Boosting iROAS by 94% with Meta’s Incremental Attribution

The Brief

Through our deep experience growing ecommerce brands with ads on Meta, we know that Meta’s default attribution settings (7-day click and 1-day view) do not accurately reflect the business impact of different audience and optimisation strategies. An example most paid social marketers are familiar with, is that when we target colder audiences by excluding recent website visitors and existing customers, the reported metrics appear worse despite these campaigns actually driving more value for the brand. 

As Meta partners for both paid media and measurement, we were able to test the beta release of their new “Incremental Attribution” setting, to see if it can improve new customer acquisition for Pom Pom London. The beta release gives advertisers a new attribution lens to assess results by, but, more importantly, it optimises campaigns for events as measured with this new method. 

“Incremental Attribution” leverages data from previous conversion lift studies on the platform to make a judgement on whether conversions were likely incremental or not (i.e. would they have happened anyway, even if we hadn’t served an ad?). This is crucial because as an advertiser, you want your activity to be held to account, driving real business value, and not just claiming results that were not directly caused by your efforts.

With increasing privacy measures in place, retargeting audiences are less reliable, so “Incremental Attribution” offers a solution where we can target broad audiences, whilst still having ad delivery optimised using high quality signals from the e-commerce store (purchases which were likely actually caused by our advertising efforts). In short, it enables our ad-spend to work harder for us to achieve the business objectives we truly care about.

The Approach

To test Incremental Attribution, we implemented a comprehensive testing strategy:

1. We conducted a Conversion Lift Study comparing two campaign groups against hold-outs who saw no ads:

• Control: Standard campaigns using Meta’s default attribution settings
• Test: Identical campaigns utilising Meta’s new ‘Incremental Attribution’ setting

1. We employed Meta’s Robyn framework (their AI-powered Marketing Mix Modeling tool) to deepen our understanding of the impact, calibrating it against real-world data from the Conversion Lift Study
2. We cross-referenced findings across multiple data sources: Meta Ads Reporting, website event data, and the Conversion Lift Study results

The Results

Our analysis revealed striking disparities between different measurement methods, highlighting the significant business impact of moving from standard campaigns to Incremental Conversions. When comparing Incremental Conversions optimised campaigns to the BAU (7-day click and 1-day view) campaigns we found that:

• The Conversion Lift Study demonstrated a 39% ROAS increase
• Robyn’s analysis estimated a 57% ROAS improvement
• Incremental ROAS (the revenue gained from increasing spend on marketing) improved 94%
• Traditional reporting tools (Meta Ads and Google Analytics) showed minimal improvement (<3% ROAS change/difference)
• We are able to apply an incrementality factor to recalibrate in-platform targets to be representative of their true value, and remain confident they were driving profitable growth for the brand

Through experimentation and Marketing Mix Modeling, we uncovered critical insights into upper-funnel digital marketing strategies. While Google Analytics (using last-click attribution) and Meta Ads showed minimal value in switching to ‘Incremental Attribution’ campaigns, the Conversion Lift Study and our Media Mix Modeling told a very different story where the new campaigns drove up to 57% better ROAS. For brands willing to move beyond last-click measurement of their marketing mix, the opportunity to get ahead is significant.

For our ‘Incremental Attribution’ campaigns, we could compare the reported conversions from the new attribution setting to the default (7dc 1dv) reporting method. Unsurprisingly to us, Incremental Attribution reporting showed fewer conversions than the 7-day click 1-day view. However, this lower number aligned much more closely with our Conversion Lift test, indicating that the reported conversions from Incremental Conversions were closer to the ‘truth’.

Future Impact

With the help of Robyn, our media mix models revealed significant opportunities for growth through increased investment in TikTok, generic search and brand awareness activities on Meta, encouraging the brand to further test diversifying marketing spend.

The success of this approach highlights the importance of going beyond surface-level metrics and traditional click-based reporting to uncover the true impact of digital advertising. To drive meaningful business growth, brands must adopt a measurement framework that not only provides a holistic view of performance but also aligns with their specific objectives. Crucially, achieving this requires expertise in selecting the most appropriate measurement methodologies – such as Media Mix Modelling or incrementality testing – and applying them to the most relevant business questions. Without this strategic approach, brands risk optimising for short-term gains rather than long-term, sustainable growth.

Meta’s Robyn analysis framework allows us to build complex models and generate useful outputs for decision-making with a strong workflow. By calibrating our Marketing Mix Modelling with lift test results in Robyn, we can make better models to support powerful commercial decisions across all marketing activities where ad-platform and multi-touch attribution metrics are not adequate.