20/20 Targeting

Watching Josh Beckett pitch a five-hit, no-run sixth game to stun the Yankees was the highlight of my weekend. He had such incredible focus, every Yankee batter seemed cross-eyed at the plate. The game reminded me of a targeting technique that delivers online creative to users with that kind of laser-like precision.

In this, the fourth in a series on targeting options, we’ll examine performance-based targeting with an overview of Ion Optimizer, one of my long-time favorite technologies. If you were a fan of AdKnowledge, the old ad-serving platform, you may recall this tool.

Getting Creative with Targeting

After Bluestreak acquired key AdKnowledge assets from Engage, the product evolved into Ion Optimizer. With it, creative optimization can be controlled across publishers and across placements.

By dynamically measuring ad creative’s current and past performance in a specific placement, Ion Optimizer can automatically serve the creative selection that drives a chosen metrics goal. Trying to decide between several designs for the same message? Want to test different call-to-action lines? Need to figure out which offer is more compelling to the target audience? Ion easily lets customers vote on the fly.

One campaign can be pitted against another through group placements. You can group placements for one position (even placements across publishers) into “virtual networks” to test response. For example, you could decide to group a number of keyword placements together across publishers to increase volume, statistical validity and results.

One Size Does Not Fit All

Ion Optimizer is a tool that offers a lot of control over the way a campaign is optimized. If you’ve run many campaigns, you know a variety of serving scenarios can affect optimization. Very broadly speaking, two things are necessary to successfully optimize:

1) Differentiated Creative
If a group of similar creative executions run against each other and get the same, or a very similar, result, obviously you won’t see much lift.

2) Volume
A statistically significant number of impressions is required. As a number of serving scenarios may need to be accommodated, Ion uses four separate algorithms fine-tuned for the task:

a) Heavy Serving Volume

  • Millions: 200,000 impressions per day (per placement).
  • Real-time experiments and performance analysis.
  • All creatives run against each other for each test.

b) Medium Serving Volume

  • 10,000 to 200,000 impressions per day (per placement).
  • Even weighting of real-time and historical analysis.
  • All creatives run against each other for each test.

c) Low Serving Volume

  • Below 10,000 impressions per day (per placement).
  • Mostly historical analysis with periodic tests.
  • Round-robin style testing: one-to-one tests, then the winner runs against the next candidate.

d) Post-Event (Beyond Banner) Analysis

  • Specialty engine for less frequent tests and specific behavior patterns of beyond-banner optimization.

Getting Crazy with Metrics

Ion Optimizer supports any metrics mix. What I like most is it’s the only solution on the market that supports optimization based on rich media criteria, such as conversions within a Flash ad. You can even mix an “internal” conversion with an “external” one, including:

  • Click actions.
  • Rich Media actions (one or many).
  • Post Event actions (Beyond Banner). Choose one, or combine post-impression and post-click per event tracked.

You can implement controls over campaign testing, for example by setting the number of impressions required before a change is considered valid, or setting the weight of the primary (winning) creative vs. the secondary (less effective) creatives for testing purposes.

Manual vs. Automatic

Can creative optimization be managed manually by a media planner or assistant mulling over reams of data? Sure, but you can’t come close to maximal results without an automated system. Such a system performs better than a human possibly could. A system can make data-determined changes that are as granular as one every five seconds. Making incredibly fast changes creates a performance improvement loop resulting in incremental gain with each pass. Bear in mind that overall poor performers may perform best at specific times. Removing overall poor performers could actually lower overall performance, because at certain moments they may have been the top performer.

Is It Right For You?

The best way to determine if this type of tool will benefit your team is to give it a try. Along with performance results, make sure to measure the operational effort and costs of optimizing. You may find letting a machine determine the right pitch provides the homerun you’ve needed.

Related reading