Client: Sugarshots, Inc.
Agency: Basement, Inc.
Ad Network: 24/7 Real Media
Ad Serving + Tracking: Atlas DMT
Site Analytics: Think Metrics
While sipping my unsweetened iced tea today, I thought about how I might come across Sugarshots in my daily life. Would I keep a bottle at work? Would I stash one at home to keep around (and keep me awake) for some late-night work? Would I know anything about Sugarshots if I wasn't asked to contribute an article to the iMedia case study? And more importantly, am I the right audience for this product at all?
Over the past couple of years, we've seen renewed interest in the notion of buying internet media based on audience and not on impressions. I may get audit statements and pass-along circulation figures from a print publication, or a Nielsen demo breakdown for a TV program, but there are few media types better suited than the web to serve ads to individuals or audiences clustered around a specific trigger event or series of events (e.g. visited the weather and gardening sections of site X in the past 24 hours; declared that they were 18-24 and went to see a movie once a week, displayed an 'interest' in entertainment content by visiting a certain number of entertainment-oriented sites within an advertising network, etc.).
After all, as Jim Meskauskas and Bennett Zucker have mentioned in articles for iMedia before, impressions don't buy things -- people do. As technology has evolved, our industry now has the unique ability to ask questions first and shoot later -- that is, we have developed products, services, and features (for individual sites and advertising networks) that all, in one way or another, help better target advertising to individuals based on displayed behaviors.
By determining a person's inclinations based on site or web wide visitation patterns, in theory, we expect to see lifts in CTR and better results for brand recall, favorability, and actual sales.
When dealing in the realm of audience, we must also carefully consider the media tactics that we employ and determine whether or not the media we buy reaches enough of the target audience over time -- one buyer is generally not enough to move your market (unless he happens to be a Pentagon official sourcing a contract for the latest weaponry). We must also ensure that we are not speaking too loudly or too frequently to that audience, and wasting impressions that might be better served against another audience (or used to remarket to our initial segments after a particular action -- a click or a visitation behavior on the destination website, for example).
As described in the last article, the following represents the testing grid applied to the Sugarshots campaign:
|Content Channel||Behavioral Segment|
|Women's Interest||Static (Control)|
|Men's Interest||Static (Control)|
At the end of this testing phase, I was able to review the reach and frequency reporting associated with each content/behavioral segment above, and the following shows those tabulated results:
|Content Channel and Behavioral Segment||Impressions||Reach||Avg Imp Freq||Clicks||CTR|
|Not measured across subsegments|
In every instance, the control group across each content channel showed better clickthrough performance than those upon which a behavioral filter was applied.
While not reflected above, 90 percent of the clicks for this portion of the campaign occurred at the 1x frequency level. The remaining 10 percent of clicks came at a 2x frequency level. No clicks occurred after 2x.
Average effective frequency (where efficacy is a measure of metrics across the brand recall/intent/favorability/purchase spectrum) is typically in the 3-5x exposure range. Overall, this campaign tips the scales at 9x and as we can see from above, certain audience clusters were exposed to a Sugarshots ad 17, 23, 61, and 68 times respectively!
Looking at the data, one could conclude that behavioral targeting doesn't work at all! But that's not true. Depending on how is it used and applied, behavioral targeting's principle weakness is its failure to identify enough people in a particular segment. This campaign isn't working because the segments are too small, there is little if any frequency-capping, and the behaviorally-defined audiences are deluged with the ad way too many times.
If we could turn back time and play this test over again, frequency caps should be implemented all around. Any leftover impressions should be moved into something else. Other content segments could be tested along with varying behavioral targets.
Behavioral targeting can be a powerful tool, but it can't exist in a vacuum. It's not just enough to pick a segment. Reach and frequency must also be taken into consideration when developing a media plan such that you reach the right number of people at the right time in the right context with just the right amount of advertising. Like the product itself, too much of a Sugarshot will make my tea too sweet, while too little won't give me the flavor I'm looking for -- it's a balancing act.
Eric Porres is COO of Underscore Marketing LLC.