Measuring Return on Investment: ROAS & iROAS

Great marketers need to know where their ad spend is most effective. But in a cookieless world with murky attribution, it’s hard to be 100% confident which ads work, and how well. For years, marketers relied on ROAS (return on ad spend) to inform future investment decisions, but this method presents an unreliable, incomplete picture. 

To solve this problem, Fetch, a leading platform in the zero-party data space with unparalleled access and influence on omnichannel shopping behavior, is setting a new standard of marketing measurement to eliminate the uncertainty of previous methods. That new, more sophisticated methodology is a KPMG-verified, test-and-control take on iROAS. 

Our methodology uses randomized holdouts—the most rigorous standard in the control vs. exposed methodology—and SKU-level data from billions of physical and digital receipts to track ad performance. Marketers can use Fetch’s methodology data to make smarter and more informed business decisions—such as whether new or lapsed users are driving a higher incremental return, or how to encourage more repeat purchases. 

To illustrate how Fetch’s test-and-control methodology improves upon these marketing measurements, let’s examine the challenges presented by ROAS, iROAS and other traditional methodologies. 

 

ROAS: Incomplete and misleading

ROAS is used to calculate the performance of a campaign over a specific period of time. It’s calculated by dividing the observed revenue by the cost of the promotion.

It’s no secret that ROAS is a less-than-desirable measurement because it doesn’t give a true picture of an ad’s effect on shopper behavior. ROAS assumes that a single factor (the ad) is responsible for all purchases, which overstates sales attributed to that specific spend. It ignores the impact of other factors, such as macroeconomic events, seasonality and other media advertising. 

As an example, let’s examine pumpkin spice lattes. Sales for the iconic fall beverage naturally heat up from September to December. Who’s to say that any ad campaign was more responsible for the boost in sales than the changing of the leaves? ROAS doesn’t account for sales that would have happened due to other factors, like seasonal change, macroeconomic trends and external shifts in product popularity.

The resulting data can inflate perceived advertising spend efficacy. In reality, that spend may actually be subsidizing existing and consistent behavior.

 

iROAS: Better, but easily biased and hard to prove

To alleviate inaccuracies of ROAS, marketers began using iROAS (Incremental ROAS) to measure the incremental lift of an ad campaign. It compares revenue between two groups: a treatment group that is targeted with ads and a control group that is not. The difference in revenue between the two groups is incremental revenue.

iROAS is an improvement over ROAS because it attempts to isolate purchases influenced by the campaign. However, since there are multiple methods (examples outlined below) to measure incremental revenue, it is impossible to compare results. Overall, these methods produce inaccurate data, which impedes marketers’ ability to adjust their strategies in real-time.

 

Three common ways to measure incremental revenue: 

• MMM (Media Mix Modeling, or Marketing Mix Modeling). MMM uses an econometric modeling approach to forecast trends.
  • Weaknesses: Lack of transparency in modeling and inconsistent measurement standards obscure data. 
• PSA Testing (Public Service Announcement Testing). PSA testing delivers a simple message to some percent of the media, brand messaging to another, and then measures the difference in desired behavior.
  • Weaknesses:
    PSAs can be prohibitively expensive, with no guarantees about the similarities between control and treatment groups.
• Causal Attribution. Also called the Control and Exposed method, causal attribution divides consumers into treatment and control groups based on data collected from marketing campaigns. 
  • Weaknesses:
    Since this is typically based on self-reported demographics, there is no way to ensure the treatment and control groups were created without bias. 

Marketers need to know they aren’t receiving artificially inflated results from messages sent to groups who may be prone to respond to specific promotions or brands. It’s clear that iROAS is still missing something: actual, verified proof at scale. 

That’s why Fetch’s test-and-control method is the most powerful—and accurate— measurement methodology available. 

 

Fetch’s Methodology: We have the receipts

Fetch’s methodology is based on the control and exposed method of iROAS but with significant improvements:

1. 1. Bias is minimized through random assignment. Minimizing bias in treatment and control groups, achieved by creating statistically significant group sizes, ensures the only difference between the two groups is exposure to the offer.  
   2. Treatment and control groups are based on observed behavior instead of self-reported demographics. Fetch’s data gets to the SKU level, providing much deeper insights into consumers’ buying across all channels. Additionally, holdouts remove the uncertainty surrounding seasonality and external market factors because Fetch can measure the purchasing behavior of both groups during the promotional period in a closed-loop environment.
   3. Results are verified through receipts. Fetch obtains physical and digital receipt data from all retailers, providing an indisputable 360°view of consumers’ shopping habits.

With Fetch, a campaign’s impact is crystal clear. Measuring the outcomes of both the treatment and control groups at the same time provides a cleaner read on consumer behavior and results. Fetch does not model results—rather, we use receipts submitted by our users to measure real impact. 

Be confident in the results of your next behavior-based campaign on Fetch when every purchase is receipt-verified. Reach out to the Fetch for Business team to execute your next campaign with unrivaled precision, crystal-clear attribution, and verified incremental results.