Logging In: The False Darwinism Of Competition

Don't make the mistake of playing the network game

How many ad networks on one media plan are too many? If you are running a performance-based campaign, the answer may be two. Does that sound crazy? After all, it's pretty common today to see online media plans that have six, seven, 10 or even more networks on the plan. Media planners include multiple networks with the idea
that they will optimize the plan over time by dropping the weakest performers. The problem with this strategy is that planners (and clients, for that matter) rarely have the data and analytics to make the right decisions, while the networks do everything they can to "game" the system.

The problem starts with the way most planners calculate attribution, or how they assign credit for the conversion or sale. Almost everyone assigns credit for the conversion to the media responsible for the last click or last view before a conversion. And pretty much everybody knows this is not a fair representation of true effectiveness, but for the sake of convenience and lack of a better alternative, they use it anyway. What few understand is that this approach creates a set of disincentives for the networks and sub-optimal results for the advertiser.

When multiple networks are on a performance-based media plan, they will fight for attribution to assure they stay on it. They do this by "pixeling" the client Web site, and bombarding the remarketing pool (i.e., previous Web site visitors) with frequency, in order to increase their odds of getting the last view and thus credit for the conversion. Almost all the networks will go to the exchanges to bid on that remarketing pool, to increase their natural reach and frequency. As they bid against each other, the cost of that remarketing pool goes up. At the same time, they are disincentivized to invest in ads that drive consumers to the client's Web site. Since few consumers convert on first visit, they wind up in the remarketing pool, where the networks know that despite their best efforts, they will only get credit for a percentage of those who ultimately convert. The more networks on the plan, the lower the percentage of conversion they can expect, and the more exaggerated their misdirection of effort.

So what are the options for a planner or marketer?

Well, you could have a plan for proper attribution, and share that with all the networks. This will take away the incentives to game the system. Unfortunately, you need strong analytical capabilities and specialized data-reporting to be able to do real attribution modeling, and few agencies or marketers have the required capabilities.

Alternatively, you could choose one or, at the most, two remarketing partners, with only one allowed to source inventory from the ad exchanges. Have other networks if you like, but do not give them remarketing pixels, and prohibit them from using text links, chat windows or email pages for view-through attribution. Judge them by their ability to drive consumers to the client's Web site, and to drive first-visit conversions, and for how they build the remarketing pool. Include among your metrics each network's unique contribution to plan reach and percentage of converters who were touched by the network. Finally, carefully track the overall contribution to frequency at the user level so you can avoid unnecessary duplication across your plan.

This approach allows agencies to flex their strategic muscles and add real value for clients. Instead of engaging in the false Darwinism of competing networks, agencies can build a solid plan dominated by direct publisher buys and one or two strategic partners. This approach will almost always deliver superior performance, happy clients and a more rewarding assignment for the agency planner.

Recommend (8)
1 comment about "Logging In: The False Darwinism Of Competition".
  1. Eric Porres from Rocket Fuel Inc , February 8, 2010 at 11:17 a.m.

    Good piece. It's surprising that what the team from Atlas invested heavily in regarding conversion attribution hasn't caught on vis-a-vis modeling attribution that considers each measurable online consumer touchpoint in the lifecycle of a campaign (note: DoubleClick also has a conversion attribution model), for it addresses unique contribution, overlap, and enables the attribution of value per touchpoint (i.e. assigning a greater value to a 300x600 display ad vs. a textlink even if the textlink generated the 'last click'). I think what we see is a hold-out of 'network planning' from the days before exchanges, meta-exchanges, and DSPs enabled access to enormous pools of biddable cookie inventory.