Business Intelligence

Where data analysis is more about what happened, business intelligence concerns decision-making and future predictions. A powerful method is developing an ad channel regression model to determine which ad channel has the greatest statistical impact on revenue when direct attribution is impossible. Further, this sort of modeling allows for the determination of synergistic mechanisms, e.g., Reddit ads may perform better than Google ads individually, but return on ad spend (ROAS) is best when Reddit works together with Google.

Below is a snippet of part of this modeling process. For this business, direct revenue-to-channel attribution is impossible due to a long lead-to-conversion time (sometimes years); thus, evaluating ad channel efficacy requires statistical modeling to find the relationship between relative ad spend and ROAS. For example, the negative slope on Google doesn’t mean it performed poorly, just that Google ads perform best when combined with other channels. Results like this (combined with ad type results not shown here) suggest that the majority of ad spend should be on top-funnel ads on social channels, with less ad spend on mid-funnel and bottom-funnel ads on Google and Amazon, to maximize ROAS.

Click icon for the Git