If you're working in marketing and advertisement, you already heard about ROAS. But, like many other topics in marketing, there are important nuances to ROAS. Understanding the intricacies of ROAS is crucial for optimizing marketing strategies and maximizing returns on ad investments.
This article aims to highlight some of the caveats around ROAS and equip you with the necessary knowledge in interpreting ROAS and similar metrics.
What is ROAS?
ROAS (read row-aas) is one of the key metrics in any marketer's toolbox. It stands for Return On Ad Spend, and as the name implies, it's an ROI-like metric.
ROAS is the change in revenue per dollar spent on ads. It can be reported as a percentage or in absolute dollar terms. For instance, 200% ROAS means you earn $2 for every $1 spent on advertising.
Instead of defining ROAS for the entire advertising, it's more beneficial to report ROAS at the channel level. Reporting ROAS at the channel level is crucial for advertisers to optimize their channel mix strategy.
ROAS is an attempt to isolate the impact of a channel on the revenue outcome, regardless of other channels within the mix.
For example, if ROAS values for Facebook Ads and Google Search are 300% and 400%, respectively, Google Ads generates $4 per dollar spent. In this campaign, Google Search outperforms Facebook Ads, making it logical to allocate more budget to Google Ads.
Example of Channel-Level ROAS
Consider a marketing campaign that includes three digital media channels: Google Ads, Facebook Ads, and YouTube Ads. The table below shows the spend in each of the three channels.
Moreover, we know that we generated $1.5M in revenue.
What is the ROAS for each of the channels? To do so, we must generate three counterfactual scenarios where we switch off a single channel and predict the revenue outcome without the channel.
The following figure shows the three counterfactual scenarios for our channels, along with the predicted revenues under each scenario.
| Channel Name | Symbol | Spend |
|---|---|---|
| Google Ads | A | $250K |
| Facebook Ads | B | $100K |
| YouTube Ads | C | $200K |
Moreover, we know that we generated $1.5M in revenue.
What is the ROAS for each of the channels? To do so, we must generate three counterfactual scenarios where we switch off a single channel and predict the revenue outcome without the channel.
The following figure shows the three counterfactual scenarios for our channels, along with the predicted revenues under each scenario.
Based on these predictions, the revenue drops by $1M when Google Ads is switched off. Given that our original spend on Google Ads was $250K, the ROAS for this channel is $4 (or 400%), as shown by the following formula:
Following table summarizes the ROAS results for every channel:
| Channel Name | ROAS |
|---|---|
| Google Ads | 400% |
| Facebook Ads | 600% |
| YouTube Ads | 400% |
Key Considerations When Interpreting ROAS
When reviewing ROAS results, consider these three important propositions:
- A channel's ROAS decreases when spending more on the channel.
- A channel's ROAS depends on other channels within the mix.
- ROAS changes over time.
In this section, we discuss each of these three propositions.
Channel's ROAS decreases when spending more on the channel
When faced with a relatively high ROAS for a particular channel, it can be tempting to over-invest in the channel. As a result, the channel becomes less effective due to the diminishing return effect on marketing spend.
See - Explaining Diminishing Returns in Marketing Spend
It's important to be mindful of the law of diminishing returns when it comes to spend on a particular channel. Overusing a channel leads to decreased effectiveness, resulting in a lower ROAS as we continue to allocate more resources to that channel.
Therefore, it's essential to find the optimal balance to maintain the channel's effectiveness and ROAS.
In practice, marketers often over-invest in specific channels due to incomplete analytics or anecdotal wisdom. We recommend measuring ROAS regularly to avoid channel saturation.
Channel's ROAS depends on other channels within the mix
In the complex landscape of advertising, the ROAS for a particular channel is not solely determined by the performance of that channel alone. Instead, it is influenced indirectly by the interactions of other channels within the marketing mix.
Understanding that different channels are intricately interwoven and rarely operate in isolation is crucial for developing a comprehensive and accurate assessment of advertising effectiveness.
In the following example, ROAS for channel X varies under two different mixes. Note that the spend on channel X is identical under both mixes.
ROAS changes over time
ROAS is a temporal metric that reflects the efficiency of a channel within a specific timeframe, such as a month, quarter, or even year. The ROAS that you measured last quarter may not be the same this quarter.
Therefore, it's essential to continuously monitor the marketing mix, by regularly measuring channels' ROAS. By regularly assessing ROAS and channel mix, marketers can ensure a sustainable and impactful marketing approach.
When set to a very short period, such as weekly, background noise overwhelms the actual signal, leading to uncertain and noisy ROAS measurements.
On the contrary, long time intervals may overlook important seasonal shifts in market conditions, ultimately diluting valuable signals within the broader scope of long-term trends.
Different types of ROAS
Having multiple ROAS measurements for the same channel is often a source of confusion and continues debates. Several sources for ROAS usually exist, which are only sometimes compatible.
