The Impact of Systemically Removing MFA Inventory from Programmatic Media Campaigns

What is a “Made for Advertising” website? 

A “Made for Advertising” (MFA) website is a low-quality website primarily created to generate revenue from programmatic advertising rather than to provide meaningful or original content. These sites often repurpose existing content or use clickbait tactics to attract traffic. They are typically overloaded with intrusive ad placements, resulting in a poor user experience, such as slow loading times, cluttered layouts, and difficulty accessing the actual content. 

MFA websites exploit ad systems to maximize profit at the expense of user satisfaction and can harm brand reputation when ads are displayed on such sites. 

The Background on Butler/Till’s Programmatic Capabilities 

For over 15 years, Butler/Till has been executing programmatic campaigns through platforms like MediaMath DSP, Google, and Meta (Facebook). Our client portfolio predominantly consists of highly regulated industries such as banking, insurance, healthcare, and pharmaceuticals, requiring a meticulous approach to ensure compliance and brand safety. 

We prioritize protections in our programmatic efforts by implementing robust policies and adopting technologies that enhance anti-fraud measures, brand safety, and campaign effectiveness. As early adopters of innovative solutions in these areas, we continuously refine our strategies to align with industry best practices and deliver results that meet the high standards of our clients. 

Our Agency Reaction to MFA’s 

The issue with MFA websites is nuanced because they are not inherently fraudulent. These sites typically generate real traffic from legitimate human users—not bot activity, as verified by platforms like Google. Impressions, clicks, and conversions from MFAs are generally valid, and in some cases, they can even deliver cost-effective results. According to Google, their DSP and algorithmic optimization will naturally deprioritize MFA sites if they underperform. The rationale is simple: if these sites did not deliver results, traders and advertisers would move away from them, making their inventory less viable. But the reality is, MFA sites remain in the ecosystem because it is possible that they can provide a lower-cost alternative for advertisers and, in certain cases, meet campaign performance metrics. However, their long-term value is questionable due to potential trade-offs in user experience and brand reputation. 

Even with these possibilities, the presence of MFA websites in the programmatic ecosystem continues to raise ethical and legal concerns. While some MFA sites operate within the boundaries of legitimate practices, others employ illegal tactics like fraud and intellectual property infringement. If a publisher prioritizes profit over legality, it calls into question the integrity of their broader business practices. What other corners might they cut? What other risks could their operations pose to brands and agencies? 

This shady business approach represents an unacceptable risk for both agency and client alike. For this reason, recognizing the risks associated with MFAs, we prioritized the identification and adoption of a comprehensive MFA solution as a key agency goal for 2024. This commitment reflects our dedication to safeguarding client campaigns and ensuring the highest standards of quality, compliance, and effectiveness in our programmatic efforts. 

Butler/Till’s MFA Mitigation Timeline 

• Summer/Fall 2023: Introduced to the concept of MFA websites and began discussions with current vendors about implementing blockers. Early versions of these tools were launched. 

• Late 2023: Insights from the 2023 ANA report highlighted gaps in early MFA mitigation technologies. 

• Q1 2024: Strategized requirements for a robust MFA solution, conducted market research, and selected Jounce Media as our supply chain research partner. 

• April 2024: New MFA blockers were rolled out across all campaigns and DSPs. Activations were handled on a rolling basis, with full implementation completed before the end of the month. 

• May–July 2024: Post-activation period, during which the effectiveness of the new blockers was monitored and evaluated. 
   

How MFA Got to Our Campaigns by DSP: 

Before leveraging Jounce Media’s source data to build our restriction lists, we consulted with several inventory and brand safety partners. While most assured us they had policies in place to combat MFA sites and were developing further solutions, these measures were not sufficient to prevent MFA appearances in our campaigns. 

Equipping our trading teams with the ability to directly implement MFA restrictions proved a turning point. This provided us with the control needed to maintain rigorous quality standards, allowing for more frequent updates and ensuring greater protection across all campaigns. 

Before = Jan 1 – March 31^st 

After = May 1 – July 31^st 

Notes:  

• Although our overall MFA rate of 6.76% was much lower than the averages reported by the ANA (~21%) it still represented a significant overall amount of investment, and a higher rate than we were comfortable with.  

