Marketing is fundamentally about prediction: what does the consumer want, and when do they want it? From New Coke in 1985, the modernisation of the Gap logo in 2010, and BIC’s pink “for her” biros just four years ago, the history of marketing is littered with flops which better consumer testing could have avoided.
Since 2011, Unilever, Disney, Panasonic, Debenhams and other giants have turned to Black Swan, a UK data science startup, for some predictive guidance. By mining and analysing public attitudes on social media, as well as first-party data sets to identify patterns and trends, Black Swan can gauge supply and demand, and predict how consumers might react to a particular product or marketing campaign.
“We sit at the intersection between big data and predictions,” says Jeff Headley, who joined as US managing director in May from Tesco-owned research firm Dunnhumby. “There are a lot of companies doing rear-view analytics. You can do some interesting things with that, but few companies have made the jump to applied prediction. That’s why I joined.”
Black Swan now has offices in London, New York, California, Hong Kong, South Africa and elsewhere. No UK-based SME enjoyed higher international sales in 2015, and just last week, it closed £6.2m of Series B funding in an effort to beef up its “Nest” prediction software platform, and expand its international presence.
Headley now counts PepsiCo as a client. Alongside having access to the Nest – the platform containing all of Black Swan’s public data – it can add in its own data to obtain unique insight into its soda-swilling target market.
The crystal ball
“The internet is the largest real-time focus group in the world”, he says. Using language processing tools, the company can analyse sentiment on social media towards almost anything, to ensure that product design, creative campaigns, and media buying strategies can be adapted to be as efficient as possible.
Read more: Why 2016 is the year of “moment marketing”
“If Disney is releasing a DVD in two weeks, where do they put their marketing? We listen to the social media conversations going on about that movie and look at where the sales of merchandise are concentrated, and we can predict where we think the deployment of marketing dollars would be most effective.”
“If you’re a food company coming up with a new flavour, for example, we can scour the web very quickly, see which flavours are trending, what kind of social media discussions are going on around cookies, and so reduce the time it takes to come up with that new product.”
Black Swan’s newest software product hopes to take prediction even further. Developed with Queen Mary University, Dragonfly aims to assess which elements of an advert are most attractive to the human eye. “When you see an ad, where are you focusing your attention? What do you hover over? Dragonfly can be used to gauge that interest,” says Headley.
Based on neural networks in the brain’s visual cortex, it can process the visual characteristics of an ad, such as orientation, contrast, texture and luminance, and assign a stimulus attention score to each individual pixel.
“However, prediction means little if that intelligence can’t be harnessed”, he says. Black Swan also has an insight team which works with clients to tweak predictive models, and build bespoke apps based on their own data.
“We’ve done some interesting work around shaping brand concepts, to find the right keywords, phrases and imagery which can then be A/B tested very quickly.”
Read more: Rescuing brands from oblivion
Predict to prevent
Black Swan has more of a consumer focus than the US-based Palantir Technologies, which provides fraud, security and disaster prediction to the US government, but clients are not limited exclusively to consumer brands.
Black Swan has also done work for non-profit organisations like the NHS. And even in the commercial sphere, its predication capabilities have been used in diverse contexts.
In 2014, the firm won a share of Google’s digital innovation fund for its work with a pharma company on an influenza early warning system in Indonesia.
“We looked at the areas where people were searching for flu remedies," says Headley. "We took the weather into account to devise a model which, four days hence, could predict where flu and allergies were going to break out. This enabled the pharma company to market remedies ahead of time and provide consumers with an early warning.”