Variance explained in true incremental conversions per £ spent
Last-click attribution: industry standard
R² = 0.19
81% of the signal in this instrument is noise interpreted as performance data.
Causal inference model (PIE methodology)
R² = 0.88
Trained and validated against 2,226 randomised controlled trials.
Gordon, Moakler, Zettelmeyer: Predicted Incrementality by Experimentation (PIE) for Ad Measurement
NBER Working Paper 35044 · Kellogg School of Management / Meta Ads Research · April 2026

What attribution actually measures

Multi-touch attribution and last-click models share a structural flaw that is not fixable through better data or more sophisticated weighting. They measure correlation between ad exposure and conversion events. They do not isolate whether the advertising caused the conversion.

The distinction matters because most conversion events would have occurred regardless of the advertising. A customer already in the purchase funnel, already brand-aware, already committed to buying, will convert after seeing an ad and be recorded as an ad-driven conversion. The model assigns credit. The conversion was not incremental.

Google's Meridian documentation describes this directly: the framework integrates organic search query volume specifically to separate category-level demand from advertising-driven demand. The reason that feature exists is that without it, the model would conflate the two. Most measurement tools in production today do exactly that.

6–12%
True incrementality on paid brand search in B2B contexts. The remainder is captured by organic regardless of paid activity.
Ebiquity: multiple independent replications
2–4×
Overstatement of ROI in multi-touch attribution versus holdout-validated incrementality, for retargeting and brand search combined.
HBS / Marketing Science Institute
54%
Of US brand and agency marketers report no improvement in measurement confidence year over year, despite increased tooling investment.
EMARKETER / TransUnion · July 2025

Why platform-native measurement does not resolve this

Google's Meridian is a technically serious framework. Its limitations are structural rather than methodological. Meridian is built by a company that sells media. Its priors, geo calibration infrastructure, and optimisation recommendations are designed to help advertisers allocate budgets more efficiently, with Google properties available as an allocation target.

An advertiser who uses Meridian integrated into Google Analytics 360 as their primary measurement framework has made Google the default arbiter of Google's own media effectiveness. The quality of the underlying methodology does not resolve this conflict of interest.

The independent MMM vendors operate at enterprise scale: complex, expensive, and designed for organisations with internal data science capacity. The 40% of organisations that HBR Analytic Services identified as unable to translate MMM outputs into business decisions are not being served by enterprise-grade consulting engagements. They are being served by platform-reported attribution, because that is what their dashboards show.

Where incrementality testing gets closer, and where it stops

Incrementality testing, run through geo holdouts, matched-market experiments, or platform conversion lift studies, answers a narrower question than attribution and answers it correctly: did this specific piece of media cause additional conversions that would not have happened otherwise. That is a genuine improvement on multi-touch and last-click models, which assign credit without testing causation at all.

The limitation is scope, not method. A holdout test tells you whether one campaign, in one period, produced incremental lift. It does not tell you where the next pound of budget should go across the full portfolio of channels, categories, and demand segments a business operates in. Running enough concurrent experiments to answer that question requires sample sizes and testing infrastructure that most organisations below enterprise media spend cannot sustain.

The Brand Demand Scan starts from the same causal premise as incrementality testing, that correlation between spend and conversion is not evidence of effect, but applies it to the full category demand picture rather than to individual campaigns. The output is not a lift percentage for one test. It is a structural map of where demand exists, where it is captured, and where the allocation decision should move next.


What the Brand Demand Scan measures, and why the distinction matters

The Brand Demand Scan does not begin with a media plan. It begins with a question: of all the search demand that exists in your category, how much of it finds your brand?

The gap between category demand and brand capture is not a marketing performance metric. It is a measure of structural leakage in the commercial pipeline. It quantifies demand that exists, that the organisation did not create, that is being captured by competitors or by nobody.

Most measurement tools answer a different question. They ask: given the media we bought, what did it produce? The Brand Demand Scan asks: given the demand that exists in the market, how much of it is reaching us, and how much is not? The first question evaluates execution. The second evaluates strategic position. Optimising on the first without understanding the second is how organisations end up with efficient campaigns operating against an insufficient share of available demand.

Brand Demand Scan

Quantify your gap before deciding where to invest.

The BDS identifies the structural gap between category-level demand and brand-level capture from your existing data. Fixed price. No retainer. No implementation required.

View the Brand Demand Scan →
From £490 · Results from existing Google Search Console data