As an industry, we look to attribution as a way of driving even more accountability out of digital. But what if the data we use to attribute credit is itself being gamed?
According to Quantcast representatives during a Workshop presentation at the iMedia Agency Summit in Scottsdale, Arizona, attribution "gaming" or fraud is very common, and is sapping brands of a big share of their marketing investments. Quantcast defines attribution fraud as a deliberate effort to manipulate performance data in ways that are outside the measurement capabilities of attribution model's measurable capabilities.
One of the most common ways this occurs is when media companies who are contracted to deliver prospecting media incorporate large amounts of retargeting into the mix to improve CPA metrics. Since someone who has already visited your site is far more likely to transact, gaming through retargeting can bring a campaign CPA down -- and help that vendor win a bigger share of spend. That would be bad enough, but when multiple media companies are competing for the last touch credit, brands end up spending huge amounts of money delivering hundreds of impressions against sales that they would have earned anyway.
Says Quantcast Head of Product Marketing Valter Sciarrillo, brands often spend more money than is prudent at the bottom of the conversion funnel as their media vendors compete to deliver the last click that will attribute a sale to them.
The company's solution: split funnel attribution. This innovation in marketing measurement separates prospecting measurement from remarketing measurement, to help agencies and clients identify how vendors actually spend their budgets. The prospecting portion of measurement relates to all activity that occurs before a user visits a brand site for the first time. The retargeting component tracks what happens between that first visit and a conversion event. The "split" is at the first site visit.
Quantcast likens the split funnel to a two-runner relay race. Most brands are spending all their time analyzing the second runner, and ignoring the first. That imbalance of attention creates an opportunity and incentive for attribution fraud.
Sciarrillo reported that by splitting marketing activity into these two discrete groups, brands and agencies can get unique insight into how their dollars are being allocated -- insight that they cannot get from an ad server alone. In addition -- implementation is relatively simple, because all the tracking can be managed with a single tag.
Using this new set of rich insights, brands and agencies can answer such important questions as:
- How are vendors really allocating dollars between prospecting and remarketing? Are they gaming the attribution model by spending more than expected on remarketing?
- What are the right levels of media activity for the prospecting and remarketing phases? By finding the inflection point where levels of activity are most efficient, brands can free up more dollars for additional prospecting that can drive real incremental growth.
- How can brands improve their assumptions and benchmarks about effective frequency and other basic media metrics?
"Think of the power there," said Sciarrillo. "If you see a high conversion rate from a first touch, then you know there is less need for retargeting. If the conversion rate from the first touch forward is low, you know that higher levels of retargeting are needed."
Perhaps the most compelling statistics from the presentation related to the observed ranges of prospecting versus retargeting effectiveness between vendors, and what that should mean for how brands allocate their resources and attention. According to Quantcast, the range of retargeter effectiveness varies by about 3X. By contrast, the range of prospecting effectiveness is more than 10X. In other words, there's a lot more variation in the speed of your first runner in that relay, than in your second. Most brands would do well to spend more of their time and resources optimizing the upper funnel than focusing so much on the last moments before a sale.