Faberlic has hired an AI that can tell the size of clothes

Faberlic has hired an AI that can tell the size of clothes

Problem


The brand, which has been included in the world ranking of the largest cosmetics companies for more than ten years, has been developing its own fashion direction since 2013, regularly releasing new collections both of its own design and together with well-known Russian designers such as Valentin Yudashkin or Alyona Akhmadulina.


If we talk about the brand name, it can be literally translated as “Master of the image”. And now Faberlic, in an effort to get closer to the buyer and help him create his own unique image, has launched a widget based on artificial intelligence for customers, which will help you choose the right size clothes for any type of figure.


How often, when ordering clothes in online stores, can we miss the chosen option? The same size for different brands has its own nuances of cut. In reality, for example, the purchased jacket does not always meet the standard of the declared one: it may be smaller or larger in terms of length, girth, etc.Incorrectly selected clothing size is what worries every buyer of an online store, and this is the most common reason for returning goods. According to Ivan Bulkin, Faberlic’s business development director, it is very important for the company that “the buyer has the opportunity to accurately determine the selected size of the model that will fit perfectly on the figure, and at the same time reduce the number of returns as much as possible.”


Decision


The Sizolution service (the Tardis company), which Faberlic used, was developed by three graduates of MIPT and Skoltech: Stanislav Podshivalov, Mikhail Matrosov and Vahe Taamazyan. It entered the market in 2016. As Stanislav Podshivalov, one of the founders of Sizolution, noted, Faberlic made a decision on technical integration very quickly. Now the service is integrated into the store’s website. The online customer fills out a questionnaire before buying: indicates the age, height, weight, gender, chest and hip circumference, chooses the type of figure or takes a full-length photo. Next, more than 80 parameters of the human body are analyzed and the widget provides information about the parameters of the figure, which can be adjusted manually if necessary. The program compares all this data with the measurements of a specific thing and selects a model of the desired size for the buyer. As a result, the widget displays recommendations of the optimal size and a description of the fit for the item, giving tips for upcoming purchases. In the future, when viewing the assortment, the user will see hints on the size of a particular clothing model that is suitable for him.


The first thing that had to be done to implement this project was to measure the available clothes. The creators of Sizolution used an automatic meter: this is a table and a webcam fixed above it. The table is located in the store’s warehouse. An employee of the warehouse puts a thing on it, takes pictures of it with a webcam, and the created program uses a neural network to take basic measurements, for example, the circumference of the chest or hips.


Result


Testing of the service from Sizolution showed high results in the accuracy of matching the recommendations of the service with the real standards of people. As a result, Faberlic decided to connect the system to all current and new clothing collections.


Now the Sizolution widget is also installed on the websites of KupiVIP, Bask, Decathlon and Ostin stores. For two years, about 350 thousand items of various brands have been scanned using the Sizolution service, more than 320 thousand customers already use the company’s widget. According to retailers, Sizolution helps brands reduce the number of returns by an average of 15-20%, and increases the conversion rate on websites by 5-10%. However, not all buyers are eager to take advantage of a technological novelty. For example, at KupiVIP, 10-15% of customers use the widget, and so far we have managed to reduce clothing returns by an average of only a few percent.


Other experience


In the world, more than 50 startups are creating clothing selection services to reduce returns. To choose clothes, you need to know the exact parameters of things. After all, on average, customers return about 40% of the purchased clothes to the stores. But mostly existing services work with customer data and do not collect measurements of things. For example, the Texel 3D virtual fitting room is gaining momentum, which takes measurements and creates a 3D model of customers. Such a platform can be installed in any shopping center or offline store, and then through a mobile application offer the customer to buy the clothes that fit the size. Her first clients are TD “Belaya Dacha” in Moscow and three international brands, including Marks & amp; Spencer in London.


However, the stores usually receive clothes from different countries, and each manufacturer has its own size grid. Almost all retailers use a table of compliance with international sizes. Warehouse workers, looking at the label of the item, compare the foreign size with the Russian size according to the table, then indicate the relevant information on the website. But the tables are not always accurate, and brands sometimes make mistakes themselves. At some points, warehouse employees manually measure things to find out the correct parameters, but this is long and troublesome.


However, startups will have a difficult time: the fashion industry is conservative, most online retailers do not want to spend time and money on integrating technological innovations, Lamoda representatives are sure. In their opinion, loose-fitting clothing has become fashionable and the exact measurements of such models can only confuse the buyer. As a result, the parameters of things remain a “headache” for many retailers, and so far the only way out for many is to offer people to try on products.

Go to our cases Get a free quote