In this section, we review the three primary sources for ROAS measurements. Going from the most credible to the least credible, here are the three ways to get ROAS measurements:
- Incrementality testing (Experimentally verified)
- MMM Measurements (Holistic and top-level)
- In-Platform Measurements (Vendor-specific and inflated)
Incrementality testing
This is the highest quality measurement, as it relies on experimentally verified measurements. By increasing or decreasing the spend in one channel on a single sub-region, marketers can measure the incremental return of a particular channel.
By experimentally measuring the impact of a channel spend on the underlying sales, this approach offers a clear picture of the true causal impact of a specific marketing channel.
While the concept of this measurement is interesting, there are specific situations where it may not be feasible to implement.
This type of measurement also demands careful planning and precise execution, which can make it require a significant amount of resources to implement effectively.
Therefore, it's recommended to perform such experimental measurements occasionally. The outcome of these experiments can improve the internal MMM system. This is conceptually similar to the idea of AI Grounding, which is very popular in the GenAI community.
In practice, we use the results of incrementality testing to inject real-world knowledge into the MMM system.
Benefits:
- Most accurate: Offers the most precise way of measuring ROAS.
Challenges:
- Resource-heavy: Setting up experiments and analyzing the data requires extensive hands-on work and sometimes a dedicated team.
- Can be implausible and risky: It can be challenging to test every channel, especially global channels. These tests can put sales in critical markets at risk, making team buy-in difficult.
Marketing Mix Modeling (MMM)
In this approach, a predictive statistical model is fitted to the historical spend and sales data. Once the MMM system is established, it can holistically measure the impact of each channel's spend on the sales.
The MMM model can be used to estimate the counterfactual scenarios, where a particular channel is completely switched off. As a result, we can estimate the impact of the channel on revenue or sales.
MMM measurement is top-level and holistic. It uses predictive modeling to estimate channels' ROAS.
It's important to note that training MMM models can be difficult due to the shortage of training data. This "small data" problem makes it especially challenging to train a predictive model because there can be several fits that offer the same level of goodness of fit.
Moreover, the estimated ROAS values from an MMM may have large error bars (confidence intervals). As a result, these ROAS can be challenging to work with and less useful for strategic decision-making.
Benefits:
- Holistic and Top-Down: A comprehensive approach to measure all the channels within a unified framework.
- Fast and Easy: When an internal MMM is established, measuring ROAS at the channel level is very quick.
Challenges:
- Error-prone: Lack of sufficient data for training a predictive MMM model makes the models less reliable.
- Uncertainty: The uncertainty in ROAS measurements from an MMM can be large, making them less ideal for channel mix optimization and strategic decision-making.
In-platform ROAS
The last source for ROAS is from platforms such as Google Analytics. Users can simply view the ROAS reports on these platforms.
However, platform owners have an incentive to over-report the ROAS for their channel. This is how they sell their service, and an inflated ROAS makes a good case for spending more on the platform.
Benefits:
- Readily available: It's easy to obtain in-platform ROAS as major platforms report their ROAS in their dashboard.
Challenges:
- Unreliable: The in-platform ROAS is usually exaggerated, as the platform owners have an incentive to inflate their numbers.
Marginal ROAS (mROAS)
mROAS of a particular channel quantifies the change in revenue due to a minor perturbation to the spend on that channel.
Instead of comparing revenue when a channel is entirely off (ROAS), mROAS compares the revenue when we modify the spend by a small fraction. For instance, when we increase/decrease a channel spend by 1%, how much does the revenue change?
mROAS offers a more practical approach for experimentally measuring return on spend. Moreover, the idea of mROAS makes it plausible to perform several tests simultaneously.
We at ELIYA utilize a combination of incrementality testing to measure mROAS across multiple channels simultaneously. We then leverage the results of these experiments to incorporate real-world insights into the MMM system.
Key Caveats of ROAS
In this section, we highlight a few crucial caveats for ROAS:
- ROAS measurements are subject to change over time.
- A channel's ROAS decreases when spending more on the channel.
- A channel's ROAS depends on the other channels within the mix.
- In practice, there is a high amount of uncertainty for ROAS measurements.
- The standard definition of ROAS estimates a channel's impact on momentary revenue, which ignores the customer lifetime value (LTV) and future revenues.
- Small campaigns can generate very high ROAS, albeit with high uncertainty.
Conclusion
In conclusion, understanding and effectively utilizing ROAS is vital for optimizing marketing strategies and maximizing the impact of advertising investments. While ROAS provides valuable insights into channel performance, it is essential to consider its limitations and the broader context in which it operates. Factors such as diminishing returns, channel interdependencies, temporal variations, and measurement methodologies all play critical roles in interpreting ROAS accurately.
By acknowledging these nuances, marketers can make more informed decisions, balancing their investments across channels and adapting strategies to changing market conditions. Ultimately, a comprehensive approach that combines different ROAS measurement techniques, such as incrementality testing and Marketing Mix Modeling, with ongoing analysis and adaptation will lead to more sustainable and effective marketing outcomes.