• Applying the Jounce Media data research, using strict updating processes and the “safest” policies for interpreting, we were able to effectively eliminate MFA sites from our buys.  

 
Performance Impacts of Removing MFA: 
Having effectively eliminated MFAs from our campaigns, our attention shifted to observing performance impacts. To create a meaningful comparison, we chose metrics that could be standard across a wide variety of campaigns, channels, and objectives. Therefore, we looked at Ad-Account level performance for the following metrics: CPM, CTR, Viewability, Fraud, and CPA. These metrics allow us to get a holistic view as they are reported by various platforms including the DSP, the Brand Safety platform, and the Ad Server.  
 

CPM Analysis 
We looked at the CPMs within each ad account on a before and after removal of MFA basis to ascertain the impact.  

What we found is that the average CPM decreased by 3.2%, indicating that removing MFA traffic did not lead to an increase in costs overall 
 
We see that about 30% of ad accounts investigated did experience a rise in CPMs. An exploratory analysis into these identified that they are primarily HCP campaigns that had extenuating factors and optimizations that drove these costs up. They also tended to be much lower spenders than other ad accounts. This causes the average change of CPMs across all ad accounts to be 20%, however, the median is 1.2% difference.  

CTR Analysis 

We looked at the CTR within each ad account on a before and after removal of MFA basis to ascertain the impact. 

Here’s what we found:  

About half of the ad accounts experienced an improvement in CTR after implementing the MFA block lists. The overall impact was the CTR was about 7% lower. This conclusion was not necessarily surprising. MFAs have been lucrative by their operators because they appear to perform, allowing them to be included in DSP targeting.  
 
However, CTR is a metric that is easily influenced by things like accidental clicks, percentage of traffic going to mobile vs. desktop, click fraud, and more. Therefore, we think it is important to take another step in the analysis and look at the quality of those clicks. To do this, we leveraged our Ad Server reporting (instead of DSP reporting) as the Ad Server typically has the most stringent click filtering methodologies including deduping, scrubbing bot traffic, and more.  

Here’s what we found:  

Using the ad servers’ click macros instead of the DSP’s shows an overall CTR increase of 32.5%. Although it is a nominal overall gain (from 0.08% to 0.1%) it does at least start to paint a picture that MFAs can demonstrate reasonable CTR, however, upon further scrutiny (and more aggressive click filtering by the ad server) having the budget be filled by non-MFA inventory will help support better outcomes.  
 

Conversion Analysis 

We looked at the conversion analysis within each ad account on a before and after removal of MFA basis to ascertain the impact. 

Here’s what we saw in terms of CPAs on non-healthcare/pharma accounts: 

 
While about half the accounts in question improved, half did not, and the overall change appears to be a net negative.  

This was not unexpected, and because these are raw conversion figures from floodlight activity without the assistance of proper attribution models, we needed to look one step closer to identify cases where the inventory before may have been assisted by conversions that took place by the user, but the ad in question did not drive that action. So, we took a secondary look at only the click-through conversions, excluding all view-throughs. 

Here we see the CPA of these ad accounts improved by 16% on non-healthcare accounts when focusing on click-through conversions only. This shows that while MFA sites often generated a large number of view-through conversions (due to serving massive volumes of ads), these conversions likely did not reflect real consumer engagement or behavior. Instead, the ads on MFA sites had minimal influence on users. Once we replaced MFA inventory with legitimate, higher-quality sites, we saw a clear increase in directly attributable conversions, meaning these ads were more effective at driving true consumer action. While total conversions (including view-through) may have dropped, the increase in meaningful, click-through conversions confirms that non-MFA traffic is much better at delivering real results. 

The above analyses of conversions are only highlighting non-healthcare/pharma campaigns for an important reason. Due to HIPPAA regulations, direct attribution becomes unfeasible. To measure the success of our campaigns within this vertical we are generally leveraging specialty 3^rd party measurement solutions who are compliantly matching script data with tag data in a clean room environment to demonstrate “script lift.” Some examples of these companies include IQVIA, Crossix, and Komodo. Because the matching of these data sources requires very careful matching of signals, and in order to prevent “fingerprinting” – a technique where data can be reverse engineered to infer Personal Health Information (PHI) – these methods require longer periods of time and larger pools of data before the results can be shared back with the agency.  

