Know your audience: Quantcast CEO Konrad Feldman talks starting out in Silicon Valley and IPO
Over the last decade, the marketing industry has laid down its scatter-gun and embraced big data, creating a virtuous circle which promises to make the ecosystem more efficient.
Extant viewability issues aside, marketers can reach those likely to buy their products, publishers can monetise better, and consumers, in theory, are served ads they are likely to be interested in.
When Konrad Feldman moved to Silicon Valley and founded Quantcast in 2006, “audience measurement” meant samples and demographics, and real-time buying was a dot on the horizon. Having applied his background in machine learning to create terrorist financing detection and anti-money laundering software, he turned his attention to advertising. “I was fascinated by internet data; its scale and the window it provides onto commerce and culture. Data is how I learn about the world. It represents activity, relationships and behaviour.”
Feldman tells City A.M. how he got into audience measurement and why “programmatic eats digital”.
How did you start Quantcast?
Google had gone public and a lot of people in Silicon Valley were discussing advertising, or rather, search advertising.
I knew that search advertising was effective because when we use search engines, we input specific keywords – relevant information about our interests. But when we click on a news article, for example, we don’t present that information, so everyone who visited the same page on a news site would see the same advertising.
I didn’t know anything about advertising, so my co-founder Peter Sutter and I put an ad on Craigslist asking to speak to people in the industry, offering to pay them $20 and buy them a drink. We were introduced to the planning and media buying processes, and saw how marketers were describing the characteristics of the audiences they wanted – their demographics, interests, lifestyle, aspirations and habits. It was top-down rather than bottom-up.
The problem was down to data. We hypothesised that, if we had enough information about internet media consumption, we might be able to use mathematical techniques to understand and characterise some of those audience dimensions.
We tested the idea in early 2006 and found some patterns – whether a website’s audience generally had a higher income than another’s, for example. We began to see the powerful potential and knew that we could substantially refine the analytics and understand audiences in a more descriptive, consistent and scalable way.
What was your thinking behind your audience measurement solution?
To be relevant for consumers, we had to create real-time advertising, which meant real-time measurement. Traditionally, audience measurement had always been done with panels or samples, with statistical estimation or extrapolation of the greater population. That’s because, with radio or TV, there was historically no way of telling who was tuning in.
But as the number of media choices grows, you get sampling error. As industry outsiders, we were able to see that the internet was not like broadcast media. Companies could place a measurement tag in their media assets, and we could observe the media being consumed. It sounds obvious today, but at the time the idea was either revolutionary or heresy.
The first problem was: how do you manage these massive amounts of data? Think about the computer technology infrastructure a company must develop to serve its website. We wanted to capture records of media consumption not just for one site’s pages, but all sites’ pages. Over two years, we were able to build the physical infrastructure as well as adoption and participation from clients.
What kinds of companies were interested?
It was classic disruptive innovation. Initially, it was a good solution for small websites which couldn’t afford expensive research solutions to measuring audiences, because they were able to talk to investors or advertisers with certainty about their audiences.
We started to incorporate more data, improved our modelling capabilities, and introduced new features which were attractive to larger websites and media companies like NBC and Viacom.
To allow publishers to monetise most effectively and advertisers to create more relevant experiences for consumers, the data needed to be applied at an atomic level in real time. More and more products were being bought online and we wanted to build a predictive model of the types of media consumption which would predict those purchases. When we launched Quantcast Advertise in June 2009, the first ad exchange was launching, allowing advertising to be bought in real-time. We were lucky, and we decided to invest heavily in this area.
What is the future of programmatic?
People say programmatic is a maturing industry because it has been going for six years. We know it still has a long way to go because there is still so much irrelevant advertising.
Any advertising which can be digital will be digital. There will come a time where we watch the same TV show but see different advertising. In the US, there are 20 minutes of adverts every hour on network television because none of them are very relevant to anyone.
That means that advertisers won’t pay as much, so networks have to show more adverts to provide the content which consumers want. Digital eats traditional. And programmatic will eat digital.
Have you got plans for an IPO?
There are not going to be many IPOs right now. We’re private at the moment and I think we have tremendous opportunity ahead of us. I hope that we can be influential in how the market as a whole embraces what’s possible with digital.
I think that the best opportunity for us to grow and innovate at present is as an independent company. And at some point, being a public company makes sense in that context. It’s something we’ll do when it makes sense for the company.