Multi-touch attribution is still worth using, but for a narrower set of purposes than it was designed for. It is a useful tool for understanding customer journey patterns and making tactical optimisation decisions. It is not a reliable tool for cross-channel budget allocation or proving channel ROI.
For several years, multi-touch attribution (MTA) was positioned as the solution to the "who gets credit?" problem in marketing measurement. Build a sophisticated enough model, the pitch went, and you could know exactly what each channel was worth. That promise has not held up.
What went wrong with MTA's promise
MTA models rely on user-level tracking data to map the full conversion journey. As third-party cookies disappear, as iOS restricts app tracking, and as consent rates decline, the data available for these models shrinks. An MTA model trained on 40 percent of actual journeys is making significant assumptions about the other 60 percent.
Even with complete data, MTA faces a fundamental challenge: correlation is not causation. A touchpoint that frequently appears in converting journeys is not necessarily the one that caused the conversion. Users who were already going to buy leave complex multi-touch trails just like users who were persuaded by advertising.
Where MTA still adds genuine value
- Journey pattern analysis: understanding how customers typically move through your funnel, which channels they encounter first, and which appear most often before conversion
- Creative and messaging insights: comparing journey patterns for different audiences or creative strategies to identify what content type appears most often in successful paths
- Tactical within-channel optimisation: using MTA signals to identify which keywords, placements, or audiences appear most frequently at key funnel stages
- Identifying gaps: spotting audience segments that convert but have thin journey data, suggesting they may have been influenced by channels that are hard to track
Where MTA falls short
MTA should not be used to set overall channel budgets, to evaluate channel effectiveness in absolute terms, or to compare channel ROAS across a portfolio. All of these require knowing the incremental contribution of each channel, which MTA cannot reliably provide.
The most common mistake is treating a channel's MTA-attributed revenue as equivalent to its incremental revenue. They are not the same. Attributed revenue includes baseline conversions. Incremental revenue is what the channel actually caused.
The future of measurement is a combination of methods: MTA for journey intelligence and tactical optimisation, incrementality tests for experimental validation of channel effectiveness, and marketing mix modelling for strategic budget allocation. Each tool has a job. Using one tool for all three jobs produces poor results.
What is replacing MTA for budget decisions
Marketing mix modelling has re-emerged as the primary tool for cross-channel budget allocation. Modern MMMs, particularly Bayesian frameworks, are faster to run and more accessible than their predecessors. They work from aggregate business data rather than user-level tracking, which makes them resilient to privacy changes.
Incrementality testing complements the MMM by providing experimental validation of key channels. Together, they provide a more reliable basis for budget decisions than any MTA model can.
The vendor landscape
Many MTA vendors are repackaging their products as unified measurement solutions that combine MTA with MMM or with incrementality components. Some of these are genuinely useful integrations. Others are marketing rebrands. Evaluate them based on how they handle the known limitations of MTA, not on the sophistication of their platform interface.
Should I cancel my MTA contract?
Not necessarily. If your MTA tool provides useful journey analysis and tactical signals, it is still earning its cost. Evaluate it on what it actually delivers: customer journey insights and tactical optimisation signals. If you are using it primarily for budget allocation decisions, that is where the review should focus.
What about first-party data MTA? Does that solve the tracking problem?
First-party data MTA, which uses only data you own such as email addresses and login IDs, solves the tracking problem for known customers. But it typically covers a minority of your total conversions. For the majority of buyers who do not log in or identify themselves, you still have incomplete journey data.
How do I explain to leadership why we are moving away from MTA for budget decisions?
Frame it as using the right tool for each job. MTA tells you how customers move through your funnel, which is useful for journey design. It does not tell you what your advertising caused, which is what budget decisions require. Incrementality tests and MMM answer the causal question more reliably. Present the first incrementality test result alongside the MTA data for the same channel: the gap between them makes the case.