This week’s blog was contributed by Andrew Van Benschoten, Senior Data Scientist, Oracle Data Cloud.
When marketers attempt to improve their Return on Ad Spend (ROAS), they often approach Oracle Data Cloud for advice on how to drive greater incremental sales. While that’s a natural response to lower-than-expected returns, it fails to consider the other side of the campaign equation: cost.
It’s particularly critical to examine media expenses, as they represent a substantial chunk of any marketing budget. We recently discussed how misallocating impressions (where the majority of impressions are served to a small fraction of the exposed audience) is a common source of waste in programmatic campaigns. For example, we’ve even seen individual households that received over 100,000 impressions during a single advertising campaign!
To better understand how over-serving impacts media costs, we expanded on this previous analysis to look at 36 recent programmatic campaigns measured by Oracle Data Cloud. Campaigns in this dataset ran on 35 different platforms and had a median campaign spend of $500,000. What we found was extremely surprising: it turns out that the typical campaign spent $255,000 of that budget on only 5% of the exposed audience! How much could be saved if these “high-exposure” households received a more reasonable number of impressions?
After adjusting the number of impressions served to the “high-exposure” households to the typical number observed by the lower 95%, we found that these 36 studies could have saved a combined $10.8 million out of the total $22 million spent:
Our end result? Even though the optimal number of impressions will vary between households, eliminating even a fraction of clearly excessive ad serving can lead to big gains in your campaign’s ROAS.
As will be discussed in an upcoming blog post, re-allocating some of these impressions to underserved households can boost incremental revenue, giving your campaign a solid path towards boosting the ROAS.
Photo: Monkey Business Images/Shutterstock