Over the past decade, the fashion industry has been shaken by the advent of online retail. Brands are now competing for an ever-increasing audience on a global scale. Today more than ever, retailers must therefore know exactly what consumers want, and when, in order to stay ahead of their competitors. And many of them are starting to see artificial intelligence as a weapon of choice to achieve just that.
daco is a French SaaS startup launched in 2016 that provides retail companies with real-time data on their market and competitors. At the heart of their activity lies a proprietary AI tool, built from scratch, and enhanced with deep-learning capacities.
We sat down with co-founder Paul Mouginot at daco’s headquarters in Paris to discuss why they went for the fashion retail industry, how they designed their product, and why they keep seeking new ways to improve it.
Automating competitive benchmarking
Here’s something you might not know: as they compete for a share of your spending, retail companies – be they brands or distributors – routinely conduct competition analyses, season after season, in order to get a sense of where they stand in their market. Their questions are simple, yet fiendishly complicated to answer. Is their selection on point? Did they miss a hot new trend? Is their pricing well adapted?
As former strategy consultants working for retail clients, Paul and his future co-founders quickly realized that such benchmarks were actually done by hand. Junior employees were tasked with picking small batches of information, online or in stores, and with drawing conclusions from that data. A tedious task which could only yield partial results. And yet, these insights were vital to any retail company. That’s when they saw a gaping need for smarter benchmarking solutions and decided to create daco.
Choosing technologies with a lasting impact
The next natural step was to define how best to tackle the job. For daco, it was pretty obvious, thanks to recent developments in the field of AI and visual recognition.
One thing, though. When gathering the building blocks of a new product, they knew how important it was for teams to define what tasks the final software should perform, which technologies arel be more appropriate… and which are most likely NOT to go stale within a few years.
Technical debt is no joke: many startups have made the mistake of coding everything in the hot-new language at the time, only to face high update costs down the road. daco thus chose Python, Node.js, and TensorFlow-derived technologies for machine learning; to keep up with new developments, they also made it a habit to read research papers on AI and enter competitions on Kaggle.
Understanding clients needs and mindset
When selling AI products, all businesses must be extra careful about giving clear instructions and providing outstanding customer care. That means translating the tech talk into everyday words. It also means paying close attention to clients’ feedback. After conducting several trial-runs with prospects, daco’s co-founders noticed that client teams were often unable to make the best use of the daco dashboard by themselves. Setting up quarterly IRL meetings to answer questions proved extremely worthwhile. While a great tech tool is a wonderful thing, it is key to remember that nothing beats human contact to create value and build lasting partnerships.
Defining your target audience
While looking for patterns in their fashion retail clients, daco’s founders divided the landscape of clothing retailers into four categories, according to production costs and degree of innovation. This helped them realize that their primary customers were actors in the “mass-market followers” category who wished to acquire powerful data in order to better compete with the leaders. They adjusted their marketing accordingly, but without giving up the idea of capturing other segments later.
Building a proprietary AI tool (without spending a fortune)
daco’s co-founders started building their tool in 2016 with two rules in mind: getting good results fast, and keeping costs low. “It was like building a plane,” Paul remembers. “We started with a wooden thing that could fly but not much else, and we improved it piece by piece as we signed clients and made money. Today, after one year of existence, we are profitable without having raised any funds.”.
The team relied heavily on libraries and free resources, which allowed them to spend more on hardware and specialized, key components (with big investments on a high-end GPU). They also devoted time and money to refining their own capacities beyond the existing material, since their requirements in visual recognition – a budding field in computer science – went beyond what was publicly available at the time. Today, they work with freelance developers now and then, and pay for a Tableau license, and that’s it.
Bringing AI and fashion together
While conventional thinking still fails to associate technology with artistic creativity, many companies – startups and GAFA alike – have been working to build bridges between the two worlds. It all hinges on training machines to seek patterns through massive amounts of data in order to produce something fresh. In 2016, Microsoft thus succeeded in delivering “the next Rembrandt”, a brand new painting, AI-designed and 3D-printed, which matches the style of the 17th-century painter with stunning precision.
Even though daco’s work on fashion does not (yet) include helping with apparel design, the three co-founders pay close attention to all tech developments in the field. They watched with interest how the Muze Project led by Google and Zalando in 2016 managed to propose new, AI-engineered designs which demonstrated a true knowledge of fashion history. Today, Paul and the daco team strongly believe that there is plenty of exciting AI projects right around the corner that will empower humans across the whole industry. And they’ll be there to watch.
YOU MAY ALSO LIKE
All companies are eager to send emails and notifications with the right messagen
Blockchain has been deemed by some “the most promising technological innovation since then