Ottawa firm’s vote-forecasting system trumps competition

After tracking U.S. election trends using patented artificial intelligence software, Ottawa company is ready to make its call on Donald vs. Hillary

With one of the most divisive presidential election campaigns in memory being waged south of the border and retailers around the world seeking every edge they can get in an increasingly fragmented marketplace, business has never been better for an Ottawa firm that predicts how voters and consumers will behave.

Advanced Symbolics uses artificial intelligence software designed at its York Street head office to forecast human behaviour, whether it’s which candidate U.S. voters will choose on Nov. 8 or what products residents in a particular Toronto neighbourhood are likely to buy a month from now.

The company’s patented AI technology has been used in more than 100 elections in Britain, Canada and the United States over the past five years, and corporations ranging from Bell and Disney to shopping mall owners Cadillac Fairview and Morguard are turning to Advanced Symbolics to get inside customers’ heads.

“There are very few companies that are using artificial intelligence well in order to predict human behaviour,” says CEO Erin Kelly.

Ms. Kelly says the company has seen “huge client growth” over the past year. Advanced Symbolics’ revenues have ballooned 400 per cent in the past 12 months despite a major reorganization that included the venture ditching its old name, ZeroPi, and jettisoning its call centre after one its co-founders left the firm.

But the changes also made Advanced Symbolics – which dropped from 21 to 11 employees as a result – a leaner and more efficient enterprise, Ms. Kelly says. 

Sharper focus

“Because we focus completely on (artificial intelligence) and we’re experts in that, we are able to do market research and predictive work very cost-effectively,” she explains.

Using a representative sample of 200,000 Americans of voting age chosen to reflect the country’s ethnicity and a host of socioeconomic factors, the company’s election-forecasting system, dubbed “Polly,” regularly tracked social media and blog posts leading up to the presidential vote in an effort to take the temperature of the typical U.S. voter.

Polly has even been programmed to understand sarcasm and other nuances of the English language.

“If somebody says, ‘I love Donald Trump’ or ‘I like Donald Trump’ or ‘Donald Trump is OK,’ how does that translate in terms of voting intentions? That’s something that the (system) learns over time,” Ms. Kelly says.

Polly also scoured thousands of online news sites and relevant blogs to help determine which way the electoral winds were blowing. For the record, Ms. Kelly says, the system is calling for a Hillary Clinton victory – albeit by a slimmer margin than most polls are predicting.

“Are Americans ready for a Trump presidency?” she says. “Our AI is saying no. Now, it’s going to be closer, I think, than people believe.”

As high-profile as campaign work might be, it accounts for only a sliver of the company’s total revenues. About two-thirds of Advanced Symbolics’ business comes from shopping mall owners, major brands such as Disney or ad agencies working on behalf of clients like Bacardi.

The company’s AI system, called “Penny,” tracks product reviews and consumer reactions on social media to predict which of those clients’ offerings are likely to be hits and which ones are doomed to be duds.

“She learns very quickly how to correlate what people say about the product and whether or not that translates into sales,” Ms. Kelly says.

Penny can even zero in on specific locations it sees as particularly promising. If a bunch of people living near St. Laurent Shopping Centre seem especially excited about a new store, for example, that retailer can target billboard campaigns for that neighbourhood.

“It can actually tell you not only what your sales will be, it tells you what your sales could be if you marketed differently,” explains Ms. Kelly. “It’s very forward-looking.”

The professional forecaster of human behaviour is predicting a bright future for her firm. But she cautions that the most accurate predictions in the world don’t mean a thing if retailers aren’t nimble enough to act on them.

“There’s no point in telling you you’ve got a $1.5-million opportunity in Toronto next month if you can’t get your marketing materials out the door fast enough,” she says. “A smart company recognizes that it is going to want to be able to capitalize on that.”