Market Intelligence Teams Find Commercial Answers Inside Free Public Datasets

The most useful market data in Alberta is already public and mostly ignored. It sits in municipal open data portals and provincial registries as raw files, published without any suggestion of what a business might do with them. Analysts who go looking are finding answers their competitors could have downloaded for free.

Dionathan dos Santos is one of those analysts. He is Senior Analyst for Market Intelligence at Avison Young in Edmonton, where he works alongside the firm's commercial real estate brokers. He started his career in commercial real estate in Brazil and later moved to Canada to study artificial intelligence, learning the client side of the business first and the technical side second.

"There's gold in there, because people are only looking at the big number and not going into the data," dos Santos says. The headline figure is what gets published, quoted and planned against. The detail underneath it stays granular, which is where a commercial answer usually sits.

The question comes before the dataset

Most of this work begins with something dos Santos overhears in the office. A colleague needed to know which small retailers could work as anchors in a strip mall, so he pulled the city's pet license file, mapped ownership density by neighborhood, and added veterinary and pet service locations from Google's API. The neighborhoods with plenty of pet owners and nobody serving them were the ones worth a phone call.

The rest of it arrives by invitation. "If you have a crazy idea, ask me. Most of the time we have the answer," he tells his team. One broker took him up on it and asked where daycare demand sat in Alberta. Dos Santos found a provincial file covering every licensed daycare, including capacity and quality ratings, and built a map showing where demand outran supply and where the existing operators scored badly. Both are openings a broker can call about.

The answer sits between two datasets

Merging files is where most of the value shows up. Dos Santos wanted to understand why Alberta's data center projects cluster where they do, so he combined operational and proposed sites with generation records from the provincial energy regulator and with hydrology files covering rivers and lakes. Sitting close to the grid was the answer he assumed he would get. "They are very close to the grid, but also looking for biomass, wind or solar. And they are close to water, because they still need a lot of it," dos Santos notes.

A retailer selling liquor and cannabis brought him the same problem in a different form. Competitor locations and demographics were easy enough to assemble. Where the retailer could legally open was not, because a licensed store has to sit a set distance from schools under provincial rules and Edmonton's zoning bylaw, and no file on any portal showed where those buffers landed. The daycare data he had pulled for another broker turned out to be the missing piece. "Nobody can give you this without going deep into the database and building a visualization that shows every school and daycare in the city," he says. The retailer gave the firm the mandate for its Alberta openings.

Old data can still be the right data

Public files arrive with two distinct problems. Some are current but built on a methodology that serves the publisher rather than the user, which means cleaning before anything else happens. Others are simply out of date. One of the files dos Santos uses most is a salary breakdown by occupation, which shows what each professional role pays across the province instead of the single citywide average that usually gets reported. It was published in 2023. "It's a matter of judging your datasets in terms of quality, but also in terms of context. Does this make sense for my scenario right now?" dos Santos says.

His test for that file was whether the economy underneath it had shifted. Edmonton still runs on energy and technology, so the sample still describes the city and the numbers hold. The same judgment tells him what to leave alone. Edmonton publishes a dataset on trees, and nothing a broker has ever asked him touches it. Deciding what to skip saves as much time as knowing what to download.

Brokers know the question, not the file

Plenty of people can work a dataset. Dos Santos points out that most of them arrive from IT or data science and have rarely sat across from a client, while the brokers who know exactly what a client cares about have no way to interrogate a file. Both skills are common. Finding them in one person is not.

"If you ask someone that doesn't know how to deal with a client to build a report, they will give you stats or median or average, and not the things your clients are really caring about," dos Santos says.

The difference shows up on the phone. A broker with nothing to offer is calling to ask whether the client happens to be looking for a property, which is a call most clients decline. "It's hard to open conversations with your clients when you don't have anything," dos Santos adds. A report built on that client's own niche changes what the call is, and the ask becomes an hour in the client's office with the findings on screen. That is what the analysis is for.

Dos Santos loads datasets into AI tools and works them conversationally, which means a broker with a question no longer needs someone fluent in the file to get something out of it. He uses the same tools to work out what he should have been asking in the first place. That leaves judgment as the scarce input, since knowing which answer a client will act on is still a human read.

What remains is a confidence problem. The files sit on public portals, the tooling is available to anyone, and the distance from signal to action is shorter than most teams assume. "People just need to trust that it's possible to do more with these datasets," he says.