In Digital, we are always on the lookout for a new ROI-producing channel.  We willed for that (unsuccessfully) with Facebook organic.  Mobile, while not a channel, continues to vex those struggling with sub-par conversion rates.  Programmatic display provided hope of a new standalone prospecting source that could yield efficient sales on a last-click basis.  Each has a solid place in the marketing mix, but none has emerged as the new high-producing direct response channel of its day.

An exception is that for marketers with a high level of resources and sophistication, programmatic display is that new channel from both a CRM and acquisition standpoint.  It is one of the key pillars to their DMPs.  For medium and smaller digital marketers, though, the amount of resources needed to build and activate a DMP may not be particularly accessible. 

That said, these marketers may still want to use programmatic display for prospecting, and there is no shortage of vendors selling it.  Even those on the supply side acknowledge that it won’t perform on a last click basis, so there are two mechanisms by which this channel gets justified for direct response offers.

The first is in measuring performance by adding in view-through conversions with click conversions and reporting on them as one aggregate sales number.  Most direct response marketers do not like this metric for two reasons.  First is that the display campaign takes credit for view-through sales that are being double counted in another channel.  This creates overlap and messy analytics.  The second is that it can’t be proven that the display banner is what actually caused the sale in the other channel. This latter point is complicated by the issues surrounding viewability and the fact that in the absence of ad verification technology, view rates hover around 45%.  Trying to pair up the display view to a sale elsewhere was the impetus for all the great attribution technologies that launched around the same time as Programmatic Display such as Adometry, VisualIQ, Convertro, etc.  The effort and expense needed to empirically match together a display view and sale in another channel is high.

The second rationale that the supply side and agencies use to justify direct response spend in programmatic display prospecting is that it can “feed the remarketing pool.”  This means that regardless of whether prospecting clicks actually yield any sales, they are still useful for building up the amount of impressions that can be remarketed to.  For direct marketers, this is more appealing than view-through tracking because remarketing is usually considered a low-hanging-fruit opportunity for driving sales.  Many marketers would love more of it, but the opportunity is limited to the amount of visitors to the site.

Fundamentally, though, there is different value in each of these cookies depending on from where they originated.  When one references feeding the cookie pool with prospecting, it is usually as a blanket effort with not much thought given to how qualified the cookies actually are.   For example, a click to a financial aid page on a .edu site would be far more valuable than a click to the .edu home page that originated from a broad display prospecting segment even if it included demos and geos, for example.  The financial aid click is further down the funnel so has more value.  This is the weak link in the effort to use prospecting display primarily as a means to feed the cookie pool.  Note that I do believe that a highly contextually targeted prospecting campaign on cpc can produce ROI by targeting very low funnel keywords.  I do not believe the same can be said of audience targeting no matter how many demos are  layered on.

We recently ran a large-scale prospecting campaign that subsequently got turned off when it didn’t produce on last click.  We then took a look at how this affected the remarketing cookie pool.  As you can see, the impact to the remarking campaign impressions was fairly dramatic.  The cookies basically fell off a cliff.

I then did a quick back-of-the-envelope analysis to see if we could have kept prospecting display live anyway primarily as a way to feed the remarketing cookie pool.  I added together the cost and performance of both campaigns to see if, combined, they produced acceptable ROI.  As you can see in the chart below, they didn’t.  Note that there are many variable to this analysis – any one of which could potentially swing success the other way such as average order value, CTR, conversion rate, and media cost.  I ran 3 scenarios using average benchmark KPIs at different AOVs.  (I chose AOV because it is one metric, in particular, that can have a big impact on the p&l.)  The cost-to-revenue calculation is higher than most advertisers would accept even in the higher AOV scenario.

While it didn’t prove out in my situation, it might for others depending on the KPIs.  I recommend this methodology as a simple way to determine whether or not prospecting display does, in fact, have value for feeding the remarketing cookie pool to ultimately produce ROI.