies gradually grew stronger, which is what led me to advocate "usable data science."

In order to increase the productivity of Japanese companies, data science education, including business science, is important from a long-term perspective.There is a limit to what I can do on my own, so I want to train students with knowledge and skills in data science and send them to the business world to work on utilizing data.

Overseas business schools have frameworks in place for researchers and practitioners to conduct joint research, but the reality is that there are very few such opportunities in Japan. Although there is a lot of academic knowledge that has been thoroughly researched and proven in academia and can be applied to business, it is hardly being used in any way, and this is partly due to issues in the educational field. There are many practitioners who believe that "academic knowledge cannot be used in the business world," and I think this is a great waste.

Even in academia, doing what you do does not directly lead to profits, so there is a tendency to stay confined to the "comfortable" academic realm. I would like to continue to promote the fusion of academia and practice, and increase the opportunities and people to apply academic knowledge to business.

── As the number of data science professionals increases and data science becomes more widespread, what kind of future do you envision, Professor Hoshino?

The big goal isCreating a society where individuals can thrive.

A society where essential value is created by making the most of each individual's abilities, sensitivities, and passion.This can only be achieved if society is highly productive and has ample room to maneuver. To increase productivity, we need to improve the quality of decision-making at all levels, including the government, local governments, companies, and individuals.An environment (organization, human resources, systems, culture) for making decisions based on science and data, without being bound by conventions or conventionsWe need to make the necessary arrangements.

Over the past 20 years, the focus has been on simply cutting costs, which is easy to do, as a means of "improving productivity." However, it seems that Japan has been unable to create value like developed countries have, and incomes have fallen relatively, resulting in a lack of leeway for Japanese society as a whole.

We will continue to work to steadily advance the implementation of data science in society, increasing productivity through decision-making using business science and data, enabling people to focus on creating essential value that utilizes their individuality, and creating a society in which that value is valued.

[Interview and text] Chiaki Saito
[photograph]Daisuke Koike
[Planning and editing] Yuko Kawabata (XICA)