How to Correct Inconsistent Results from Your Programmatic Media Buying
ADOTAS (2nd of 5 parts) – No one is perfect when they first set out to master a new skill. Programmatic advertising is no different. This is why I have set out to diagnose five common pain points associated with mastering programmatic real-time media buying. In this post, the second of the series, I will explore the causes of erratic campaign performance and prescribe some solutions.
Erratic performance is when audiences that perform well in tests deliver unexpectedly low performance at scale, or when initially successful campaigns do not sustain performance over time.
Segments that are similar on the surface, yet fundamentally different, might be one cause of erratic performance. These segments appear identical, and might even have the same name, but could ultimately have very different audience composition and performance with respect to a product or brand. Thus, targeting seemingly similar segments to increase scale might actually result in targeting the wrong audience altogether. As an example, our team found that the Consumer Electronics segment constructed using search activity worked well on a mass-market consumer electronics product, but performed poorly for a high-end audio product. On the other hand, the Consumer Electronics segment that was built from website browsing activity worked well for the high-end audio product and fell flat for the mass-market product.
To avoid challenges like this one, advertisers should run initial tests that collect data from a broad set of sites and audiences. If you are working with a full-service DSP, its learning algorithm can evaluate all data in combination to achieve two objectives. First, the algorithm should be able to discern the difference between seemingly similar segments based on the actual performance of each. Second, the algorithm should discover the sub-audiences and independent niche audiences that perform well. Only through the assembly of many highly targeted audiences can a campaign reach scale without compromising performance.
Another cause of erratic performance can be a dynamic audience. Not every audience segment responds the same way all the time; some shift in many ways, even in very short periods of time. The Ferrari segment, for instance, typically contains consumers with high net worth and a luxury life style. After a massive Ferrari crash in Japan, however, the segment shifted within a matter of days to include a wide array of newsreaders. Even when the composition of a segment remains constant, the segment’s relation to a brand can shift. Take, for example, a Spanish-speaking segment that performed particularly well at the outset of a TV provider’s campaign. The high performance of that segment ceased abruptly at the conclusion of the World Cup games. It turned out that many Spanish speakers had enrolled at the start of the campaign in order to tune into the World Cup, since coverage was available through that specific TV service provider.
To prevent this issue, advertisers should constantly reevaluate their targeting by using an automatic optimization system that keeps up with dynamic audiences in real time. Systems that cycle as quickly as every 15 minutes can detect shifts in performance to drop segments that have stopped performing as they should. More importantly, these systems can also discover replacement audiences to maintain scale and pacing as well as performance.
Automated optimization allows brands to find the sub-segments within an audience that perform well. Sometimes the defining attributes of the sub-segment identify incremental, independent audiences that perform well. At other times, only the sub-segment responds well. For example: while targeting mini-van drivers, it may appear that small business owners within that segment respond exceptionally well. Through testing, an automated optimization system can determine if all small business owners (independent audience) perform well, or if it is only the small business owning mini-van drivers (sub-segment) that perform well.
In sum, segments aren’t always what they seem, and even when they are, they can change rapidly. Detecting differences and keeping up with changes will allow you to avoid erratic campaign performance through deployment of an automated optimization system built on your initial guidance.
No comments yet
Leave a Comment
- Game, [Data] Set, Match: How Deep Learning Turned Sci-Fi-esque Predictive Intelligence Into Reality
- Growth of Location-Based Mobile Video Ads Is 2016′s Top Mobile Ad Trend
- Social Media Ads: What’s New, What’s Next
- Idomoo: Empowering Dynamic Video Ads on Facebook
- Want Programmatic to Work for Tech Marketing? Dismiss everything you’ve heard!