Last week we explored the origins of Quantcast, the introduction of our free audience measurement service Quantcast Measure and the path leading up to the launch of Quantcast Advertise.
Beyond Big Data + Predictive Intelligence
The concept behind Quantcast Advertise is powerful, yet simple: Rather than rely on subjective reading of a desirable audience in a brief, use math and massive data to automatically determine the appropriate audience on an advertiser-by-advertiser basis.
We believed that the difference would be in the data and that to make good on the promise of relevant real-time advertising, it was essential to directly measure and understand real-time patterns of media consumption at massive scale. This is why we spent the best part of three years building the world’s best audience measurement service, as we described in our post last week.
When Quantcast Advertise launched, adoption of Quantcast Measure was so widespread that our anonymous records of media consumption spanned the entire US Internet population. This enabled us to directly observe an advertiser’s desired outcome (perhaps a purchase on their website) and then reverse engineer the patterns of activity that were predictive of that specific outcome by analyzing hundreds of billions of prior media consumption events. Starting with as few as a thousand outcomes on the advertiser’s site, our technology could stack rank the entire Internet population for the next million or ten million most likely to respond like those thousand (we called this set lookalikes). This subset of the population most appropriate to reach for a given advertiser’s campaign is the most valuable data for programmatic advertising.
It was revolutionary, a completely automated approach for using an advertiser’s own data to construct an advertising plan entirely unique to that advertiser. Better still, the approach was completely dynamic, updating continuously based on changes in the advertiser’s customer base and changes in the entire population’s media consumption habits. It leveraged our Predictive Intelligence to anticipate and influence audience behavior.
We took Quantcast Advertise to market as a data product, providing the list of lookalikes to a media partner who could match them against inventory on an impression-by-impression basis to deliver tailored advertising. Our partners paid a CPM rate for the data when it was used to power an individual impression decision.
Business was good, and while the difference was certainly in the data, when the hottest trend in display advertising emerged it became clear that our definition of data might be too narrow.
The Rise of RTB
Quantcast Advertise launched in June 2009. Just a few months later in September 2009, Google launched the DoubleClick Ad Exchange and the age of real-time bidding (RTB) was born. RTB exchanges conduct auctions one impression at a time with the highest bidder earning the right to show an advertisement. A technological marvel, RTB elegantly aligns interests: The publisher obtains the highest price for each impression, the advertiser buys only impressions they want and at a price they determine and consumers receive more relevant advertisements.
RTB brought with it an explosion in the number of companies applying programmatic methods to the buying and selling of advertising. Our lookalike data appeared perfectly suited to helping them and so in addition to supplying Quantcast Advertise data to the many publishers and advertising networks we had on-boarded, we started working with the new demand-side platforms (DSPs).
The market response was fantastic – Quantcast Advertise was quickly adopted for campaigns across practically every industry vertical, and we went into 2010 with a great deal of optimism for this data business.
We were growing revenue and headcount quickly, but we couldn’t shake a niggling concern. The RTB campaigns using our data weren’t renewing or growing at the rate we expected. Exchange-based buys were typically performance oriented, and we believed that if a campaign performed, it should renew at greater scale. We weren’t seeing the growth that premise implied, so we dug deeper.
Knowing “who” is necessary, but not sufficient
We began to realize that as good as we were at identifying who an ad campaign should be delivered to, the ultimate performance of the campaign required that those consumers were reached in the right place at the right time and at the right price in order to reflect the value of each individual impression for the specific campaign.
Our data was only one component of the total solution, and a customer’s perception of Quantcast was highly dependent upon decisions outside of our control made when a DSP ran a campaign. Many DSPs were combining our data with bidding technology they’d (sometimes) built or (typically) licensed, and then layering on their own logic to execute the campaign. There were many potential points of failure in this scenario, and we wondered whether we could do better.
The bidding technology was a natural for us as we already operated a worldwide real-time computing infrastructure for Quantcast Measure that handled volumes far in excess of the then nascent exchanges. But what about all of the other decisions that had to be made: the sites, placements, time, frequency and so on? Well, it struck us that digital media is data, so we applied our artificial intelligence approach to every aspect of a campaign. We set out not to just select the consumers who should receive advertising but also to calculate where the advertisements should be shown and how much each available impression would be worth to a given advertiser.
We worked around the clock to build our first version of a combined bidder, data management and campaign optimization system and then deployed it side-by-side on campaigns where our lookalike data was also being used by a third party.
For Real-Time Advertising, All-In-One Beats Mix-and-Match
The results were conclusive; our integrated solution delivered superior performance in over 95% of the campaigns – not bad for a v1.0! Most importantly, customers renewed their buys with the incremental scale we expected from advertising campaigns that performed well.
In retrospect, we shouldn’t have been surprised by our triggering concern. Advertising is complex and several distinct decision systems are required to drive performance and scale. Individual components could be independently optimized, but optimized sub-components, when assembled by a third party, do not necessarily produce an optimal whole. By developing all of the components in-house, each had an understanding of the others’ operating characteristics, leading our integrated approach to deliver superior performance to the mix-and-match status quo.
So we took the tough but important decision to cease selling data to third-party exchange buyers. Advertisers perceived Quantcast based on the performance of campaigns using our data, and having seen the side-by-side results, we knew we had to control our own destiny. Since 2011, all exchange-based buys utilizing Quantcast data are executed through our integrated real-time stack.
Turning off the revenue stream that brought our first $10M of revenue was a tough call. But we started Quantcast to build a company that could play a meaningful role in transforming an entire industry, and to us that meant a model that scaled to $1B and beyond. Selling data into RTB was never going to get us there, and while we can’t prove our model will (we’re not yet at $1B), we’re confident we’re on the right path.
Today Quantcast is a growing business with an incredibly talented team of nearly 400. We’re building world-class technology and solving hard problems through innovation.
We have close to twenty data centers, handle as many as one million real-time transactions every second, collate more than one trillion new records each month and process as much as 30 Petabytes of data daily. This compute infrastructure saw us described as “easily one of the top five data processing organizations in the world”. We’ve filed close to 50 patents and have open sourced the Quantcast File System we created to deliver storage, cost and energy efficiency as our data processing needs scaled.
Quantcast Measure is used around the globe, interpreting media consumption from billions of consumer devices across millions of web destinations, from the smallest blogs to the largest international media conglomerates. Today Quantcast Measure operates cross-platform to understand audiences across websites (desktop and mobile) as well as the emerging app economy of smartphones and tablets.
Quantcast Advertise is used by over a thousand of the world’s leading advertisers, both across all the major RTB exchanges and through our sell-side partners – major publishers and premium advertising networks – who create compelling audience solutions for brand advertisers.
Programmatic advertising is in the ascendency. RTB alone is projected to reach $13.9B by 2016 (source: IDC) and in time will permeate every aspect of digital advertising, including television. The intelligent application of data will create ever more compelling consumer experiences and power a vibrant digital ecosystem that promotes, supports and respects the interests of buyers, sellers and consumers. We’re thrilled to be part of this ecosystem and excited for what the future holds.
Posted by Konrad Feldman, CEO & Co-founder Quantcast