Two years ago, TFR included Heuritech as one of the AI startups that were taking the world of fashion by storm. The Paris-based company has been recognized by Vogue, BoF, Forbes, WWD, Elle and partners with Google Cloud, Microsoft AI Factory, NYFTL or La Maison des Startups (LVMH), just to mention a few.
Fashion keeps adapting their business models and operations to digitalization and covid-19 just accelerated the process. Latest events such as weather extremes, social and political disruptions or Coronavirus reveal we must prepare for more volatility. Uncertain times in our infoxication era require making data-driven decisions. Having the right information is essential and Artificial Intelligence is a key enabler in that sense.
Recently, Heuritech and Pantone partnered to bring the latest colors trends to the fashion industry. I felt it would be very interesting to have Heuritech´s point of view on the future of retail and know their latest innovations and success stories. The Fashion Retailer interviewed Tony Pinville, founder and CEO of Heuritech, and Célia Poncelin, Heuritech´s CMO.
Tony Pinville (Tony P.): Hello, my name is Tony, I am the CEO of Heuritech, based in Paris, a company I founded in 2013 after my thesis in Machine Learning at the Pierre and Marie Curie University in Paris. My background is technical, I was head of research and development in the finance sector for 5 years before my PhD. I am also the founder of Paris Deep Learning Meetup where we gather a scientific community of 5000 members. My passions lie at the crossroads between sports, music, and computer science. I was appointed Young Leaders in 2019 by the French-American Foundation.
Célia Poncelin (Célia P.): Hello! I am Célia, the Chief Marketing Officer of Heuritech, French as well. My background is communication, entrepreneurship and sustainable fashion – I founded my own glasses brand (Lunetist) before joining Heuritech three years ago.
The Fashion Retailer (TFR): What is Heuritech?
Tony P. / Célia P.: Heuritech is a fashion technology company that offers trend forecasting. Leading brands worldwide use us to anticipate the upcoming trends, plan their collection assortment and reduce their uncertainty when launching new products.
This is based on the analysis of social media, with millions of pictures from consumers and influencers processed each day by our cutting-edge technology. We automatically detect all fashion elements – colors, shapes, textures, print – and process it to anticipate trends up to one year in advance.
The company was launched in 2013 and we have been working since then to research and develop the most advanced computer vision dedicated to fashion, based on machine learning.
TFR: What is the entrepreneur story that inspired the launch of the company?
TP: I launched Heuritech after my thesis in Machine Learning at the Sorbonne University. Back then, in 2013, no one was talking about Deep Learning! Our goal was to bridge the gap between scientific research and fashion. We created Heuritech with first the technology in mind, developing cutting-edge machine learning algorithms capable of recognizing anything inside images. During the first years of the company, we tried to apply our technology to different industries, to see where it would have the biggest impact. The click arrived when we met with Louis Vuitton in 2016 – they saw the power of the technology and together, we co-constructed our first product with a use case in mind.
TFR: The Fashion Retailer commented about the social impact, and the short to mid term strategy to face Covid-19 by fashion retail companies. Even stores will have to reinvent their purpose. New Retail is a about adaption to omnichannel and digitalization in an industry where the customer is at the center and information enhances designing, planning, supply chain, retailing, marketing… It´s a 360º adaption from people and processes that are empowered by technology.
How do you feel Covid-19 impacted the way traditional retailers see technology or digitalization?
Tony P. / Célia P.: Brands realized they can no longer count only on their intuition, but they should adopt a data driven approach and integrate external data, such as social media. During the confinement, we helped our clients to navigate with key data of trends that were showing up through our analysis. We could indeed identify changes in consumer preferences for our clients to react. We publish each month public reports available to download for free on our website. Covid has accelerated digitalization of brands because they now realize they can no longer avoid it.
Heuritech´s latest report on Fall 2020 Colors with Pantone
TFR: What is your vision for the future of fashion retail?
