The data could not say it more clearly: 91% of the children who played with Lego pieces were boys. Ten years ago the directors of the Danish multinational could have surrendered to the umpteenth evidence: the girls did not like to play with their constructions. However, the president, Jørgen Vig Knudstorp, a young engineer of 40 years (the first outside the founding family and a Lego fan since his childhood), was not convinced. The company embarked on a research study with 3,500 families for four years and found that girls were more interested in collaborative games and small construction. This is how the successful Lego Friends line was born.
This example illustrates very well why the data can not be treated as pure evidence. The explosion of big data He has been among executives and many believe that data should be at the center of decision making. Thanks to them, business management is now more scientific than ever. This approach has its limitations, according to R.L Martin and T. Golsby-Smith. According to these authors, decisions have to be made deconstruct between "you can" and "you can not". "A true genius has the ability to imagine processes and products that simply have not existed before," they say.
Lego Friends was a novelty and the data did not encourage much to investigate in the girls' market, quite the opposite. The vision of the "master builder" Vig Knudstorp managed to revive the multinational, on the verge of being swallowed up by its millionaire losses.
- ¿Mores ready with many mtos data?
With so many new metrics available, do managers take the risk of making wrong decisions based precisely on what the data does not say?
"The problem of big data is that science is done the other way around: first you establish a hypothesis and then you contrast it. In the case of Lego, what they said was that the girls did not play. Nothing else. They did not explain why. The data can lead you to interpret them as you are interested if you do not have a previous hypothesis, "says Pedro Rey Biel, professor of behavioral economics at ESADE.
For years that has been the focus: you generate a hypothesis and you go to reality to see if it responds to your model. The arrival of big data has made the "return to the tortilla", as Pablo Haya, director of Social Business Analytics of the Institute of Knowledge Engineering (IIC) explains. Machine learning techniques are being applied to unexplored terrain, according to this expert, and the way forward is different because data scientists do not assume anything about the variables. "The revolution is coming around. The algorithms generate the a posteriori hypotheses and keep the best. If they are well built, they are usually quite successful. How that conclusion has been reached, however, can not be explained. It's like it's a black box. "
There are examples of fraud detection on cards. "It would be difficult to explain why the algorithm signals certain operations as suspicious. In this case the important thing is to get it right and that when it is fraud, the charge is denied, but when it is not, the card is not blocked ".
The data scientist has to be, but the big data it has to open up to other fields in order to exploit the full potential of the figures.
- The why,It is important?
The answer to this question remains in the hands of managers, a world where the directions have been taken many times because the president had a hunch or simply to imitate the competition.
Do I need to know why my sales go up, what is the reason why the customers are leaving or the reason why a certain product does not work? Do I need to know what is inside the black box or to guess the future result? "For certain business problems the explicability it is not fundamental and in that case the algorithms have advantages, "says Haya.
"Companies will make better decisions if they analyze the data well and do not embrace them simply because they are there. To take advantage of the big data you need people who know how to ask questions and where to find the answers, "says Rey.
What do Airbnb, Disney, Uber, ING, Virgin, Google or Walmart have in common? They all have a lot of data from their clients, yes, but they also apply behavioral economics to their decisions.
- Two worlds with a single profile
The synergies of the two worlds seem clear, but apparently only one profile is being imposed: that of the data scientist. Maybe it's the first step, because the revolution big data It is quite recent and many companies do not have a history or have not kept all the relevant information for your business. "This also happens in large companies, it is not a problem for SMEs," explains Haya.
Once this phase is over, the next step comes: what now? Many have invested enormous amounts in saving thousands and thousands of data, they have sold them that the big data It's a wonder, but they feel frustrated. "There's everything, it's true," Haya admits, "and a repeated complaint among some managers is that the models, while still good, do not answer a business question."
The data scientist has to be, but the big data it has to open up to other fields in order to exploit the full potential of the figures. Rey believes that data capture technology has advanced much faster than the theory of what we can do with them. "I think that only 5% of the data is being used," he concludes.
The data is the oil of the 21st century, says one of the fashionable claims. But even black gold needed many more disciplines to get the most out of it. Something similar happens now. The revolution of big data coincides in time with another, that of behavioral economics (as Richard Thaler's Nobel Prize shows). To think that they are parallel worlds would be to arrive at a hasty conclusion. Let 's be friends.