Understanding the true business value of initiatives is crucial to optimizing budgets and Media Mix Modeling (MMM) is an increasingly popular approach for holistic marketing measurement. It strives to analyze historical data on marketing spend and sales to identify correlations and recommend optimal budget allocation across different marketing channels. However, MMM has limitations, particularly with affiliate marketing.  

CJ recognizes the specific challenge of Media Mix Models measuring affiliate marketing alongside other marketing channels. To address this need, we've developed this comprehensive guide and FAQs. We believe this information will empower your company to make data-driven decisions and optimize your marketing investments. 

To address this challenge, CJ suggests marketers: 

• Understand the limitations of MMM and advocate for a nuanced interpretation of its results. 

• Employ supplemental measurement techniques, like multi-channel customer journey analysis, to explore affiliate’s contributions alongside MMM data. 

• Emphasize the unique strengths of affiliate marketing, such as its performance-based model, media diversification, and strong ROI. 

Same Data, Different Outcomes 

MMM model designers must make assumptions about the data and the model, which can lead to vastly different outcomes even when using the same data and methodology - see figure 1. 

Fig. 1. The same overall model with the same data but different analyst assumptions can lead to vastly different outcomes, and different spend recommendations, despite the data the model remaining constant. Even with variations, models can exhibit a high level of precision and statistical significance. This doesn't mean the results are incorrect, but rather that they're based on different sets of assumptions.  Source: Google Research  

Strong vs. Weak Correlation  

Imagine that whenever you increase your media activity in a channel (impressions, clicks, etc.), you also see a similar increase or decrease in sales. This indicates a strong correlation. Conversely, if sales fluctuate significantly regardless of your channel spending, there's a weak correlation. 

The charts below illustrate these concepts: 

• Left Chart: Shows changes in monthly media spending accompanied by similar changes in sales (upward or downward trends). This suggests a strong correlation. 

• Right Chart: Shows sales fluctuating independently of a relatively steady spending rate. This indicates a weak correlation. 

MMM provides a powerful tool for understanding the impact of traditionally hard-to-measure channels, like broadcast and cable TV, on overall sales. But MMM are not always an objective, infallible measure of channel value. 

However, it's important to understand the limitations of MMM: 

• Accuracy for Specific Channels:  While MMM companies often claim high overall accuracy (e.g., 95%), this might not hold true for every channel represented in the model. Relatively smaller budget channels with steadier spending patterns may not have enough variation in data points for a statistically significant correlation. 

• Wide Variation in Results: In working with clients on attribution, we have seen dramatic changes in outcomes when companies move from one measurement vendor to another due to different assumptions in the models. Models with similar data can produce different outcomes (as shown in fig. 1). This further underscores the need to leverage MMM analyses with a degree of caution—MMM are not always objective, infallible measures of channel value.  

Lower Data Volume

Affiliate programs with smaller budgets compared to other channels, such as display or paid search, could result in comparatively lower data volume for analysis, making it harder to isolate affiliate's contribution to sales within the overall "noise" of the data.

In marketing programs where affiliate captures a relatively smaller share of total marketing spend, assessing spend correlation with sales is more challenging.

Always-On Activity

Unlike channels with scheduled campaigns (e.g., TV ads), affiliate programs are generally "always on” due to the evergreen nature of the content created by affiliate partners. Content longevity and consistency can make it challenging to pinpoint correlations with sales fluctuations.

While there may be fluctuations in affiliate activity for promotions, the changes tend to be small relative to total brand marketing investment, making it more challenging to isolate their impact on sales.

Limited Segmentation

Many affiliate programs operate on a national level, due to the openly accessible nature of publisher sites. Segmentation, when it does occur, is often behind a user login or achieved via targeted email campaigns.

The commercial structure of affiliate partners’ segmentation makes it harder for MMM to associate affiliate with changes in sales performance in one segment or region.

The Probability Challenge 

Correlation analysis in MMM is like trying to hear a single instrument in an orchestra. Many factors influence sales, such as other marketing efforts, seasonality, and economic trends. This "noise" in the data can make it difficult to isolate the true impact of each marketing channel. 

In MMM, correlation analysis is usually easiest for channels with the largest investment, channels that vary spending significantly throughout the year, and channels that can segment marketing efforts to measure specific outcomes in a region. 

The more data points and spending variations a mixed media model has to work with, the clearer the picture becomes. This is reflected in the P-value, the probability of a statistical measure of significance. A lower P-value (typically below 0.05) indicates more reliable findings. 

Here's where things get tricky for MMM: 

• Smaller budgets, steadier spending.  Relatively smaller budget channels with steadier spending patterns may not have enough variation in data points for a statistically significant correlation. This makes it harder to detect a correlation with sales. 

• Limited segmentation. If a channel lacks clear segmentation (e.g., by region or audience), it's difficult to isolate its impact within the overall data because the marketing effort is compared to total outcomes vs segmented outcomes. (For example, it is more difficult to measure a response on a national campaign vs. a regional campaign.)   

Educate Marketing Leaders

Advocate for a nuanced understanding of MMM's strengths and weaknesses. MMM's apples to apples view is the ideal for CMOs, but there are shortcomings to having one tool that measures all channels. Relatively smaller, always-on channels like affiliate deserve additional examination beyond MMM. 

Highlight Affiliate's Advantages

Emphasize the benefits of affiliate marketing, many of which are highly valued by organizations, yet not viewed via these measurement models: 

• Quality Traffic:  Affiliate partners deliver transparent performance with verified clicks, visits, and purchases and attract audiences with desirable demographics. 
• Performance-Based Model:  Affiliate marketing aligns perfectly with the growing focus on performance marketing, as it directly links costs to business value. 
• Media Diversification: Investing in affiliate marketing helps reduce dependence on large media platforms, mitigating risks associated with concentrated spending. 
• Scalability: Affiliate programs can reach vast audiences comparable to mainstream publishers and television programs.  
• Strong ROI/ROAS: Numerous studies show affiliate marketing drives transactions efficiently, often at a lower cost than top digital channels like search and social, where media costs are rising.  
• Amplification Effect: Data analysis shows that affiliate touchpoints within a customer journey can significantly enhance the effectiveness of other marketing channels, even if those channels ultimately receive credit for the sale.