Back in 2005, Konrad Feldman was looking for his next challenge. He had recently sold off his company Searchspace Corp, which specialised in detecting finance fraud in an increasingly digital world, and he had no clear plan of what he wanted to do next.
Around this time, he moved to San Francisco where he started to take a particular interest in the way digital advertising, particularly search, was changing the landscape.
Google had just gone public, and the quantity of ad spend chips spent on advertising were already growing into the giant stacks that have turned the company into one of the high rollers around the digital table.
Feldman says that from those early days, it was clear why Google was winning the game.
“I understood why search worked,” he says. “It worked because it applied data to create a relevant experience. It was modifying results in real time.”
The problem Feldman immediately identified was that this was not happening in other digital mediums, despite all the promises of perfectly targeted advertising being made.
“As I got into it, I realised that the problem was data,” he says.
“The data you have on search is presented by the consumer when they enter their search item. If you move out of search, that data doesn’t exist. And that was the starting point for my business, Quantcast.”
Feldman says his aim from the early days was to create a means by which the levels of relevance in search advertising could also be applied to other forms of online advertising.
The trick for Feldman lay in finding a way to connect the desires of the consumer with the type of advertising they wanted to see.
Of course, Quantcast isn’t alone in looking to achieve this objective. The promise of virtually every programmatic provider is that it provides perfectly targeted advertising specific to the individual web user. Despite some decent progress in this space, Feldman believes there’s still a long way to go.
“If anyone knew exactly how to do that, they would do it. And that’s why we have the experimentation at the moment. We have many, many years to go.”
Earlier this year, New Zealand was given a front seat to Quantcast’s continued drive to solve this issue, with the company opening a local office and appointing former Vevo head Brendan Muller as the sales director and Brook McGregor as sales manager.
Feldman says the timing for this move boils down to the positive response the Australian team (launched in 2015) was receiving from local marketers and agencies.
“It’s a really interesting opportunity,” he says. “The market here is eager to try things out.”
It’s all about the AI
Although Feldman is optimistic about where Quantcast is headed, he believes programmatic advertising is still in the discovery stage and that it’s set to evolve quickly over the next few years.
One thing that’s he’s certain about is that success is not going to be achieved through human optimisation. Relying on humans to enter the correct data is simply too slow, he says.
“There’s no doubt in my mind that the future of optimisation will be owned by computers,” he says. “But that’s a good thing because it’s going to free up people to think creatively about strategy.”
For this reason, he’s built the Quantcast system with machine learning at its core.
“It’s not based on the human’s hypothesis,” he says. “Our technology actually learns the ideal patterns of behaviour uniquely for each advertiser and keeps track of changes. The reasons why people might be interested in buying a certain product or service might vary over time and this platform is able to keep up with those changes and keep up with consumers.”
The system follows online users in real time, tracking their habits to establish whether they’re in the sweet spot for an advertiser.
“Having a live audience, where you’re able to update your models and make decisions in real time really stands out in terms of the amount of scale we can deliver,” he says.
Feldman says it has some parallels to Pandora’s Genome Project recommendation engine but adds one important differentiator.
“The Pandora engine relies on a human’s interpretation of music categorisation and interpretation, whereas we don’t pre-categorise anything. Our models are actually built on massive amounts of raw data that generate patterns.”
Quantcast’s live-tracking chews up a mammoth 40 petabytes of data a day, a figure so enormous that Feldman and his team were forced to develop bespoke filing architecture, dubbed Quantcast Filing System, just to house all the information.
But following people across the internet also carries some risks, as Quantcast learned in 2010 when it paid US$2.4 million to settle a class-action lawsuit alleging the company secretly recreated tracking cookies after users deleted them (so-called ‘zombie cookies’).
However, zombie cookies aren’t the only digital pests Quantcast faces as the company continues its journey. Marketers and agency folk are starting to ask some serious questions about the effectiveness of the measurement and targeting methodologies employed in digital advertising.
But having spent much of his career working in the innovation space, Feldman says these questions are nothing new. He says that people always question technology in its early stages, pointing out mistakes and condemning inefficiencies.
“As the industry starts developing, it generally starts aligning around certain standards that have consistency,” he says.
But progression to a clear standard, in this case, hasn’t been linear because of all the forces at play in advertising.
“You’ve got a lot of different fields that really need to play together well to make good advertising,” Feldman says.
“There’s creativity, statistics and machine learning, data management, psychology and strategy. That’s the thing about really good advertising, it’s the culmination and integration of all these disciplines.”
So, in this sense, winning the algorithm battle will essentially require an integration of the human mind, and all the contradictions, hypocrisies and insecurities it brings with it.
Easy enough, if you didn’t also have to find some room for the heart.