Biz Dev in NYC by Rob Tsai

Scott Karp poses the question at Publishing 2.0 - “What is The Value Of Online Display Ads?”   He writes: 

Without clicks or some kind of response mechanism, how do advertisers determine the ROI of their display advertising? One reason why Google’s AdWords program has been so successful is that the ads use the format of the universal call to action on the web — the text link. Unless a display is itself a call to action, which usually involves text that calls upon users to click through from some benefit, there is no way to really know what impact that ad had. Many advertisers use random sampling “brand lift” surveys through companies like Dynamic Logic, but in a medium so rich with data, this is hardly an exact science.

One possible challenge with survey companies and methodologies is sample bias. Are people who have the time to fill out surveys on brand impact of the ads they see different from the vast majority of people who ignore these surveys? How does this skew the data?

Yes, click-through rate is the most direct and measurable metric from an online ad, but it’s clear that online display ads provide value above and beyond the click through. For example, a user may be exposed to a BMW banner ad while reading an article on a news site. If that user later types in a search query on Google for “BMW car”, after being exposed to the BMW car ad, the search engine gets credit for the click-through, though the display ad played a role in increasing awareness.

Yes, this challenge has been around for a long time, as articulated in Julian Smith’s article in clickZ in April 2005.

Measuring direct converters often comes at the expenses of measuring the equally important target consumers who don’t immediately convert, convert offline, or never convert at all… With greater online activity, marketers should no longer rely on click data and conversion metrics alone to provide an understanding of return on investment (ROI). Increasingly, they must take a holistic view of performance and measure the greater, aggregated effect their online messages have on consumers’ minds, offline behavior, and overall attitudes if they are to fully optimize campaign performance… As the medium matures, a wider range of traditional measures and metrics will become important. Reach and frequency are offered by the key audience measurement companies, such as Neilsen//NetRatings with WebRF and comScore Media Metrix with Campaign R/F. Those measures will enable marketers to build campaigns against branding metrics, such as gross rating points (GRPs). They’ll make pre- and post-campaign management more analogous to and comparable with offline advertising. Brand engagement measures, gained through an understanding of advertising or Web site content interaction, will enable comparisons with exposure to 30-second TV spots. Marketers should look to adopt analytics tools and interpret and act on behavioral data.

Probably the best third party data sets out there on measuring the impact of an advertising campaign can be found in a few data repositories - the search engine databases at Google, Yahoo and Microsoft - and the databases at research companies like Hitwise who mine the traffic logs of Internet Service Providers.

Imagine if an advertiser like BMW had visibility on which of their advertising campaigns (and on which publishers’ sites) were viewed by consumers who later typed in a generic search query like “new car” or better yet a branded search query like “bmw lease.” All of a sudden, you have new metrics like % Ads Associated with Generic Search Queries or % Ads Associated with Branded Search Queries.

Or better yet, imagine if you were able to tell what portion of your web visitors were exposed to which advertising campaigns? Leading behavioral targeting providers like Revenue Science and Tacoda are enabling publishers to target ads to profiles of users based on these users’ prior behaviors. For exmaple - Did they visit an auto site in the last 10 days? If so, let’s serve them an auto ad even if they’re on an email page view, and so forth.

Does the technology exist, and is it economically viable, to turn the problem on its head? Instead of looking at a users’ cookie to see what behavioral data is known about that user in order to determine what ad to show him, would it be possible and worthwhile to look at the cookie data of incoming visitors to an advertisers’ site and “see” what advertising campaigns were shown to that visitor? At some point, the technology probably breaks down. The cost of storing, analyzing and mining the data (terabytes worth) exceeds the value of the added insight. But you never know, with Moore’s law breathing down our neck….

With so many cool analytical problems to solve, it’s no wonder why so many scholars, business folks, engineers, technologists and mathematicians are attracted to the media space.

Techememe picked up a great article in the New York Times, “Reaping Results: Data-Mining Goes Mainstream”

Internet marketing and advertising is a social market made for the use of heavy-duty computing and sophisticated mathematics. Investment and start-up money is pouring into the market, and so are many high-powered computing brains.

Basem Nayfeh has a Ph.D. from Stanford, where he did his graduate research down the hall from one of Google’s founders, Sergey Brin. Mr. Nayfeh’s thesis was on multiprocessor chips, and he has worked in corporate labs in Silicon Valley on things as diverse as climate and computer design.

Today Mr. Nayfeh, 37, is the chief technology officer of Revenue Science, which tracks, analyzes and predicts online behavior to help advertisers find people most likely to buy their products. Many of his fellow computer wizards are in online marketing.

“If you asked any of us 5 or 10 years ago if we would be in advertising,” he says, “none of us would have said yes.”