Tony P. / Célia P.: Our vision is that no product in the world should be produced without the certainty it is going to be sold, or it is already sold. We are big advocates for sustainable fashion and believe data can help achieve that: by producing clothes that consumers want and producing it locally, on-demand, we can accelerate the go-to-market and radically eradicate overstock.
TFR: Artificial Intelligence in Fashion mentioned some of the benefits of AI that, in my opinion, Heuritech is covering mainly in the designing and planning phases of the fashion business. In Design & Buying, AI analyses trends and customers ‘behavior providing useful information to designers or buyers when creating new collections, selecting colors, styles, etc. Then, Planners will use the information to plan the right quantity, at the right time, in the right place.
How Heuritech helps fashion companies regarding this?
Tony P. / Célia P.: Our software is designed to anticipate the future behaviors of trends in the market up to one year ahead, to make the right decisions. We cover all of the collection process, from product brief, to collection planning, design brief and process. Some of the questions we help to answer are: which trends should I not miss because they will be key for next collections? What share of my collection’s offer should pants represent for next collections? Should I plan more midi skirts in Europe than in the US? When is the ideal month to deliver this trend in store? Is my consumer segment wearing this trend? Is this trend growing faster than this other trend?
Our product is based on both proprietary machine learning and algorithms. We analyze millions of images shared on social media each day to recognize specific fashion trends. By analyzing what consumers and influencers are wearing on social media, we can detect early signals of future trends by one year in advance with a 90% precision. An example of an insight we provide to brands is that the saffron belted dress will increase by 15% next year in the US. Those insights help Design and Buying, as well as Demand Forecasting & Planning.
TFR: Some business, or even IT, executives might see image tracking or social listening as a non-priority project and perhaps more “science fiction”, meaning there is not a clear ROI.
How do you mitigate this bias?
Tony P. / Célia P.: While we understand these concerns, we play in another category: trend forecasting. The value we bring to our clients is to increase sales, decrease overstock, and as a result, increase profitability. By anticipating future trends, they can plan better, produce in the right quantities, allocate the stock to the relevant geographies, and have a more sustainable business. This is already what brands like Zara and Uniqlo are doing by being super reactive and listening to their consumers through data, and they achieved amazing financial results before Covid. To quote McKinsey on this “Top fashion performers are using data analytics and consumer insights to stay ahead of the pack”.
For example, one of the results we achieved for a client was +9% in sales and +13% in overstock in one capsule collection by optimizing assortment and avoiding inventory obsolescence.
TFR: Fashion companies are updating their softwares and PLM (Product Lifecycle Management) is one of the most demanded ones. In summary, PLM empowers fashion companies to seamlessly link the creative areas of the value chain to their business processes—from inception, assortment planning, design, and development to supply chain sourcing and in-season demand-driven design changes.
Data curation could filter and analyze fashion-related content such as colors, shapes, sizes, styles, patterns, silhouettes, shopping occasions, materials…, resulting on strategical and tactical attributes. Designers, and then planners, will work with these attributes to optimize the assortment (right product, time, place, price and quantity).
If I´m right, Heuritech could nurture these systems and attributes with external data (e.g. social media).
How Heuritech connects to fashion retailers’ systems and converts raw data into smart attributes?
Tony P. / Célia P.: You’re totally right, we bring external data to brands. This data is processed by us, with both machine learning and forecasting algorithms trained by our data scientists. So we directly convert, in-house, raw data into smart attributes. To make sure we fit the brands’ attributes, we have developed an advanced trends architecture that covers everything in fashion, so we can pick the relevant attributes for brands, no matter that they have different attributes than another brand, for example.
TFR: In many retail companies, IT architecture and systems are obsolete and updating them is slow and highly expensive, adding that people resist to change. How to convince a CIO to change when he/she has a team of people that have been using the same tool for 20 years? Maybe it´s about talking to the right people like Chief Strategy Office, Chief Transformation Officer, CEO…
In fashion we still see companies with self-made systems that are not ready for the omnichannel or digital era. How does your company face this challenge?
