Online media paired with an influx of data sources has made it easier for advertisers to reach consumers at scale and target the best audience possible to receive their ad messages. But one single piece of creative will not elicit the same reaction with each and every consumer. Real-time marketing has made it so advertisers can alter their messaging and creative elements, depending on the traits a consumer exhibits.
Big data must enable big creative. Pairing real-time marketing with dynamic creative technology is a powerful tool for advertisers looking to deliver specific, targeted messages to their online audience, while adding flexibility and efficiency that were previously impossible.
While real-time marketing is about matching the right ad to the right person at the right time, the tactics and best practices can vary tremendously depending on the advertising vertical and the brand or agency's overall goals. There are several common challenges that arise when brands and agencies begin combining real-time data with creative permutations. For brands that master the strategy, dynamic creative is a great asset that can deliver results.
There are two ways in which brands and agencies currently leverage dynamic content, and they are often confused. The first is dynamic versioning, which introduces efficiency when using multiple creative executions and drives performance. The second -- dynamic optimization -- involves learning what works best and optimizing that result, improving the performance with each new impression delivered or campaign flight.
One easy way to remember the difference is to follow this rule: Dynamic versioning is used when an advertiser has very specific localized messages and offers that might need to be updated frequently. Dynamic optimization is used when an advertiser wants to learn more about their audience by trying different executions (messages/creative) and allows the system to optimize executions for optimal performance.
The ability to dynamically change creative makes it very easy to get carried away and provide too many executions, resulting in too many personalized messages without leaving room to learn what works. Let's look at an auto advertiser that ran into this problem with a series of campaigns that ran on Tier 1 and Tier 2 websites.
The Tier 1 strategy involved eight campaigns, with three ad sizes, and 200 dynamic creative variations per campaign, resulting in 4,800 different creative permutations of the ads.
With the Tier 2 strategy, there were 125 campaigns, three ad sizes, and 160 different creative element variations per campaign, creating 60,000 possible permutations.
Added together, that's a total of 64,800 creative ad variations for one brand. The goal of optimization is to push out one wave with multiple permutations, monitor what drives toward the performance goal, and then try to repeat that result in the next wave. The first optimization effort should come early, following no more than 10 percent of the campaign flight. If this auto advertiser was trying to drive conversions, and wanted to hit a goal rate of 0.25 percent, then they would need more than 100 million impressions just to get enough scale to understand which of the 64,800 executions drives the best results. Following the 10 percent rule, the advertiser would have to buy 1 billion impressions to deliver an effective campaign that leverages the optimization learnings.
Targeting on a granular level that produces thousands of permutations makes it impossible to measure performance lift, which really kills the power of optimization. Optimization engines are built to look for a statistically significant number of positives before they begin to trend in a particular direction. So, let's use interactions as another example key-performance indicator, and use 50 as our goal before optimization. A 5 percent interaction rate (fairly average) means the campaign needs to serve 1,000 impressions to detect the statistical positive trend. With 200 versions of creative created by data points like geotargeting, age, and gender, the minimum impression count becomes 200,000 (1,000 x 200) to hit that 5 percent goal. The number quickly increases when more variations or audience segmentations are added in.
By limiting the number of permutations advertisers let technology learn what works and then push forward the best-performing creative and message, resulting in higher conversions. Advertisers planning data-driven creative campaigns can often reverse engineer, starting with a total impression count and working backwards.
The easiest way to approach optimization and versioning is to put learning first and above all else. Limit the number of variables that can change in an ad so that the technology can learn what works best and then adjust the future waves of ad delivery based on that data. After learning about the audience, the advertiser can experiment with new elements via versioning.
The ability to adapt data-driven creative in real-time is a tremendous asset for marketers. Don't treat the technology like a kid in a candy store and get overexcited at the possibilities, cramming too much into one campaign. Overindulging washes away the potential for concrete learnings that can make incremental yet marked impact. Savvy marketers exhibit restraint and follow best practices, avoiding this common misstep unlocking the potential of real-time, dynamic digital advertising.
Brian Goss is the chief technology officer at PointRoll.
On Twitter? Follow iMedia Connection at @iMediaTweet.