Someone who didn’t know much about advertising still founded a company that is now helping to revolutionise the industry, all the same. Lesson to be learned.But then that is so often the way, it can take an outsider with little industry knowledge but an awful of lot knowledge about the very thing that the industry needs to transform it.
Konrad didn’t know much about advertising, he did know about artificial intelligence and gaining insights from data.
He was an expert on the application of artificial intelligence, he had even worked on a PhD on neural networks. He knew about machine learning, too, back in the days when it wasn’t fashionable.
Quantcast – a beginning
The origins of the thinking that later became Quancast began 1999, via a party.
Konrad, working on a contract with Wells Fargo, found himself in San Francisco. It was the beginning of the dotcom era, companies raised money and then they celebrated. “It seemed like people were spending 20 per cent of the money they raised on parties.” And one such party, presumably with no shortage of glitz along with popping corks, was thrown by Alta Vista. Back then, Alta Vista was the world’s leading search engine — Google it, if you are not convinced — and Konrad found himself engulfed in a conversation with Alta Vista discussing AI.
Konrad’s mission was to create software to identify the bots. He presented his findings to Alta Vista, they shut the programme down – the dotcom crash had begun.
beginning of Quantcast……
Konrad, a good deal better off, decided he wanted to do something else involving internet data, and figured that everything internet was happening in the ‘Bay’, so he moved to San Francisco.
He took some time off. He got back into coding and looked at ways to apply the computer analysis skills he developed to internet data. He had become friends with Paul Sutter, a serial entrepreneur who was leading engineering at Alta Vista after it had acquired his startup, Transium. They had been talking about doing something together, and in 2006 the company Sutter started after Alta Vista was acquired.
Paul’s background was in what we now call big data, building computer systems that can handle massive amounts of information. Konrad had a background in analysing massive amounts of information. So they agreed to work together, applying AI techniques to internet data.
Konrad developed some analytical software for which the goal was to rank sites in order based on certain characteristics, for example income level.
It was like that with internet advertising, too: only the biggest were noticed by advertisers when lack of data meant it was too difficult to target ads.
So, Konrad had to fill in the gaps. At the end of the testing, he was able to show that there was a link between sales, target audience and rank of sites for that audience and by cross referencing data about information where the businesses resides with census data
Konrad thought he had something. He felt it was inevitable that the advertising world will look more search like such that the experience would be tailored to create more relevance to consumers.
He shut down the other projects, leaving him with a small team of around six people which he called Quantcast.
The idea behind Quantcast was good, but there was a problem. It was difficult to scale — for serious data, crunching was a hardware constraint. At this time, Google published a paper on the paradigm they created called MapReduce — the ideas for which was for a programmer to distribute the load of their program across many computers. Google, itself, wanted to analyse web pages, but Quantcast wanted to replicate what a relational database did. So that was one piece of serendipity.
Finally came the insight and a dollop of courage Konrad was focused on advertising, of using AI to analyse data. But to do, that a lot of data was needed.
Industry, with half a trillion pounds a year spent on advertising, but neither buyers nor sellers liked this critical service. It was an opportunity for disruption.It was disruptive for two reasons: one it was free, while other systems costs tens of thousands of dollars a year. It also allowed any website owner to directly contribute their website visitation information, so rather than have a sample, they can directly measure every page.
And from this, Quantcast worked its way up the market. The data created a network effect. At first, each site appeared as an island, but as more data was collected, a picture emerged of traffic passing between sites, “and that connective tissue – the visit graph – is the data structure that we could apply machine learning to.”
This is classic tech — scale creates a network effect, creating a more desirable product, creating more users, creating a bigger network effect. After about a year from the launch of Quantcast Measure, they got requests from bigger media companies to use the tool.
By the end of 2006, a lot of companies were using Quantcast data when pitching to investors, so investors became aware of Quantcast. The company got approaches.
With support from the Founders Fund, Quantcast raised $5 million from a series A in April of 2007, and focused on building out Quantcast Measure.They raised a further $20 million in December of 2007, but by the beginning of 2009, even patient investors wanted to know when they would start making money. in June of 2009, they finally launched a commercial product: ‘Quantcast Advertise.
“AI is going to transform every industry, it is a bigger shift than the internet was. So when I think about our role, which is AI in the marketing arena, every consumer interaction, between a brand and a consumer, if you can make their interaction more timely, more relevant, then of course the people who do it will have a competitive advantage.”
The Quantcast approach is different. Rather than think of the various things that shows someone is relevant, with Quantcast it is case of “show us examples,” and use machine learning to find all those things that are relevant. “