Tony P. / Célia P.: It’s a very good question because change requires time and often, brands don’t have it because they are losing market share.
What should be kept in mind is that companies no longer have the choice: they have to adapt to the new normal, i.e. digitalization of brands, otherwise they will be left behind. To quote McKinsey and Business of Fashion that sum up quite well, “AI-driven innovation may further expand the income gap between high-performing firms and their competitors”. Brands do not want to wait too long.
Our take on that is that it is very important to ask the “Why” question before planning anything. Today, there exists as many technologies as there are use cases in the industry, and everyone wants to be “digital”. However, if a technology does not answer a need, it won’t be adopted inside the company. Each company is different and should first think about “Why” they need this technology before looking for a solution.
This is what we start with at Heuritech to make sure we solve a concrete problem – “why do you need us?”. Then we define key metrics of success to measure our achievements for the collaboration with a brand. Since the beginning, we designed our product to be “independent”, i.e. no need to integrate with the brands’ data. The platform is personalized according to the brand’s needs, and the data is ours, so we are in full control, which allow brands to keep on having their digitalization roadmap while we collaborate together – and usually, it is not the same team that we discuss with than the one in charge of digitalization.
TFR: The startup environment is highly competitive. What is the experience of being part of La Maison des startups (LVMH Accelerator)?
Tony P. / Célia P.: We are very happy that we had the opportunity to be part of La Maison des Startups since the launch. It is a great opportunity to be among an ecosystem of innovators and game-changers. The startup environment is indeed competitive but we are lucky to be one of the first fashiontech company launched, back in 2013, and we also won the LVMH Innovation Award in 2017 which allowed us to have a huge visibility.
We were also members of other accelerators: New York Fashion Tech Lab in 2020, Sap.Io, Microsoft Factory, Adidas Platform A, and Plug and Play. They all boosted us with the relevant connections.
Heuritech and Tony have won the LVMH Innovation Award in 2017
TFR: Data scientists, machine learning engineers, predictive modelers,…, professions are in high demand. How do you manage to find these profiles? How do you keep them motivated as Millennials aren´t afraid to change jobs?
Tony P. / Célia P.: We are teaching at Paris top universities like Ecole Polytechnique, Epita, Sciences Po, HEC and ESCP. This helps us to recruit top talents and we have a very high enterprise culture at Heuritech that allows people to evolve rapidly to the job they want. We have a very low turnover rate and our first employees are still here, seven years later!
TFR: Nike´s change of CEO is a metaphor of what is happening in the industry. Mark Parker, who joined Nike forty years ago (1979) as a footwear designer, steps down. John Donahoe, a tech person and former CEO of eBay will replace him. Parker said about Donahoe that “His expertise in digital commerce, technology, global strategy and leadership combined with his strong relationship with the brand, make him ideally suited to accelerate our digital transformation and to build on the positive impact of our Consumer Direct Offense.”
Apparel, retail, businesses in general, are pivoting to technology. Do you feel Fashion is becoming more Science than Art?
Tony P. / Célia P.: Fashion is becoming a science in the way it is runned by CEOs, but it has always been, and will continue to stay, an art. Let me explain: in recent years, we have seen many companies closing because they were not competitive anymore. “The retail apocalypse” was theorized about the brick-and-mortar stores closing since early 2010s as a result of e-commerce arrival, change in consumer spending and the increased competition with new brands. This is why companies started to be managed more like a science, with data analytics to understand and optimize profits and loss – which is a good thing because it allows for a more sustainable business with less waste!
On the other hand, design is still an art in the way we design clothes. Brands have become of course more customer centric, but for the best, because it helps them to create designs that resonate with consumers. Design process could never be properly done by machines, because they don’t have the creativity and intelligence we have as humans, and they will never be able to replace it.