An “Algorithmization” of Business Operations
A company’s data should be delivering real, quantifiable value. BCG Gamma’s Sylvain Duranton talks about what a company can expect when AI experts apply analytics.
This interview, conducted by Muriel Motte, was published in L’Opinion, a French newspaper, in April 2017
BCG Gamma, which uses advanced analytics to achieve breakthrough business results, was launched in 2016. Under the supervision of Sylvain Duranton, a senior partner in the Paris office of The Boston Consulting Group, 250 expert data scientists (400 by the end of the year) in North America, Europe, and Asia provide support to businesses in the field of artificial intelligence.
Muriel Motte: What is the mission of BCG Gamma?
Sylvain Duranton: We aim to assume responsibility for the analytics, or artificial intelligence, modules of BCG’s client projects. We operate in every domain, for example, optimizing container transport and improving the management of yield per passenger for an airline company. The distribution sector is particularly fertile terrain for the utilization of data. For one major US group, we set up a system of personalized promotions that addressed 12 million loyalty card holders. Every week, this group had been sending out different offers to 30 customer segments. Using internal customer data, we established a personalized marketing system that can produce almost 400,000 promotions that correspond to specific customer profiles. For a fashion house, our algorithms can even enable the company to predict, with 80% accuracy, whether or not a particular loyalty card holder will buy any clothing within the next 30 days and, if so, to anticipate that he will buy a trendy beige cotton jacket. For companies, the potential gain from data analytics can easily represent 1% to 3% of turnover. So this is huge.
Motte: On what is your work based?
Duranton: Essentially, we use the data stored by our clients and connect that with the external data collected from outside the company. For example, we have developed an application in New York that helps taxi drivers look for customers where they actually are. To do that, we divided the city into a grid of 200-meter squares and then identified, for example, all the bars, restaurants, museums, and offices in each square. We developed an algorithm that is based on this information and historical data of taxi transportation in New York. We tested the model to check its relevance and to improve it with data obtained in real time. Finally, we studied the way drivers use the data: What level of precision is relevant to making the right decision? Once the technology was working well with the users, we launched the industrialization phase. Today, our model is running also in London and other big cities. In summary, starting with data from the past, we developed a predictive algorithm that can even be prescriptive: in other words, it can either make decisions or aid in making decisions.
Motte: Are businesses underprepared for the data revolution?
Duranton: There is considerable literature on the gold rush represented by the conquest of data, the oil of tomorrow. Many companies have already recruited analytics teams — young people fresh out of the top universities who have PhDs in computer science and data science. The next major difficulty is to convert this science into action. We often see these teams working in isolation, like goldfish in their bowl, with no connection to the rest of the company. Or else we see teams that are well integrated but are working simply as prototype factories. They construct an algorithm and demonstrate that it can work, which reassures the board of directors. And then, nothing happens. But if you want to make money with data analysis, the rule is that the algorithm itself represents only 10% of the total effort. Technology and the information systems account for 20%. The essential part — 70% — is the adaptation of working processes, so that everyone is capable of using these new technologies. The devotees of analytics too often do the first 10% of the work ten times over and then simply stop short of the rest. My recommendation to companies is to do less but to do it all the way. That’s when analytics is really worth it: the positive impact can affect a company’s turnover by several percentage points. In the field of predictive maintenance, analytics can bring considerable savings by enabling companies to prevent breakdowns of extremely expensive machines.
Motte: After the digital divide, should we now be fearing a data divide?
Duranton: In many fields, companies have already been exploiting data for 20 or 30 years: telecom operators have been using it to model churn, the rate of subscriber cancellations, and airline and passenger rail companies and the hotel industry, for yield management, or setting the right prices as a function of varying demand over time. The rupture stems from the fact that today all industries, all sectors, and all functions — including HR — are concerned. In general, startups are well-equipped for this new revolution. Many are founded on the data universe. Among major corporations, taken broadly, industrial companies are trailing services. Small and midsize enterprises will have to progress on the cultural level, but they are more adept than others at concentrating on a few very important subjects. They do not suffer from the prototype factory syndrome. By contrast, all businesses, small and large, are facing difficulties of recruitment. The job called data scientist was voted the hottest job of the year in 2016, and it will likely be so again this year. To hire a data scientist, a small company in the provinces is now in competition with Google and Facebook. This HR battle is very different from recruiting a production manager that another small company down the road also wants to hire. A profound overhaul of HR career paths is necessary because these populations have very different expectations. This is a major upheaval.
Motte: Ray Dalio, the founder, cochief investment officer, and cochairman of the speculative investment firm Bridgewater Associates, envisions being replaced by an algorithm that will define the company’s fund strategy and manage the career of its employees. Is that the future?
Duranton: I’m going to use a word that is not very elegant: we are certainly going to see an “algorithmization” of business operations. However, corporate strategy is one of the areas that cannot be modeled. Anticipating economic crises is very difficult. With corporate strategy, it’s more or less the same picture. Machines are unbeatable at games of go and chess. These are closed environments, where it is possible to test millions of combinations. With very open subjects, it’s not the same at all.
Motte: So we shouldn’t be afraid of the emergence of a world of robots?
Duranton: Our big area of expertise is the man-machine interface. Instances of entirely automated universes are quite rare. For example, medical diagnosis will probably always be the work of a doctor. By contrast, computer-assisted diagnosis will be the product of hundreds of millions of consultations or studies analyzed by algorithms. We are heading not for a world of robots but for a world of droids — in other words, robots that assist humans but do not replace them. On the societal level, this means that we will undoubtedly be able to avoid popular revolts. But that also means allowing for a long period of familiarization and hands-on training in the new technologies. This corresponds to the 70% that I mentioned earlier. We will have to appropriate this new world. One outcome of the revolution already underway: the companies that collect the most data and are best able to exploit their data will be giving themselves a significant competitive advantage. Like brands, data will become an asset that will have to be managed and nurtured in compliance with clearly defined ethics and principles. How will Google be able to guarantee us that its search engine is not biased? The societal challenges are immense.
Motte: What do you expect from your collaboration with Cédric Villani?
Duranton: We recruit the teams of BCG Gamma from advanced mathematics laboratories. We formed our partnership with Inria, the French Institute for Research in Computer Science and Automation, because, for certain subjects, we want to have the capacity to put researchers to work on client problems. In fact, we are currently in the process of signing the same types of partnerships with other major research institutes in England, the United States, and Germany. Our work with Cédric Villani is vital. He is a great expert on the connection between the academic and the business worlds. He is one of the scientists who are most involved in creating bridges between these two universes. We share with him a common aspiration for the development of the mathematics sector in France. Mathematics, being less hampered by social predeterminism, is an extremely important sector for professional integration and insertion.
