Global cosmetics manufacturer L’Oréal has developed a system called TrendSpotter that uses artificial intelligence to analyze millions of comments, images and videos posted online. The aim is to spot trends before rivals do so that L’Oréal can develop on-trend products and incorporate into its online marketing the terms and topics that are engaging consumers now.

New trends in fashion and cosmetics don’t necessarily announce themselves with headlines and brass brands. They may start with signals on social media that are easy to miss, such as a scene from a popular TV show sparking a growing number of comments, or a blog by a popular social media influencer going viral.

By the time they blow up into the latest rage, all the brands selling related products know about it, and no single competitor can get an edge. Now, global cosmetics manufacturer L’Oréal S.A. (No. 19 in the Digital Commerce 360 Europe 500) is aiming to spot trends faster than its rivals so it can quickly respond to consumer preferences.

An example of a trend that emerged unexpectedly a few years ago was interest in calendula oil among cosmetics shoppers, says Charles Besson, global social insights and AI director at L’Oréal Group. Marketers at Kiehl’s, a facial cream brand owned by L’Oréal, did spot this trend quickly through its normal monitoring of online conversations. But if all the company’s brands had been aware of it “it would have been a major added value,” Besson says.

That’s because it takes L’Oréal about a year to develop a product, from conception to showing up on store shelves. Identifying a trend even a couple of months earlier than rivals could give L’Oréal an edge in new product development, Besson says.

The main idea is to make sure we can detect before the competition the trends of tomorrow.
Charles Besson

Aiming to spot future trends and to disseminate those insights throughout the company, Besson and his colleagues at L’Oréal’s technology hub in Paris developed a tool two years ago called TrendSpotter. It scans 3,500 online sources—social networks like Facebook and YouTube, as well as cosmetic-focused online publications and bloggers—looking for what’s new, including “what’s hot now and weak signals.”


“The main idea is to make sure we can detect before the competition the trends of tomorrow,” Besson says.

Charles Besson, global social insights and AI director, L’Oréal

Charles Besson, global social insights and AI director, L’Oréal

The team’s algorithms each year pick up some 25 million bits of data, mostly text and hashtags from posts and articles, as well as words embedded in online images and videos. It’s worth noting that all that data is publicly available, contained in posts consumers and others voluntarily share online. That means access to that information isn’t blocked by privacy laws, such as the European Union’s General Data Protection Regulation or the California Consumer Privacy Act, or by consumers taking steps to keep from being tracked online.

This system is “a perfect use case” for AI and machine learning, technologies that can process far more data than humans can, says John Coniglio, a senior consultant at ecommerce advisory firm FitForCommerce. And, because they search through publicly available content, such as websites and social networks, properly designed web crawlers like this should not run afoul of privacy regulations.


“The process can be built to ignore any or all personal data it may encounter while crawling these websites, and only collect the relevant data it needs to run through the algorithms,” Coniglio says. “The risk of running into privacy issues using this type of methodology is low.”

FitForCommerce was acquired in January by OSF Digital, a technology consulting and implementation company based in Quebec, Canada.

Six countries that influence cosmetics trends

Because millions of pieces of data would be impossible for human beings to process efficiently, the inputs flow into a social listening platform from Synthesio, Inc. that uses artificial intelligence and machine learning to distill patterns from online speech.

The system for now focuses on six countries from which cosmetics trends often emerge: the United States, United Kingdom, France, South Korea, Japan and Brazil. In time, it may be expanded to other geographies, Besson says.

AI detects hundreds of possible cosmetics trends

Those patterns Synthesio detected are turned over to human moderators who assess whether they represent trends that could be useful for L’Oréal product development and marketing teams. Besson says the TrendSpotter system has, to date, come up with between 700 and 800 trends that merited study.

For example? Besson won’t say—after all, the point is to get a leg up on the competition by identifying trends they’ve not yet noticed.

But he does give one example from early in the coronavirus pandemic, when a growing number of consumers working from home started using filters to make themselves look more attractive in video chats.


That played into a L’Oréal strength because the company, having seen the emergence of such filters in Japan a few years ago, acquired in 2018 a Canadian company called Modiface whose augmented reality technology allows online shoppers to see how they would look in various types of makeup or hairstyles.

The sudden interest in altering one’s online appearance led L’Oréal to encourage greater use of that technology by its 40 brands. The tool allows consumers, for example, to virtually try on different colors of lipstick or types of makeup, and to scan their faces for tips on the best types of skin-care products to use.

While that kind of online feature had not gained much traction outside of Japan previously, Besson says, when “the pandemic changed the rules of the game on the cosmetic business, we were ready to push on our internal tech to our consumer to respond to their needs.”

Marketing to online shoppers with the terms they use themselves

Besides aiding in product development, spotting trends quickly can help L’Oréal’s open innovation digital marketing team incorporate terms consumers are using today into their website product pages, social media posts and online ads.


“With Trendspotter, you have access to the hashtags and the main terms sources are mentioning,” Besson says. “All that new information is gold for the digital, marketing and consumer intelligence teams because they can use the right content that comes from the consumer voice.”

In all, 1,800 of L’Oréal’s 88,000 employees worldwide have access to insights from TrendSpotter, Besson says. He says it’s part of making L’Oréal not just a cosmetics brand but “a beauty tech company.”

The TrendSpotter team consists of seven L’Oréal employees and five outside contractors, part of the 30-person technology hub in Paris. L’Oréal operates similar innovation centers in New York and Shanghai that, Besson says, employ the kind of data-crunching specialists that rarely worked at L’Oréal in the past.

“It’s a new era that is happening with a huge diversification of expertise,” he says.


It’s a significant investment, but Besson says, “In the end, if we are able to be ahead of the competition and achieve faster time to market, it will be worth it.”