Taming an algorithm. Predictive analytics from commercial device to a Masters degree course. The case of Evo Pricing.
Italy, March 31, 2017
Economics is a matter of numbers but to do business numbers are not enough, human intuition is necessary. This story starts from the idea around which a team of young people tamed an algorithm and made it available to managers all around the world.
Following a ten-year consultancy experience at McKinsey in Italy, USA and England, during which he achieved an MBA at Harvard Business School, Fabrizio Fantini started a new business experience that would reconnect him to its Italian origins.
Evo Pricing was founded in 2013: a big data startup operating in predictive analysis to help companies becoming more competitive. "After starting the company in London, we immediately focused on Turin" said Fantini, Evo Pricing founder and CEO. "We already have customers around the world from Mexico to California, obviously in England and Italy too. In our Turin office, we hire young professionals data scientist, however I have to keep hiring also abroad considering that professionals such as computer scientist and developers for the Anglo-Saxon markets are unfortunately very difficult if not impossible to find in Italy."
Evo Pricing works with world class business experts, data scientists and researchers of international universities such as Harvard and MIT while also looking to find excellences made in Turin. In fact the Piedmont capital is set to become a very attractive international big data excellence center. An evolution made possible by the synergy between companies and universities. Data scientists, in fact, are among the most demanded professional figures in Italy and in the world. A new Master of Science in Data Science is now offered in Turin and it is the first of its kind in Italy. The syllabus is heavily based on mathematics and entirely taught in English. By July 2017 the first students will graduate.
Big Italian institutions, the traditional companies and the startups have no time to lose and started hunting for undergraduates in "Stochastics and Data Science" (official name of this new course) allowing many students to find placements or internship before the completion of their studies. This is the case of Elena Pesce, 23 years old. Elena is writing her dissertation on the impact of promotions on retail sales. In March Elena began her internship at Evo Pricing to analyze the impact of multiple promotions which is a marketing term relating to the availability of multiple promotions in parallel. "The knowledge I am developing in university is very inspiring and stimulating however I'm happy to be able to apply it with real data and customers” says Elena who continues “This is not only an opportunity, but also a challenge."
The Masters’ program in Stochastics and Data Science is a two year program that started at the beginning of the 2014-15 academic year and is currently approaching the completion of its first graduates. The program was initiated thanks to the intuition of Laura Sacerdote, Professor of Probability and Statistics, and a group of academics that got inspired by European and US models. The Turin University program is at the forefront of the international scene in data analytics. "The offer is multidisciplinary, entirely taught in English and thus attracting a great pool of foreign students and teaching staff. This is the reason behind our decision to base our Italian office in Turin" explains Fabrizio Fantini, who does not hide there is a healthy competition among companies of all sizes to hire these professionals.
This is the first academic offer of this kind made in Italy - applied mathematics, probability, statistics, machine learning and computer science are essential skills for emerging professional people in an era with an enormous data availability. This foundation will enable graduates to build and manipulate mathematical models under uncertainty (Stochastics) and analyze large datasets and complex structures (Data Science), leveraging the deep understanding of the underlying mathematical structures.
The university collaborates with innovative research institutes and companies; for example, Evo Pricing regularly organizes meetings and seminars dedicated to students: "As early as three years before Amazon announced the opening of a data science center in Turin we sensed the great potential of this area" adds Fantini, who is also a member of the steering committee of this new course of study, along with other fifteen members, from McKinsey to Intesa Sanpaolo.
"Artificial Intelligence will not replace humans, but the collaborating between the two is the winning move". Can sales assistants in apparel shops forecast future sales better than big data analytics? This may very well be a “miracle” that would be extremely useful for businesses throughout the year, not just during the holiday season or in end of season promotions.
Evo Pricing found the winning formula: that is the combination of artificial intelligence and human experience, even in retail, which outperforms any computer system. Algorithms are fed with data, and as such bounded to what was observed in the past. Instead, operators can use their experience to quickly formulate hypotheses about the future. In a dynamic world like fashion, where speed is essential, the algorithm alone is therefore less powerful, without human intuition.
Every week over 300 people including retailers and store managers in Italy alone contribute to Evo Pricing’s system which is built on a solid input system, based on non-monetary incentives and a simple process. "Human intuition improves the results of artificial intelligence, their combination reduces errors by 40 percent, decisions become more accurate and sales grow by more than 10 percent" ensures Fabrizio Fantini. His intuition was triggered after years of work at McKinsey, alongside top managers around the world. "By watching the untapped potential inside companies I got the idea" he adds. Variables such as past sales, geographical area, climate and characteristics of the product are used for the initial forecast based on data. But then, thanks to a real-time platform, stores can change this forecast, and even exchange goods directly between different stores as well as improving inventory management, promotions and marketing strategies.
It is not sci-fi: just like in the famous movie Next - inspired by the short story “The Golden Man” by Philip K. Dick - Nicolas Cage says: “Here is the thing about the future. Every time you look at it, it changes, because you looked at it, and that changes everything else.” This cycle of lifelong learning is just what Evo Pricing helps automating.
“In Turin, for example, seasons are different then Palermo” – explains us a store manager – “It’s easy to describe, but with so many shops it becomes impossible for the head office to take it carefully into account. So, in the spring, when the padded warmer jackets aren’t selling in warm regions any more, the flexibility of this new system allows us to take advantage of this information and to stock up to satisfy our real big demands. The results are excellent.”
Another example is Novara, which tested for the first time in their history, the demand for beach products: when the small initial sample has sold out quickly, the system has discovered the opportunity and learned of a trend till then unknown and developed it further.
"We address customers in retail who do not have precise predictions about their customers’ behavior and for this reason miss many revenue opportunities. Evo gives them access to information that allow them to always have the correct inventory thus avoiding sales loss. Additionally, Evo gives pricing services aimed at maximizing margins and reduce waste" explains Marco Palminiello, business developer.
“We are publishing our results with Harvard University, demonstrating that the demand of the “real time platform” can approximately forecast the performance of supply-demand binomial eight week in advance” - ensures Giuseppe Craparotta, senior data scientist at Evo Pricing – “this is a one of kind work, based on effective measures of product popularity for which the human contribution is essential. Many retailers are surprised when they discover that their daily contact with the public and their product is the decisive factor: the algorithm’s soul”.