How To Measure The True Return On Your Ad Spend

Proprietary campaign management systems (eg AdWords, AdCenter, Facebook, etc) allow advertisers to track conversions and manage towards efficiency metrics such as Cost per Acquisition or Order (CPA / CPO).

Some, such as AdWords and AdCenter, take performance management a bit further and allow advertisers to return a revenue value per conversion to report Return On Ad Spend (ROAS). But these systems stop here; failing to allow advertisers to optimize based on true, billed, revenue rather than booked revenue.

Booked orders are the gross – total orders captured via a standard online purchase process. Billed orders are the net – orders which are actually shipped and as a result, are a more accurate reflection of the revenue recognized by an advertising campaign.

Booked revenue rarely equals billed revenue because of several potential situations including but not limited to:

1. Credit Card Rejection
2. Order Returns / Exchanges
3. Order changes via phone
4. Cancellations

There is no proprietary campaign management system that allows advertisers to adjust order values retroactively. As a result, if you want to look at billed revenue for optimization purposes, it takes crafty reporting, well-tagged URLs for clean analytics, and a very dedicated analyst capable to merge data across multiple systems.

There are third party campaign bid management and reporting systems offering solutions for advertisers to import analytics via FTP and are allow optimization and reporting using billed revenue rather than booked revenue. That said, you don’t need the extra software/expense to understand and optimize off of billed revenue metrics.

Any decent analytics system allows an order ID to be associated with a purchase event. If the analytics system is properly tagging inbound traffic, each order ID is linked with the appropriate channel, campaign, keyword or placement, etc. Through order ID level reporting and reconciliation against adjusted order values, an analyst can easily update all metrics to more accurately reflect billed revenue.

The only real downside to this process is time and reactivity. Depending on the ease of the reconciliation it can take up to a day to execute all the required reporting and put the output in an actionable format which can be imported into AdWords, AdCenter, Facebook, etc to adjust bids.

To ensure data is statistically significant and previous bid changes into account, bid adjustments in a system like AdWords or AdCenter should only be made once every few days and must be done manually. But if a completely manual reconciliation and bid management extract doesn’t get you excited and you would prefer to use automated bid strategies available in proprietary systems, read on.

More interesting than cleaning up reporting is identifying deeper opportunities within new data sets. Truing up revenue is a time intensive endeavor but the net result can be leveraged to quickly adjust more real-time metrics. The relationship between booked and billed revenue allows an analyst to adjust booked revenue goals, effectively predicting billed revenue based on historical trends.

With enough time, a normalized trend line forms; defining forecasted billed revenue as a percentage of booked revenue. While not perfect, you can use this predictive analysis to get more reactive and manage campaign/channels using adjusted revenue goals to more accurately optimize to profit margins.

The same analysis can be done for each channel to identify appropriate multipliers and improve reactivity and predictions of true return on ad spend.

calculating billed revenue as a percentage of booked revenue

Operating without knowledge of billed return on ad spend creates a risk of over-reporting on marketing channel performance.

If your only insight into channel performance is through basic analytics or information provided directly from proprietary tools, you could be assigning too much revenue to various revenue streams and mismanaging leading to non-profitable campaign as a result of prior performance assumptions.

Why operate under assumptions when you don’t have to?