Platform bias has long been a known problem, whether it be Google, Microsoft, Meta or Tik Tok. Marketers, and more crucially, finance and C-suite, maintain a healthy distrust of in-platform attribution. Attribution is, by nature, ‘greedy’ as the platforms attribute any conversion that has the platform as a touchpoint back to that platform, and the numbers rarely add up across a diverse media mix.
However, a more acute and recent crisis has emerged. The foundation of the attribution system, reliable user journey tracking, is deteriorating. For years, the ‘trust contract’ rested on this foundation: even if value attribution across touchpoints was controversial, the underlying user journey data was solid. Now, that is no longer the case.
The most recent shockwave was the move by Safari in September to strip click IDs, including Google Click Identifier (gclid) and Microsoft Click Identifier (msclkid) from URLs in standard browsing sessions, a measure previously limited to Private Browsing. Without these click IDs, the essential mechanism for mapping conversions back to a user journey fails, leading to an immediate rise in unattributed conversions.
This stacks onto historic tracking degradation such as Apple’s App Tracking Transparency implementation in 2021, requiring apps to surface users a consent banner, for those opting out, meaning advertisers lose tracking of users going through to site from apps such as Instagram and FB.
More untracked conversions and a greater reliance on platform ‘black box’ modelling only deepen the mistrust in platform-reported ROI. Marketers increasingly need a new source of truth.
In response to the need for better evaluation, platforms offer incrementality testing solutions such as Google’s GeoLite and Meta’s GeoLift. While these can offer directional insight, they come with fundamental limitations that prevent them from restoring trust in measurement.
Stakeholders who are further removed from the digital ecosystem, such as CFOs, are rightly wary of in-platform testing. Their distrust of platform attribution often extends to platform-owned testing tools because of their suspected bias.
In addition, the key performance indicator (KPI) for these tests is often restricted to in-platform attributed revenue, rather than objective, business-level metrics. This confines the robustness of the findings and adds to the distrust.
The most critical limitation is scope. In-platform tests can only measure the activity of that particular platform. They cannot test or measure the incremental impact of non-digital efforts, such as Above The Line, Out Of Home, and TV which form a vital part of the media mix. They also cannot test or measure the impact of non-media efforts such as store openings, product launches.
How do we solve this? One solution we’ve been working with is independent, statistically rigorous geo-testing.
In an environment where platform attribution is failing and trust is eroding, geo-testing is emerging as the scientifically grounded, independent solution that can restore confidence in how we measure the impact of marketing activity. Better measurement ultimately means better investment decisions.
Geo-testing works because it isolates the impact of media activity on an objective business KPI (e.g., store visits, sales, business-level revenue) by comparing performance across highly correlated control and test regions.
Done transparently and with rigorous implementation, the methodology provides a clean, causal link, answering the C-suite’s most important question: What is the true incremental value of this investment?
At Kinase we use an approach that combines permutations testing and bootstrapped resampling to determine the statistical significance of results and produce confidence intervals. By generating a null hypothesis distribution simulating where marketing activity had no effect, we can assess whether our observed outcome sits within this distribution, this allows us to discern signal from noise in a meaningful way which client teams and stakeholders can trust.
For one of our clients, our geo-analysis following a major, geo-targeted TV brand awareness campaign found its cost per lead (CPL) to be twice that of our estimated CPL for YouTube branding activity. Armed with this information, the client is now pivoting a significant portion of their brand budget to a YouTube-focused strategy for the new year, which itself will be subjected to a new geo-test to examine the hypothesised CPL at scale.
This is the power of robust measurement: it doesn’t just produce a graph; it drives high-value, data-led decisions.
Every pound spent at a lower marginal ROI than in your strongest channel is a pound that could have been better invested. In an increasingly uncertain landscape, with inventory and algorithmic changes like the Meta Andromeda update, or the emergence of AI summaries on the Google SERP, prior assumptions, however right or wrong, must be constantly updated.
More robust and statistically grounded measurement solutions are not a silver bullet, but they provide the essential, trusted hand that can guide media investment. An agile, flexible measurement approach is critical to test a wide range of areas quickly, ensuring you truly know your channel’s worth and are never investing blindly.
Paul Artunduaga is an analytics engineer at Kinase
