Marketing Sherpa’s Chart of the Week this week comes from their 2008 Online Advertising Handbook & Benchmarks. In their article Display Ads Do Work and They’re Getting Better they posted the following chart.

They go on to point out that for years, online display advertising has been a confusing challenge for brand builders and the object of suspicion for the direct marketing crowd. They surveyed 577 online advertisers from a range of companies and there is a lot of no-nonsense information in their report like; “From an ROI perspective, eliminating wasted impressions, then making a good impression by serving up great advertising, is consistently the best option for advertisers.” But from the advertisers perspective I was delighted to see how well contextual targeting scored in the report. They found that 40.5% of the advertisers surveyed felt that contextual targeting yielded a higher ROI while only 36.7% preferred behavioral. This speaks volumes to how well contextual has matured as a technology in the display ad space. Search advertisers like Google have always had the luxury of having their users input keywords to tell them exactly what the relevance is but in the display space it can be much more difficult to find the most relevant ad in just a few milliseconds. To do this you have to run a rich semantic library against an intelligent rules engine to match phrases to meaning all astride a robust data crawler scarping content in real-time. Few ad networks bring this kind of horsepower to the game and usually fall back to simple keyword spotting against a flat database of user generated synsets. When you really put a potent contextual engine behind an intelligent optimization algorithm you can get much higher returns than even what Marketing Sherpa’s survey set has found. In our early case studies we found that we could easily lift eCPM by more than 70% and return a minimum cost savings of 30% by eliminating waste for the advertiser and focusing their ad spend on just the highest performing combinations of hyper-targeted micro-segments. To me this report found that basic contextual is the top targeting method. You should see what good contextual can do.