Many marketers avoid testing the validity of their return on ad spend because they worry that they might not acquire as many new customers during the process. Reality is, testing is the only way to truly understand the incremental value of your ad spend, whether it’s new, retained or reactivated customers.
Incrementality testing is paramount to orchestrating a best-in-class digital media strategy that drives enterprise results. However, when doing incrementality tests, it’s important to distinguish between “predictive” and “causal” analytics.
Predictive analytics creates models from historical data, and these models can be used to inform marketing decisions. For example, a brand may uncover a relationship between historical paid search spend and total revenue that indicates every $1 spent in paid search correlates with an additional $2.50 in revenue. A return-on-ad-spend (ROAS) of $2.50 implies that the brand should increase their paid search budget. However, in this scenario, predictive analytics has two key flaws:
- It cannot distinguish between correlation and causation. How can I be sure that an unknown factor wasn’t the true cause of the increase in sales? Many factors are at play here. For example, the target audience may have been influenced by the halo effect from a category competitor’s brand push happening concurrent to the paid search efforts.
- The models produced by predictive analytics work best when spending decisions are close to historical levels. In the paid search example, let’s assume the brand has historically spent between $10,000 to $20,000 per month in that channel. If they spend $30,000, we might safely assume that the $2.50 ROAS trend will continue. But should they be comfortable continuing this assumption as spending levels increase to $50K, $500K or $5 million?
Predictive analytics works well when you want to keep spending at or near your current level. As your business grows, however, you’ll inevitably need to increase your ad budget. If you were to increase your monthly paid search budget from $20,000 to $50,000, you would need to use causal analytics to determine how much additional revenue that shift would generate. Causal analytics not only answers whether your increased paid search spend led to increased revenue, but also to what extent.
Most digital media agencies confuse predictive analytics with causal: they build machine learning algorithms and analyze Media Mix Models to give a sense of causality that simply isn’t present. At Ovative, we capture causality through the gold standard of randomized controlled tests. By testing your media strategy across geography, customer base and other meaningful factors in our Marketing Analytics Platform, we identify the true incremental impact of your media.