Data management experts discussed the specifics of artificial intelligence and machine learning at the "Management of the Future" conference at GSOM SPbU. 

 

Can all tasks be solved using Data Science? Is it worth relying on the "opinion" of machines in a situation of uncertainty? What kind of analytics to conduct before launching a product on the market and why quantitative research will not enoght without qualitative research? These issues were discussed by data management experts at the IX annual "Management of the Future" conference at the Graduate School of Management of St. Petersburg University (GSOM SPbU).

 

The exchange of opinions took place at the panel discussion “Prediction cannot be predicted: how the practice of analyzing the future is changing”. Data Science is one of the fastest growing areas in the field of analytics. Arguing that this is a fashion or a stable trend, Leonid Cherny, MegaFon's Data Management Director, noted that Data Science today is already an urgent need for companies. Although, of course, this direction is popular among young specialists also because it is in trend.

 

Andrey Revyashko, Chief Transformation Officer of Eldorado, added that, at the same time, the use of Data Science by specialists without expertise will only increase the deadline for completing tasks. In addition, so far there are no clear rules on how to work with the machines.

 

The sought-after qualified experts in this area will be trained as part of the updated Master in Business Analytics and Big Data program, in which Vladimir Gorovoy, Senior Lecturer of the Department of Information Technologies of GSOM SPbU, product manager of Yandex.Vertical, will take part in the reboot of the educational track.

 

Discussing the practice of using neural networks in companies, the experts listed the “merits” of artificial intelligence: by analyzing the behavior and preferences of users, it helps to recommend products to them, effectively prevents fraudulent activities on websites, and can identify fake reviews and recommendations for goods.

 

Leonid Cherny noted that MegaFon uses artificial intelligence and machine learning in both online and offline products. In particular, neural networks help to calculate the required power to provide good communication coverage in a specific area. It was they who, according to the expert, helped to quickly adjust the work plan during the corona crisis in 2020, when communication was required "not in offices, but in suburbs." To cope with the uncertainty, the team calculated various scenarios and possible customer behavior.

 

For all the accuracy of mechanical calculations, reliance on them can lead to errors. However, when working with Data Science, this is a routine part of the process, and not at all a surprise. For example, Vladimir Gorovoy told how the Yandex.Vertical team faced a problem when the value of real estate, responding to the coronavirus crisis, changed rapidly and the valuation models that were previously used gave incorrect results.

 

To avoid these same mistakes, Data Scientists need not only quantitative, but also qualitative research. “An increase of 300% of some indicator, upon closer examination, may turn out to be not at all what we thought about,” stressed Vladimir Gorovoy.

 

Andrey Revyashko noted that high-quality research and hypothesis testing helps to ensure the potential success of a product. But, in his experience, cross-cultural research is also important, which can reveal what the machine will not catch: for example, the prevailing habits of users, or local characteristics that can affect the success of a product at a particular place and time.

 

Dmitry Nikitin, Head of Data Science, Cardsmobile, Wallet, agreed with this opinion: before quantitative research, the Wallet company conducts, among other things, make qualitative research, and products are often tested on colleagues.

 

But getting constant feedback from users is also not a panacea, said Leonid Cherny. In his opinion, there is no need to ask the audience about all stages of work on the product, firstly, because it is not always possible to be sure that the audience for testing and polling is chosen correctly. Second, Customer Development (custdev) may simply be an attempt to relieve the responsibility for making decisions.

 

“Customer Development is sociology, not mathematics. It is important to correctly interpret the result,” said Vladimir Gorovoy. “A professional manager must be able to combine different tools and determine whether it is necessary to use Data Science at a particular moment, or at this stage it is more important to invest in building a reliable team. For example, if the business is still small and collecting the data required for machine learning is difficult,” he concluded.

 

The discussion was moderated by Margarita Gladkova, Deputy Dean of the Higher School of Business at the Higher School of Economics. Students participating in the conference were interested in the use of artificial intelligence in small business, approaches to predicting the success of a product and how technologies help to cope with a situation of uncertainty.