The Origins of Quantcast
This week we are celebrating the seventh anniversary of Quantcast. In honor of this we want to share with you a feature on the origins of Quantcast.
Quantcast launched in September 2006. In the year prior to this, my co-founder and I had been researching the digital advertising market, and we kept coming back to a seemingly simple question: “Why would you broadcast, when you could Quantcast?”
When we broadcast we transmit content broadly – the quantum of delivery is everyone – and everyone receives the same content. But the Internet is not a broadcast medium; everyone has content transmitted to them individually. Yet, outside of search, digital advertising still used a mostly broadcast-like model.
It didn’t make much sense to us. Search was successful precisely because it delivered tailored, data-driven content and advertising to each individual consumer. Outside of search, the industry was using inaccurate aggregate data and human hypothesis to guide advertising delivery. When advertising is less relevant, it’s less valuable, and when it’s less valuable, a media owner must sell more of it to cover the costs of content production. Ever wonder why there are nearly 20 minutes of advertising for each hour of network TV?
We saw an opportunity to change the status quo and take display advertising from a broadcast model to a search-like model with corresponding gains in consumer relevance and advertising performance. And so Quantcast was born, with the “quant” referring both to the quantum of delivery we would enable – individual impressions – and the quantitative methods we would use to create relevance.
The Core Premise
Our central hypothesis was that given a sufficient volume of HTTP requests (the underlying protocol for Internet data communication and present in every request made to a web server) and sufficiently advanced math, it would be possible to discern the characteristics of web audiences. This concept was by no means trivial. Not only had it never been done before, but HTTP headers are themselves inherently anonymous and don’t provide information on the consumer behind the request.
So we set out to test this hypothesis. My co-founder Paul Sutter wrote an ad server one weekend (from scratch!), and I purchased advertising impressions across a few thousands sites from an advertising network. The ads themselves didn’t matter; what we wanted were the HTTP headers received with each ad delivered. The ads ran for a couple of months, and we fed them into our rapidly prototyped analytical engine.
Often the output from prototypes is inconclusive, perhaps suggesting incremental steps to further evaluate a hypothesis. Fortunately for us, and perhaps due to the eternal optimism of entrepreneurs, we interpreted the results as conclusive proof that our hypothesis was valid. Using HTTP headers, a variety of reference data sets and some novel mathematical techniques we’d developed, we could infer audience characteristics for thousands of web sites!
The output from this analysis was rough, at best directional in nature and offered no clear path to the detailed composition reports and insights that Quantcast delivers today, but our first glimpse of these results stands as our “ah-ha” moment, that instant when we knew this could be big.
We identified three significant challenges to productizing our prototype at Internet scale:
- Developing sufficiently advanced analytical methods (we’d only scratched the surface in our prototype)
- The need for massive data processing capabilities (now referred to as Big Data)
- A sufficiently large set of fresh HTTP headers to conduct the analysis on an ongoing basis.
We were confident #1 and #2 were within our grasp: I’d spent the prior decade applying machine learning to large data sets, and my co-founder Paul had extensive experience creating large-scale computing platforms. So we set to work on these while pondering how we would solve #3.
The Collaborative Model
The solution to #3 emerged over time. With many friends at consumer Internet start-ups, we knew that website owners didn’t care for the inaccurate, sample-based estimates of site traffic commonly used in the industry. The disparity with site owners’ own web logs, their HTTP headers, was glaring. Through our research into advertising we discovered that buyers of digital advertising had just as much disdain for these services as the site owners. It became increasingly clear to us that real-time advertising was inextricably linked to, and couldn’t work effectively at scale without, real-time measurement. “Ah-ha” moment #2: we might attain the scale of real-time measurement necessary to influence an industry’s shift to real-time advertising—all by changing the status quo in digital audience measurement.
In September 2006, we launched Quantcast Measure as an open ratings solution that allowed any web site owner to directly participate by placing a Quantcast measurement tag on their site, and we offered this service for free. Data from all participating publishers would be pooled to improve the accuracy of the analysis for all. Central to this collaborative approach was the unique value of each publisher’s audience being carefully protected, principles that stand today. Publishers choose the extent to which they want to share data publicly, no individual visitor to a publisher site is identified and no one is permitted to retarget any publishers’ audience.
Our approach was classic disruptive innovation. At first, we could accurately present information on the volume of site visitors, visits and page views to an individual site, and little more. But this was just good enough for a set of early adopters who desperately needed third-party validation of their traffic. Then, as more publishers participated, our data visibility grew, and we began to construct a visit-graph for the Internet – a model that helped us understand the dynamics and nuance of visitation patterns and audience commonality. As this visit-graph grew, our algorithms could start to discern and quantify key audience characteristic. We even used this as our slogan: “Get Quantified!” And the service became valuable to a broader set of publishers and web site owners, who in turn began to participate. The visit-graph grew again, accuracy improved and we were able to introduce yet further audience analyses. We committed 100% of our energy to fueling this virtuous cycle.
A Clear Vision
We made a pivotal decision early on not to get into the paid research market. Many venture capital firms implored us to take their money and start charging for the superior measurement solution that we were giving away. For sure, market research was a multi-billion-dollar market and was hugely influential in how the then $600B in annual advertising expenditure was allocated. But we believed wholeheartedly that the value of real-time impression data would be at least an order of magnitude greater than the value of the retrospective aggregate data that was then the market norm.
Many potential investors agreed with this long-term vision, but they just couldn’t get past the temptation of selling aggregate research data as a starting point. We believed that our big vision required absolute focus and to build a sales and marketing operation to pursue the research market would have been both a distraction and a severe impediment to the widespread adoption of Quantcast Measure. We were fortunate that circumstance allowed us to grow the business while we searched for an investor who was aligned with our vision and path. We discovered that investor in Founders Fund and raised our $5M Series A in early 2007.
The Dawn of Quantcasting
For the next two years we focused maniacally on building Quantcast Measure into the world’s best audience measurement service. Not the world’s best free service, but the best service period. The market responded and participation grew at an astounding rate. Today millions of web destinations from individual blogs and videos to many of the world’s most well-known Internet brands use Quantcast Measure.
Our vision for commercialization always centered on the concept of quantcasting – making real-time decisions to create more relevant advertising experiences for consumers – and in 2009 we launched Quantcast Advertise.
Come back next week to learn more about the dawn of quantcasting and how RTB changed the game.
If you’re not already using Quantcast Measure, sign up now. It’s completely free and available for your desktop and mobile websites and iOS and Android apps.
Posted by Konrad Feldman, CEO & Co-founder Quantcast