A true measure of retargeting effect

With the recent explosion of new vendors, targeting options, and prediction models, advertisers are testing and evaluating more ways of buying online inventory. However, as more advertisers today are eager to test various tools like demand-side platform (DSPs), data management platforms (DMPs), or dynamic creative technology, they are rightfully nervous about disrupting current successful efforts. Unfortunately, these desires to preserve performance anchors often lead to sub-optimal testing efforts that can waste budget and return biased results.

Before the rise of ad exchanges, advertisers would often allow almost all their inventory sourcing vendors to tap into retargeting. Now, there's a legitimate concern about how cross-bidding (i.e., bidding against yourself for a single ad impression) on the exchanges may impact effective cost-per-thousand (CPM) impressions. This leads many advertisers to wall off their retargeting campaigns, leaving one vendor responsible for retargeting and testing and other vendors for additional scale. The idea behind this is that the non-retargeting campaign will provide incremental conversions, while the retargeting campaign sits happily with the preferred vendor providing the performance anchor. This thinking is siloed and doesn't function well with today's holistic marketing strategies.

On the surface, segmenting your display efforts into lower-funnel retargeting and upper-funnel methods makes sense. Retargeting is easy -- right? But simply throwing a wall between upper- (retargeting) and lower-funnel campaigns and evaluating incremental conversions based on a single metric doesn't work. Here's why:

1. With an attribution model that values adjacency to the conversion but with a hampered ability to target to that adjacency, the upper-funnel campaigns will have a very difficult time achieving the required performance metric. This eliminates the potential to scale the upper-funnel campaign, and the test will fail. With multiple failures to expand beyond retargeting, efforts will return to focus only on the lower funnel.
   
   
2. As interest driving campaign efforts ebb, retargeting will slowly starve out unless new interest is driven from other sources, such as branding or offline campaigns.

When we look back at our ill-fated upper-funnel campaigns, it's evident that some of these efforts did an excellent job of generating interest and luring new consumers into the funnel, but since they were being judged on a lower-funnel metric, they appear to have provided little value. Marketers should explore solutions that leverage non-retargeting campaigns to grow audience for their lower-funnel efforts.

For advertisers that do not want to touch their current retargeting efforts at all, tests can be constructed to focus on upper-funnel interaction with corresponding metrics. For example, if a major automotive brand is looking to increase the number of site visitors who configure a vehicle using a "build a car" tool, you know that users who are more engaged with your brand are more likely to participate. Thus, remarketing performs very well. In this case, the automotive advertiser wants additional scale, and needs a metric to evaluate how a non-remarketing campaign contributed to the overall volume of users who configured cars. This funnel-stage bridging metric tracks new users who later convert, and allows for a fairer comparison of upper-funnel campaigns while ensuring lower-funnel value still is a primary driver of success. This metric is calculated by the probability of a new user converting (anywhere) after exposure to the upper-funnel campaign divided by the probability of a new user converting (anywhere) without being exposed to the upper-funnel campaign.

With today's eco-system relying more heavily on targeting, there are many advertisers that want to test entire targeting solutions. In this case, retargeting data can be split into discrete groups by vendor, along with a hold-out control group. Each vendor's efforts can be compared to the performance of the control group so that retargeting effectiveness can be measured as true incremental lift. This prevents cross-bidding, and is my preferred method of testing. Additionally, segmented full-funnel tests are a great way to evaluate the growing number of media and technology vendors.

Building test plans with goals that more accurately reflect the role of the media buy will result in greater campaign insight, and more efficient media spend. Hopefully, with a little extra effort advertisers can break the trend cycle of test-and-crash that seems to affect so many non-retargeting campaigns.

Maxwell Knight is director of optimization and professional services at Turn.

On Twitter? Follow iMedia Connection at @iMediaTweet.