Algorithms are set to change your future targeting efforts
From robo-journalists to robo-bricklayers, artificial intelligence (AI) is after our jobs – but along the way, it’s also set to revolutionize investor targeting.
Machine learning for targeting is still in the early stages but one firm that already has more than 180 corporates signed up to its beta system – ‘including three of the 10 largest companies in the world’ – is Intro-act. Peter Wright, co-founder of the company, explains that while traditional targeting techniques have been very much a qualitative and subjective process, AI takes an approach that is all about the data process – one he likens to the evolution of Netflix recommendations. ‘Twelve years ago, Netflix would say, You’re a five-year-old boy, you must like programs about dinosaurs and sharks,’ he explains. ‘Those recommendations had nothing to do with studying that boy’s behavior.’
And it’s the same with targeting. Rather than simply look at who is investing in your peers, or which funds say they’re focused on dividend-paying stocks, Intro-act’s algorithms take in 600 different factors to predict which institutions are most likely to buy, sell or hold a stock in the next 90 days.
The company had a goal to pass 50 percent accuracy before its beta launch, which it did in just five months. ‘[As of December 1, 2016], across the entire Russell 3000 we had a system that can predict buys, sells or holds, on a back-tested eight-quarter basis, with around 52 percent accuracy,’ says Wright. ‘Our goal is to go above 70 percent in the first 18 months of beta. That’s when we’ll go from beta to commercial.’
Which brings us to the question of cost. While the details haven’t been ironed out yet, Wright says the company plans to launch with three levels, the first of which will offer IROs access to a selection of targeting recommendations for free.
But it’s not just Intro-act getting in on the, ahem, act. Q4 is working on its own AI targeting service, which Adam Frederick, senior vice president of intelligence at the firm, says is ready – with 70 percent accuracy – and awaiting a launch date.
Frederick echoes much of what Wright says. Traditionally, ‘there’s not been a whole lot of science behind [targeting techniques]. The only analysis behind it is however much research someone did into what names made the most sense,’ he explains. With AI, however, ‘[targeting] becomes much more predictive, rather than this idea of filtering for which funds best fit your profile.’
Q4 has looked at data going back 10 years and says it has thrown up some surprises. A fund might say its mandate is as a low-turnover, high-growth biotech fund, for example. ‘But when we look at the data over the past 10 years, we might see something else,’ says Frederick. It’s these findings that can make all the difference in the targeting process.
And like Intro-act, Q4 is also looking at a 90-day timeline; it seems that’s the Goldilocks zone. A shorter timeframe wouldn’t give IROs a chance to do any actual targeting, while a longer outlook would sacrifice accuracy. ‘You can point the algorithm at anything you want, but we found that 90 days is the optimal timeframe,’ says Frederick.
IROs will no doubt look at mixing in tech aspects with traditional targeting techniques as AI becomes more mainstream. But does this technology have the ability to make the old ways obsolete? ‘Absolutely,’ Frederick affirms.
This article appeared in the summer 2017 issue of IR Magazine