Cross-Channel Attribution

Cross-channel attribution exists to solve the fragmentation problem in modern customer journeys. A customer typically encounters 5–8 touchpoints before purchasing, across different channels and devices. Single-touch attribution models (first-click or last-click) arbitrarily assign all credit to one end of that journey. Cross-channel models distribute credit across the full path, using rules (linear, time-decay, position-based) or machine learning (data-driven attribution) to weight each touchpoint.

The core challenge is that cross-channel attribution requires tracking users across sessions, devices, and channels — a technical capability that has been significantly degraded by iOS 14, cookie deprecation, and the proliferation of ad-blocking. Even the best cross-channel model today has significant gaps, particularly in cross-device journeys and iOS user paths.

Exactius uses cross-channel data-driven attribution (DDA) in GA4 as a directional input for understanding journey patterns, but does not use it as the governing signal for capital allocation. The Capital Allocation Loop relies on MER and incrementality data, which do not depend on cross-channel tracking.

Cross-channel attribution is important because customer journeys are genuinely multi-channel — optimising each channel in isolation will not optimise the system. A brand might cut email on the basis of low attributed ROAS, not realising that email is the touchpoint that converts prospects generated by Meta prospecting campaigns into purchasers. Remove email, and Meta prospecting performance degrades.

The insight cross-channel attribution uniquely provides is channel interaction effects — which channels work together, which cannibalize each other, and which are genuinely additive. This is the level of understanding needed to optimise a full-funnel growth system, not just individual channels.

For LTV:CAC measurement, cross-channel attribution helps identify which channel combinations produce the highest-value customers — because the customer journey often predicts the customer's future behaviour. Multi-touch customers (who engage with brand, consideration, and conversion channels) often have higher LTV than single-touch customers.

Cross-channel attribution models

Linear: Equal credit to all touchpoints. Simple, removes last-click bias, but ignores which touchpoints were more influential.

Time decay: More credit to recent touchpoints. Better for short consideration cycles; still biases toward bottom-of-funnel.

Data-driven (DDA): Uses ML to weight touchpoints based on their observed contribution to conversion. Best available model, but requires 300+ conversions per month and significant traffic volume. Available in GA4 and Google Ads.

None of these models is immune to tracking gaps. Supplement cross-channel attribution with MER and incrementality testing to catch the conversions that attribution models cannot see.

Cross-channel attribution is more honest than last-click, and data-driven attribution is the best version of it. But it remains a model that distributes credit based on observed patterns — it cannot distinguish causation from correlation, and it cannot see cross-device journeys where cookies are blocked.

Exactius uses DDA in GA4 to understand journey patterns and channel interaction effects, but cross-references those findings with MER and holdout data before acting on them. The goal is convergent evidence: when DDA, MER, and incrementality data all point the same direction, the capital allocation signal is reliable. When they diverge, it flags a measurement gap to investigate.

The Growth Operating System uses cross-channel attribution for what it is good at — understanding the sequence and interaction of channels in the customer journey — and incrementality testing for what DDA cannot do: measuring causal impact free from attribution assumptions.

Exactius embeds growth squads that maintain cross-channel attribution infrastructure in GA4 as a directional analytical tool, while anchoring capital allocation decisions in the causal measurement layer of the Capital Allocation Loop.