Impact on viewability: 

MFA websites can significantly influence viewability within a campaign in several ways: 

1. Low-Quality Content: MFA sites often prioritize ad placements over high-quality content. This can lead to poor user experience and lower engagement, resulting in fewer opportunities for ads to be viewed. 

2. User Distrust: Users may be skeptical of MFA sites, associating them with clickbait or misleading content. This distrust can lead to higher bounce rates, reducing overall viewability. 

3. Ad Placement and Visibility: MFA sites frequently use aggressive ad placement strategies, which may lead to ads being blocked or overlooked by users. While some ads may load, their effectiveness can be compromised if they are not seen or engaged with. 

4. Short Attention Span: Users visiting MFA sites may have a short attention span, quickly scrolling or leaving without fully engaging with the content. This behavior can negatively impact the viewability of ads. 

5. Measurement Challenges: The nature of MFA sites can complicate the measurement of viewability and engagement metrics, making it difficult to assess the effectiveness of campaigns accurately. 

Viewability increased by a total of 8% across all accounts with five of the seven accounts increasing by at least 1.4% and two accounts viewability staying consistent. To optimize campaigns in the context of MFA sites, it’s essential to focus on high-quality, relevant placements and to ensure that the ad content aligns with user expectations to foster better engagement and viewability. 

Impact on Fraud 

MFA websites can contribute to increased fraud within digital advertising campaigns in several ways: 

1. Ad Misplacement: Ads on MFA sites may appear alongside misleading or irrelevant content, increasing the risk of click fraud, where users may click on ads without real intent due to the nature of the surrounding content. 

2. Fake Engagement Metrics: MFA sites can manipulate metrics, providing false impressions and engagement data that can mislead advertisers about the effectiveness of their campaigns. This can lead to wasted ad spend on fraudulent impressions. 

3. Bot Activity: Many MFA sites are susceptible to bot traffic, which can further distort campaign analytics. This can result in inflated CTR and skewed conversion metrics, making it difficult for advertisers to gauge true performance. 

4. Ad Network Risks: Some ad networks that partner with MFA sites may not prioritize brand safety or fraud prevention, increasing the risk of ads being served in low-quality environments that facilitate fraudulent activity. 

In the table above, you see fraud decreased by 3.24% across all associated accounts with three of the accounts seeing a decrease in fraud by at minimum 15%. To mitigate these risks, advertisers should employ stringent vetting processes for ad placements, utilize fraud detection tools, and focus on partnering with reputable publishers to ensure higher quality traffic and reduce exposure to fraudulent activities. 

Conclusion: 

Conclusion 

The removal of MFA inventory from our programmatic campaigns yielded clear benefits in key performance metrics.  

• What improved: After removing low-quality MFA traffic, we saw several key improvements. CPM rates generally decreased, showing that eliminating poor-quality inventory did not lead to higher costs. At the same time, CTR improved, as reflected in our ad server data, indicating better engagement with higher-quality inventory. Additionally, cost per acquisition (CPA) metrics for non-healthcare accounts improved when focusing solely on click-through conversions, confirming that legitimate traffic was more effective at driving meaningful consumer actions. These results highlight the benefits of prioritizing quality over quantity in our campaigns. 

• What did not improve but why it is okay: While overall CTR saw a slight decrease, this was likely influenced by factors like accidental clicks and click fraud, which are common with MFA traffic. Similarly, the reduced total CPA was anticipated, as MFA-driven view-through conversions had previously inflated performance metrics. In both cases, we believe the improved quality of interactions outweighs these apparent drops in numbers, as our focus is on driving true business outcomes rather than relying on vanity metrics. The shift to more legitimate traffic ensures more meaningful and sustainable results for our campaigns.  

• Final thoughts: While a few ad accounts saw increases in CPMs and declines in CTR, we attribute these changes to specific campaign nuances. Despite this, we are confident that moving away from MFA inventory has strengthened both the integrity and effectiveness of our campaigns, especially in terms of improving conversion quality. Looking ahead, we will continue to refine our inventory selection to ensure sustained, high-impact results for our clients. 

Contributors: Ryan Lammela