Harnessing Social Insights for Successful Trend Prediction



In order to drive a trend through innovation, it’s vital to be on the forefront, not back, of information.

All sessions from #SMWONE, our four-week virtual conference program, are now available on-demand.

Staying abreast of trends in our industry is imperative – but what does the phrase ‘trend prediction’ really mean? What role does human expertise play in this process?

In this article, we recap a #SMWLDN session featuring Linkfluence CEO, Guillaume Decugis and Danone’s Chief Strategy and Insights Officer, Elaine Rodrigo, who took a close look at these important questions.

Here are the primary insights and takeaways:

  • Relevant and Accurate Results Depend on Refined Data
  • Don’t Overlook the Power of Human Nature
  • Trend Prediction is Costly for a Reason

Rodrigo notes, specifically concerning Danone, “in order to affect health, one really needs to understand how people consume food and drink products. Food culture is the dominant culture out there; breakfast is the most Instagrammed meal of the day.”

Danone works with Linkfluence in order to obtain and translate and interpret trend data – this, in turn, helps them make business-based decisions, for example, introducing yogurt in pouches so customers can make their own breakfast bowls out of them.

Relevant and Accurate Results Depend on Refined Data

The desire to be more accurate and ahead of the culinary playing field sparked looking into AI-driven data prediction.

Danone asked Linkfluence whether it was possible to design predictable models using social data to identify tomorrow’s star ingredients and could they leverage these ingredients to inform product innovation. Decugis interpreted this brief as giving an edge on the market to Danone. Yet, he explains, “whilst we have AI that can predict the end of a time series, and therefore a trend, raw social data is too massive and heterogeneous to monitor everything. When you give [AI] all the data, you’re giving it rogue data. It has to be structured.”

Data has to be refined in order to achieve accurate and relevant results. Algorithms and exhaustive social media social listening apps such as Linkfluence’s Radarly, can learn to identify patterns – these can determine whether there’s a high probability of what ingredients are going to be popular.

Don’t Overlook the Power of Human Nature

“When you have likely ingredient suspects, you can go back to human data. You need to combine AI with market expertise – people who can look at the most interesting way to frame the data,” explains Decugis, “we have a lot of customers in different industries and the same methodology can be applied. It’s about taking a large set of things that could be potential trends and then narrowing it down using statistical technologies.”

Naturally, a trend in one company may not be applicable in another. “We try to tailor what we do by industry, application and by market,” says Decugis. “You have some express data and there’s inferred data. At Linkfluence, we have an interface model that has a 92% precision score that infers the location of people on Instagram or Twitter based on AI prediction.” Rodrigo mentions that Danone asks Linkfluence to look at somewhere between 16 to 20 country’s trends. “Local languages is extremely important. What conversations are happening in a country itself?” She asks.

Trend Prediction is Costly for a Reason

Free trend prediction tools aren’t readily available for non-profit organizations just yet, however, Google Analytics makes a good starting point. “We can’t make low cost because we can’t do it with just software. We need analysts and scientists and expertise and they have a cost,” announces Decugis.

As well as trend predictions, it’s extremely important to be able to catch something before it goes into decline, as well as whether it’s seasonal, and the same tools and methodology is extremely useful for doing this.

What about Mintel?

Rodrigo concludes that Danone hasn’t given up one source of data insight entirely though. “We can no longer make decisions for certain business questions based on one piece of data. It’s all about collecting data, which leads to data analytics. For example, Mintel data is complimentary and will always feed into the ‘what’s next?’ model.

Join 100,000+ fellow marketers who advance their skills and knowledge by subscribing to our weekly newsletter.

Newsletter Subscription

Get the latest insights, trends and best practices from today's leading industry voices.

Learn More

Write for Us

Interested in sharing your ideas and insights with the world? Become a SMW News contributor and reach 300k readers each month.

Apply Here