Most market research expenses go to polls and client surveys, while only $1.5 billion — as of 2015— are used to analyse what actual people are spontaneously saying on social media like Facebook and Twitter.
“For market research, we shouldn’t actually need to ask anybody to tell us anything,” Michael explains. “We can just get all the information from ‘listening’ to what they say online.”
An aerospace engineer by training, Cyprus-born Michael has nonetheless spent all his professional life in market research. Over almost three decades, he has brought his talent and experience to uncountable firms in several countries: from Cyprus, to Poland, to Hungary, to Germany, to the US—where he started the first online marketing research division at leading market research company Synovate.
Now settled in the UK, Michael—a published author and a revered thought leader in the industry— is championing a new approach to market research through DigitalMR, his London-based company, which, he says, is “on a mission to redefine market research.” Michael founded the company in 2010, convinced that the innovative Artificial Intelligence (AI) technique known as machine learning could radically change the way brands make sense of what the public think, say, and feel about them.
To do so, 20-employee-strong DigitalMR created Listening 24/7, a product able to gauge a brand’s reputation by gathering large amounts of user-generated information online, and automatically assessing its meaning.
“Social media posts—like, in fact, 80 percent of available human knowledge— are unstructured: they are comprised of text, images, videos, whose meaning is not immediate,” Michael says. “What our machine learning algorithms is able to do is give them a structure, and analyse them.”
To train their algorithms to understand social media content, DigitalMR asks teams of “curators” to go through about 20,000 posts, and label them depending on their literal meaning, general sentiment (positive, negative or neutral), and emotive connotations. This data is then looped back to the system, teaching it to recognise the telltale signs of how a brand is perceived by the public. “When a company asks us to carry out a market research, we harvest millions of social media posts about their brands—in any language— and we compare them to their competitors in term of sentiment and emotion, through our machine learning algorithms,” Michael says. “This information is then summarised in a standard report, or in an interactive dashboard format.” Michael says DigitalMR algorithms’ rate of accuracy in assessing content meaning, sentiment, and emotion nears 80 percent—which he maintains is 20 percent higher of market research’s average.
In addition, it only takes three weeks for DigitalMR to train their system to understand a new language, which ensures the company can conduct its market research at a global scale. It is not surprising, therefore, that DigitalMR boasts major brands such as Diageo, Vodafone, Heineken and Nielsen among its clients.
Being in London also helps, Michael says: “It is really a high-energy city, and its time zone is convenient: we can talk with clients both in Asia and in the US over the same working day.”
Since 2012— two years after its kick-start— the company has been headquartered in Workspace’s Vox Studios in South London. “Early days, we chose it mainly because the rent was quite affordable for us—but now we really feel part of a larger community,” he says. “Just think that, right now, we are planning an event for all the space’s tenants, in order to talk and offer advice on how artificial intelligence can be used by any business, no matter how small or large,” he says. The space’s stimulating ecosystem has helped the company to thrive and shine in London’s bustling AI scene.
Over the last five years, DigitalMR has received six innovation grants from government agency Innovate UK to carry out targeted research into Artificial Intelligence technology. Among the funded projects were DigitalMR’s signature technologies for sentiment and emotion analysis. The latest tranche of funding, granted in 2016, went to financing DigitalMR’s new tool to assess the meaning and the sentiment of images (it leverages image recognition to caption them, and then interprets the captions’ meaning); another tool to analyse videos for market research purposes is currently in the works, Michael says.
“Machine learning video analysis would open the door to a lot of new applications. It’s not only good for sentiment analysis: even just understanding if a video is about people using a certain product would be enough,” he explains. “We could analyse videos on social media and understand what is the typical occasion on which people drink beer.”
So, what’s next? Michael hopes that as companies understand the power of AI, the whole market research sector will shift from the current survey-based model, to one predicated on silently “listening” to social media narratives around brands and products. “We have to steer the industry in that direction,” he says. “Our ultimate goal is to completely take over market research.”
If you’d like to join companies like Digital MR and work from one of our state-of-the-art spaces, why not see what’s on offer? As the leading provider of commercial business premises across the capital, Workspace offer a host of attractive offices, studios and industrial spaces in 69 prime locations.