Multi-Touch Attribution (MTA)

MTA operates by stitching together the sequence of ad interactions a customer had before converting, then distributing credit across those interactions according to a model. Rule-based MTA models (linear, time-decay, position-based) apply fixed weighting formulas. Algorithmic MTA models use machine learning to infer each touchpoint's contribution from observed conversion path patterns.

The fundamental limitation of MTA is that it can only measure touchpoints it can see. After iOS 14, a large share of paid social impressions and clicks are invisible to MTA models because the users who engaged with those ads opted out of cross-app tracking. This creates a systematic gap: MTA under-reports the contribution of Meta and other iOS-heavy channels, and over-reports channels with better tracking coverage.

Exactius uses data-driven MTA (via GA4) as a directional signal for channel interaction and journey analysis, but does not use MTA outputs as the primary input for capital allocation decisions. Those decisions are governed by the Capital Allocation Loop, which relies on MER and incrementality data.

MTA matters because it exposes channel interaction effects that single-touch models miss entirely. A brand using last-click attribution might conclude that its email program is underperforming based on low attributed ROAS — without realizing that email is the final touchpoint that converts prospects that Meta prospecting introduced to the brand. MTA reveals these dependencies.

The value of MTA is not in its attribution accuracy — it is systematically incomplete. The value is in the journey patterns it reveals: which channels appear early in paths, which appear late, which channels appear together in high-value customer journeys. These patterns inform full-funnel strategy in ways that channel-level ROAS never can.

For businesses building toward high LTV:CAC, MTA path analysis helps identify which customer journeys produce the most valuable customers — not just which journeys convert. A short, single-touch journey through branded search might convert efficiently but produce lower-LTV customers than a multi-touch journey that begins with brand awareness content.

Implementing MTA in practice

GA4's data-driven attribution is the most accessible MTA tool for most mid-market brands. It requires 300+ conversions per month and uses a counterfactual approach — comparing converting paths against non-converting paths with similar touchpoints — to estimate each touchpoint's marginal contribution. Enable DDA as the default attribution model in GA4's attribution settings.

Dedicated MTA platforms (Rockerbox, Northbeam, Triple Whale) offer more granular cross-channel tracking for brands with complex media mixes. These tools invest heavily in identity resolution to stitch cross-device journeys — though their coverage for iOS users remains limited.

Critical limitation: All MTA tools will underreport channels with poor tracking coverage. Cross-reference MTA outputs with channel-specific incrementality tests to identify where MTA is missing significant volume.

MTA spent a decade being positioned as the solution to last-click attribution's failures. It is better than last-click — but iOS 14 revealed that MTA's accuracy was always contingent on the quality of user-level tracking, and that tracking was never as complete as the industry assumed.

Exactius's position is that MTA should be used for journey analysis and channel interaction insight, not for capital allocation. The capital allocation decision requires causal measurement — which MTA cannot provide. MTA tells you which paths were observed; incrementality testing tells you which paths caused revenue.

The Growth Operating System uses MTA as one input layer — useful for understanding how channels relate to each other in the customer journey — alongside MER and incrementality data. No single layer is sufficient; the Capital Allocation Loop requires convergent evidence from all three.

Exactius embeds growth squads that implement GA4 data-driven attribution as the baseline MTA layer, supplement it with third-party MTA tools for brands with complex multi-channel mixes, and validate MTA channel attribution with holdout tests on an ongoing basis.