Commentary

Web U: Optimization Overdose

DSPs and dynamic ads can optimize your performance away

Demand side platforms (DSPs) are a giant leap forward for Adkind. They put the power back in the hands of the marketer to decide how much to pay for each audience segment, target them in real time, and hyper-optimize the campaign with the help of ingenious black boxes with Einstein-quality math equations inside.

Dynamic ads are the other superheros. They empower marketers to tailor their message or offers based on performance (among other factors). More math equations that put more dollars into marketers' pockets!

Put them together and you get super-hyper-mega-optimized performance, right? Wrong! You get a mess. Here's why: The DSP is optimizing against a specific creative. Let's call it Big-Box Retailer Creative X. As the DSP sees a gradient of performance across different audiences shown Creative X, it optimizes your media buy to bid for more of those audiences. This scenario works great.

Dynamic ads complicate this scenario in that Creative X is undefined. Put simply, the execution of the creative is determined by optimization performed behind the scenes, and whether that means showing pictures of a DVD player or a washing machine is determined each time the ad loads.

So, when the DSP hits on a good audience that it thinks is responding well to Creative X, it doesn't know whether that's actually Creative X.DVD, Creative X.WASHER, or any of the literally millions of combinations possible with dynamic ads. Even if the DSP was able to foresee all the creative executions possible, it would unlikely be able to find enough volume for each micro-segment.

And there you have it: optimization overdose. Neither technology stops spinning plates long enough to get a good read on what the other is doing.

Luckily, there are ways you can avoid this problem and exploit the synergy between DSPs and dynamic ads:

1) Make sure your DSP and dynamic ad provider are talking to one another (programmatically). Essentially, the DSP has to pass audience information through, and the dynamic ad provider has to pass which creative execution it will show. Ask a lot of questions to make sure this is being done.

2) Stay away from non-product centric dynamic ads. It's much easier for a DSP to scalably find audiences responding to DVD players vs. washing machines than "logo position=3X390F" vs. "logo position=3X939Q." Why? Because with good product taxonomy, the DSP can always instruct the dynamic ad to tone down the granularity a step. Instead of "product=DVD player," the ad can be requested to show "category=consumer electronics," so that a broader audience can be targeted if necessary. What's the broader audience for "logo position=3X939Q"? Yeah, I don't know either.

3) Test your DSP with a static control to establish a baseline. Then you'll have a frame of reference when you add the dynamic ad functionality to the mix.

For marketers, myself included, this is a lot to think about at a time when we have a lot to think about. But this is the future of display ads, and each day more marketers uncover the incredible performance (at scale!) that DSPs and dynamic ads (separately or together) enable. Let's make sure the next one is you and not one of your competitors, okay?

4 comments about "Web U: Optimization Overdose ".
Check to receive email when comments are posted.
  1. Peter Marino from reelWebDesign.com, March 8, 2010 at 10:20 a.m.

    Looks like we marketers have to start becoming more analytical than ever. I guess a degree in Calculus can help you now! :)
    <P>
    Peter Marino<BR>
    <A HREF="http://reelwebdesign.com">Marketing Firm NYC</A>

  2. Nick Talbert from MediaMind, March 15, 2010 at 2:43 p.m.

    This will become a relative non-issue if your creative optimization engine and your RTB are operating from the same platform.

    The issue is not over optimization of ads, its using systems that are not truly integrated with one and other. I suggest marketers pay close attention to what functionality will be passed from one interface to another when a creative optimization engine (separate log in) is compatible with you RTB vendor (yet, another log in). Makes more sense that these features would come from a single technology vendor? Doesn't it?

  3. Paul Knegten from Dapper, Inc., March 17, 2010 at 5:57 p.m.

    I suppose it makes sense Nick, but dynamic creative is not easy to do and it's a tremendous development undertaking to nail both well.

  4. Kevin Pike from Kevin Pike, May 12, 2010 at 4:11 p.m.

    For PPC, I 100% agree with this article. I also wish more people could see it this way.

    The way I can tell if someone REALLY knows how to run an optimal PPC campaign is to ask them about dynamic keyword insertion.

    The lazy and inexperienced tend to love it. The people who know how to build an account, and have had their feet held to the online marketing fire know it's not optimal.

    I always think back to the university ad a few years back that was displaying for the search "kill a cop". Needless to say, dynamic insertion was an #EpicFail for their criminal justice degree program.

    Basically, higher CTR does not always = better ROI in a CPC model.

    On a CPM buying model dynamic insertion is more helpful. You already paid for the ad space, so you might as well do everything in your power to get the click.

Next story loading loading..