AI can give fundies an edge over the market: Invesco

Using interesting data sets and compute power can help investors get an edge over the market, Invesco Solutions director Scott Bennett said.

Speaking at the Financial Standard Advisers Big Day Out (ABDO) in Sydney, Bennett said the fund manager is using the data sets to drive better return and risk outcomes in their portfolios.

"While the fundamentals of investing are relatively straightforward and simple, we do think you can gain a slight edge by using some of the more interesting data sets and compute power that we now have available relative to what we had 20 years ago," he said.

Bennett said Invesco has recently started looking at credit card data to evaluate a company's earning momentum and sentiment.

Rather than waiting quarterly to get information on key metrics like company sales, Invesco is assessing credit card transactions data that happen every day globally.

"The ability to not have to wait means we can actually get a slight edge on the market but it also allows us to effectively get a better predictor of what those earnings announcements will be ahead of them," he said.

Another way Invesco is trying to get an edge is by comparing what the management says on the day to what they have said in the past.

"It's not so much what they're saying in their scripted comments but we like to measure the tone of how the actual Q and A session is going, so when management is talking to sell side analysts," he said.

"We track every single manager for every company around the world, so whether they were working at BHP and then switched over to Rio or whether they were working at NAB and now they're working at ANZ. We effectively track their career and their history, and we track everything that they've ever said in history over time."

Bennett said they do the same for sell side analysts, as the tone of their questions can help determine the sentiment.

"The sell side analyst is probably where we get more insight than from the management, because we track every single analyst and we can measure the questions they're asking a particular company relative to questions they've asked other companies around the same time," Bennett said.

However, he said strong compute power and scalability is required to process these large data sets over time.

Bennett also said an investor must have an economic intuition when looking at any of the data that comes in. He gives the example of Glassdoor data.

While the initial proposition with the Glassdoor data set was that if a company has a lot of positive ratings from employees, that will likely mean it's a positive company with a good culture and will provide positive returns. Bennett contends the data shows something completely opposite.

"The worse your Glassdoor rating was, the better your share price performance was," he said, noting the narrative was the company mistreats their employees because they are rewarding their shareholders.

"That's the type of data set that we're very, very skeptical of and doesn't hold a lot of economic intuition. Interesting data set, but not really useful for how would you actually go about building a long term fundamental portfolio," he said.