Every industry claims that its marketplace is innovative and changing at light speed. I hate to throw out this cliché, but it never was more apparent to me than when I saw what transpired since my last article, "Lesson Learned," regarding Google's print ad auction test.
"Rewind," an ad industry task force led by Julie Roehm, Wal-Mart's senior vice president of marketing, announced plans to explore the creation of a Nasdaq-like trading system for purchasing television media. Roehm first hinted at the idea publicly over a year ago during the Association of National Advertisers' March 2005 Television Advertising Forum in New York, when she was director of marketing communications at DaimlerChrysler. This go-around, she brought in big hitters, such as Hewlett-Packard, Microsoft and Toyota to help lead the cause--and a proposed $50 million budget for the test. Shortly thereafter, eBay won the bid to build such a system.
The Future is Now
In my last article, I predicted that the days of advertisers accepting standardized ad rates based only on yearly readership or viewership audits and not bottom-line business results are ending--sooner rather than later.
Looks like sooner just became now.
If the aforementioned isn't enough proof, consider this: Major TV advertiser Johnson & Johnson also announced that it would not buy ad time during this year's TV upfront. Even Coca-Cola Co. hinted it might decide not to buy during this period. Both moves are another indication of how advertisers want to plan and buy media on their terms--with more efficiency and higher return on investment.
Predicting the Future
When I sat down to write my last article, I didn't intend it to be the first in a series of commentaries on ad auction marketplaces. However, who could have predicted that these recent events would happen so soon? This led me to thinking about the practice of predicting.
Previously, I discussed the dynamics of a biddable marketplace and how media buying firms, including search marketing agencies, will have to evolve and adopt new technologies to succeed. I suggested we look at Wall Street and its "auction-like marketplace," specifically how mutual fund managers manage their portfolios with sophisticated software tools and portfolio-management expertise. I'd like to expand on that analogy a little further.
The fundamental approach to managing a portfolio is to predict performance. Mutual fund managers do not buy and sell securities for real-time performance. They make their trading decisions based on a desired ROI and use forecasting--how securities will perform in the future--tomorrow, next week, month, quarter and beyond to help achieve their ROI goals. They are able to do this with advanced, predictive-modeling applications that help lead them to make better-informed predictions. Yet another lesson learned for search marketers, as well as the next generation of media buyers who will be bidding for ad space and time for their clients.
Predictive modeling is the practice of creating statistical models to anticipate future behavior. Think Amazon. In a nutshell, data is gathered, integrated, analyzed and used in a varying degree of equations to forecast future probabilities and trends. It allows practitioners to establish patterns and relationships for making better-informed decisions and reducing risks.
The same principle applies to a "biddable" ad marketplace, including search, where keywords, prices and return metrics change dynamically. We all know that when you place a bid, competitor bids hit the market within seconds. What's important to remember is that it really doesn't matter whether those bids are higher or lower at that moment in time. Mutual fund managers don't care; and neither should SEMers. What matters is, if you can bid effectively on the entire portfolio to achieve the level of ROI you desire.
So how do you do that?
Predictive modeling starts with data collection and mining. The old adage that "history repeats itself" holds true to some extent. In order to make better-informed decisions, you need endless quantities of historical data and the ability to analyze multiple metrics. In search, it's terabytes of data such as bids, keywords (the more popular keywords or the head and the more obscure and inexpensive keywords or the tail), match types, time of day/week/month, seasonality, demographics, brand awareness, linguistics and copy, landing pages, geography, etc., etc., etc.
Where the rubber meets the road is how that data is used. In auction marketplaces, there basically are two fundamental factors that affect accurate decision-making: First, the mathematical and statistical formulas used to analyze mounds of data via probability predictions, regression analysis and so forth; second, the analysis--how the multiple variables of data are used to form connections and simulate outcomes.
The complex analysis is accomplished with optimization algorithms that allow you to generate "what-if" scenarios and tradeoffs to determine the optimal solution. This level of analysis takes millions and millions of data points, models the expected return on all variations of data, and automatically selects the optimal mix to create a reliable and consistent higher rate of return. As a result, you reduce risk and thus, increase the reliability level for performance.
Media buyers beware. An "Adsdaq" system is on the way. Take it from mutual fund managers and SEMers, there's more to buying in a biddable marketplace than meets the eye. It's about being able to see into the future.
Google's print ad auction test. EBay building a TV ad exchange. What's next?