Playing The Attribution Game: VivaKi Analytics Exec Weighs In

Publicis Groupe’s VivaKi recently released a whitepaper on programmatic media attribution, arguing that marketers can be gamed if they aren’t careful.

Real-Time Daily spoke with Shirley Xu-Weldon, VP of analytics at VivaKi, about the whitepaper, and she said the core question marketers need to understand when it comes to programmatic attribution is: “How do you understand the difference between someone trying to steal attribution credit and making it look ‘good,’ versus a partner that’s really adding incremental and true value to your media plan?”

The whitepaper gives examples of “attribution gaming,” the first of which can also be called “impression bombing.” In this example, Partners A, B, and C all serve ads throughout the day, at an average CPM of $2.50. Partner D swings in late in the day and buys a ton of impressions at $0.01 CPM and the conversion is ultimately attributed to Partner D. (This is also an example of “last touch” attribution.)

Because Partner D paid so little to buy those impressions, VivaKi points out that it “virtually guarantees that CPA (cost-per-action) performance” from Partner D will meet campaign goals. Thus Partner D is hailed, but VivaKi points out that Partners A, B and C need to be credited, as well.

And crediting Partners A, B and C is not best done in the name of fairness. Rather, it’s about getting the most out of every dollar.

In another real example VivaKi saw, a media-buyer had three partners, all of which were given a $20,000 budget. Each partner bought ads for the client at different parts of the day, and at different CPMs. In the end, the campaign generated 1,000 conversions ($60 cost-per-acquisition).

The same advertiser then added a fourth partner and split up its budget by giving each partner $15,000 (so the total budget remained the same at $60,000). Once again Partner D bought a ton of impressions late in the day at just $0.01 CPM.

The end result was a low cost-per-acquisition of $43.48 for Partner D, which could convince marketers that Partner D was a great addition to the campaign. However, the campaign as a whole generated 910 conversions ($65.93 per), nearly 10% less than before Partner D joined the party.

“The reason for this is that Partner D wasn’t adding incremental value, but rather taking credit for the conversions that may have been driven by A, B and C,” explains Xu-Weldon. “This is a net loss for the marketer as the budget allocated to Partner D could have been spent to generate the 90 incremental conversions.”

So who is to blame for the “gaming”? It’s surely intentional, right? Not in every case, says Xu-Weldon. 

“It may not be intentional,” she said. “It may be algorithmic. We’ve seen that some algorithms are set to bid the lowest CPM with the highest response rate, regardless of frequency. In those cases, the algorithms are set up to game. It’s not malicious, but the algorithms are doing that, and we have to be very aware of it.”

Xu-Weldon suggested that “clean, head-to-head” tests are the best way to determine the true value of a DSP (demand-side platform), rather than using a handful for the same media plan.

“Just adding players isn’t going to help your end-game,” she said.

The whitepaper also spells out ways to spot gaming via analysis in the absence of a multi-touch attribution partner, including reviewing performance in context, holding partners to similar inventory standards, ensuring that CPM, frequency and time-of-day delivery is similar and comparing both the ad server’s and DSP’s cost-per-acquisition from each partner, among others. The full report can be found here.

2 comments about "Playing The Attribution Game: VivaKi Analytics Exec Weighs In".
Check to receive email when comments are posted.
  1. Seth Ulinski from Independent Analyst and Consultant, August 14, 2014 at 9:36 a.m.

    To change behavior, one needs to change the rules. Giving value to vendors/impressions that actually add value should be paramount. Additionally, I've witnessed advertisers "bid up" their cost of media in DSP bake-offs: multiple DSPs targeting/-re-targeting the same audiences drives up cost significantly.

  2. Andrew Eklund from Ciceron, August 14, 2014 at 10:55 a.m.

    Sh** in, sh** out. Let's do away with CPM altogether. It's a TV model applied to the Internet. Talk about a market in need of total overhaul.

Next story loading loading..