La Stampa

The project Evo Pricing was born around a thesis

Data, weather and mathematics
A start-up that tells the shops what to showcase in their windows

HISTORY

By Fabrizio Assandri

The Novara store had never sold bathing suits, but an algorithm suggested putting them on display.
It guessed it: they sold like hot cakes. The algorithm was developed by Evo Pricing, a start-up based in Turin which collaborates with the University. It’s very ambitious aim is to predict which sizes and apparels models will customers demand when entering into a store.

Making the customer happy, who is always looking for the right size but never finds it, and even making the store owner happy, who can reduce his stock. A live example is store chain Miroglio, which has a 1,000 stores worldwide, and came to rely on this start up.

The idea came to life based on a thesis written by Elena Marocco, who graduated in July from the Department of Mathematics, and sampled twenty stores as a start. From the first fall-winter season tested, the project objective was to propose the right product, at the right place and at the right time; it has now expanded to 500 stores of Fiorella Rubino and Elena Mirò in Italy, France, Spain and Portugal. This does not say that it won’t reach the other 500 stores as well.

Explaining how the algorithm works is Fabrizio Fantini, founder of Evo Pricing, and a Harvard graduate. Everything is based on "big data", which is the massive amount of data that surround us. The algorithm was developed to manage the stock of the store, and can process 1.2 million variables per week. In practice, it considers 3 factors: weather, competition and historical data of each point of sale.

The algorithm generates a proposal once week for every store: it suggests which sweaters to hold in stock and which ones are most likely to be sold. As in a game, it gives the shops a virtual budget, those with larger stocks have a negative budget, and those with just a few garments have a positive one. This creates a purse of supply and demand around the chains of the store; and the budget can be used to get rid of garments that will never be sold and to stock up on the most "sellable" products.

"So this increases business: every customer who comes into the store and doesn’t find what he’s sought after - it’s a missed opportunity," adds Giuseppe Craparotta, joint PhD in mathematics by the University and Polytechnic, who works for the startup. Evo Pricing has a funding of 500,000 Euros and collaborations in America with a retailer of museum gadgets, in Mexico with a mobile company and in England with an eCommerce store.

"The collaboration with the University of Turin - explains David Garelli of Miroglio - had set off from a study done by a doctorate for the Zara group. We realized to have the same need. For us it breaks new ground, but also for many of our competitors." Garelli explains that so far "the movement and the renewal of stock was done in a traditional way, manual, and we were forced to keep large amounts of immobilized goods." The added value of the algorithm "is the ability to predict what will happen. It doesn’t only tell me that if there are twenty degrees I do not need coats, but it also indicates where, on the basis of past trends and learnings, I have more chances to sell them."

Once stores receive the mastermind proposals, store managers are not forced to implement them slavishly: they may amend the proposals, and then on the basis of actual sales, it will take the risk.

Thus, the AI doesn’t get the better of man, but it cooperates with him. As explained by researcher of the University, Roberta Sirovich, the system combines statistics and "human experience that managed to build a sensitivity comparable to the most sophisticated systems".

The University is investing a lot in big data nowadays: in addition to having a specific degree in "Data Science", yesterday it held, at the Carlo Alberto College, a conference on the various applications already in place for data science, from fighting against cancer, to the study of bank customers.

 

Download full article

Transcript:   English   /   Italiano