The whole concept of artificial intelligence sounds like the kind of theoretical science left to universities and research institutes, if not the pages of science fiction novels.
But over the past decade, AI has moved out of its cradle, and is now being increasingly leveraged to help businesses fine-tune their brick-and-mortar operations, from making decisions about inventory, to working out where to market their wares.
That’s where Pelmorex Data Solutions comes in. Aside from using AI to make ever more accurate forecasts – themselves a useful tool for businesses – Pelmorex has the benefit of years of tracking billions of data points, sourced from The Weather Network’s app with the consent of its users. For Pelmorex Data clients, access to that kind of aggregate data lets them know their customers like they never have before.
“Before the age of data, you still had credit cards, you can maybe see what else they’re buying in your store, but it doesn’t really give them a view to what their customers are doing when they’re not in their store,” Pelmorex data scientist Marek Kos says. “Which competitors do they shop at? What do they do for fun?”
On their own, those billions of “breadcrumbs” are just a hopeless jumble. But when fed into an AI model, the AI can analyze and re-analyze them, looking for patterns and trends on a scale that human observation could never comprehend.
Working with an individual client, Pelmorex Data stakes out a “polygon,” containing the retailer and a chunk of the surrounding area that could reasonably expect to be served by it, and then uses the AI to parse the comings and goings of users to find patterns that would be useful for good business decisions.
The kinds of threads that AI can pick out from the gigantic data pool are many, and multiplying, with major intersections between things like demographics, zoning density, time of day and, yes, even the weather.
One major box store chain client that includes snow shovels among its wares used the data to differentiate between stores primarily serving single-family homes, and stores serving mostly apartment-dwellers with little or no need of shovelling. Another client, a golf supplies retailer, found that it had been sending flyers to postal codes with few digital breadcrumbs leading to golf courses, and was able to redirect future efforts toward more golf-friendly neighbourhoods.
For businesses that need to keep a watchful eye on the forecast, whether for traffic or inventory reasons, AI has been a real boon for forecasters at The Weather Network, Pelmorex’s flagship TV station and online digital platform. Backed by data from thousands of data points across Canada, going back decades, forecasting AI has been fed plenty to work with, according to Chris Scott, the Weather Network’s Head of Meteorology.
“Weather forecasting will always have a grey zone associated with it, there will always be probabilities,” Scott says. “What AI allows us to do is to be more precise and sharp with what we think the actual answer is going to be. It’s improving the precision of what we do, and allowing us to zero in on the bullseye.”
Scott says aside from honing the day-to-day forecast, AI is increasingly showing potential in longer-range forecasting, typically a tricky endeavour even for the finest meteorologists.
“We’re starting to look at patterns at that time frame, where the physics-based predictions that we used tend to fall apart, because the atmosphere is a crazy ball of chaos,” Scott says. “AI isn’t going to help us solve the chaos, but it may allow us to see some of the patterns in chaos to give us a better prediction.”
As versatile as the AI can be, however, Kos says it’s not quite magic, limited as it is by privacy constraints on where user data can and cannot be tracked. That includes sensitive areas such as hospitals and schools, or postal codes with very low numbers of users, all with a view to preventing any data point being linked to any particular real-world user.
Pelmorex values users’ trust and privacy, and accepts those restraints willingly — and Kos says the company wouldn’t seek users’ personal information even if it were allowed.
“In general, we’re interested in what PEOPLE are doing, not what one specific person is doing.”