Over the last few years, Programmatic Buying has fast emerged as the mainstream and irreversible model of media buying on display ads globally (across desktop, video and mobile). As for India, almost every digital marketer we have spoken to in 2016 is already using programmatic at scale (usually starting with remarketing or branding before moving to prospecting), or is keenly evaluating a pilot, or has heard about programmatic and can’t wait to try it out.
It’s well established in the minds of marketers that programmatic buying is the way to go if you are looking for a scalable, efficient, and transparent way of media buying – to extend your reach and scale economically beyond Google Display Network (GDN) and Facebook, while the “wild west” of ad network and affiliate marketing is scripting its own obsolescence.
That being said, programmatic buying has brought its set of complexities. Massive scale of programmatic inventory results into longer optimization cycles; CPM buying model at per impression level creates some barrier or friction for testing; in addition to the concerns around inventory quality – viewability and fraud concerns (although, top 20-30 programmatic exchanges have pretty effective fraud control).
This inherent complexity in programmatic leads to multiple confounding factors when running pilots or testing new strategies for audience or contextual targeting. For instance, how do you attribute downstream performance if the ad wasn’t even viewable in the first place?
Here are some interesting articles around the nuances of viewability – http://marketingland.com/affects-ad-viewability-5-factors-google-study-109876 and http://www.adnews.com.au/opinion/what-you-didn-t-know-about-optimising-viewability-in-rtb and http://marketingmagazine.com.my/breaking-news/viewability-viewpoint-what-does-bala-have-to-say
The Peer39 study shows that there are publishers who achieve consistently high viewability, and there are others who always fall below benchmarks. In fact, 65% of domains have average viewable rates that hardly vary (delivering a standard deviation of less than 0.1), from one day to the next. And the majority of those domains actually have a standard deviation of less than 0.01 in their average daily viewable rates.
In our experience, one solution that works fairly well is to test new programmatic strategies at 80-100% viewability employing real time viewability prediction tools. Targeting high viewability does reduce the scale of inventory available. However, it helps evaluate whether the targeted audience or the context performs as expected, taking viewability out of the equation. It also accelerates the test as CTR improves 2-3x at 80-100% viewability (discounting for poor quality ad formats with 100% viewability that are prone to accidental clicks). Once the outcome is measured, you may relax viewability thresholds, if required, to open up scale, and bid at viewability adjusted CPM.
What’s been your experience with testing new programmatic strategies? What are some of the best practices you’d recommend? Would love to hear your insights.