Multi-Touch Attribution Is Expensive, Complex, and Usually Wrong
The attribution industry has sold you a fantasy: if you just track every touchpoint precisely enough, you'll know exactly which channels drive value. The reality? MTA is a money pit that creates analysis paralysis while the answers you actually need come from much simpler methods.
Attribution Paralysis
We've seen brands spend £50-100k on attribution platforms, take 6 months to implement them, and then... make the same decisions they would have made with last-click data. The problem isn't the data - it's the organisational capacity to act on attribution complexity.
When your MTA model says Display deserves 12% credit for a conversion that last-clicked through Brand Search, what do you actually do differently? Increase Display budget by 12%? The connection between attribution insight and bidding action is weaker than vendors admit.
Model Complexity vs Utility
Every attribution model is wrong. Some are useful. The question is whether the additional complexity of MTA produces better decisions than simpler models:
- • Last-click: Biased but actionable. Everyone understands it. Good for daily optimisation.
- • First-click: Useful for understanding top-of-funnel acquisition sources. Rarely used for bidding.
- • Linear/time-decay: Marginally better than last-click but introduces model risk. Hard to validate.
- • Data-driven: Google's black box model that structurally favours Google channels.
- • Full MTA: Theoretically perfect, practically impossible. Requires complete data, cross-device tracking, and offline conversion stitching that doesn't exist in a cookie-less world.
What MTA Gets Wrong
MTA's foundational assumption - that observing touchpoints equals understanding causation - is flawed:
- • Correlation ≠ causation: A customer seeing a Display ad before purchasing doesn't prove the Display ad caused the purchase
- • Missing data: In a world of iOS privacy, ad blockers, and Consent Mode, you're missing 30-50% of touchpoints anyway
- • Cross-device gaps: A customer researches on mobile, buys on desktop. MTA often treats these as two separate journeys
- • Impression attribution: Should a YouTube ad impression that wasn't watched count as a touchpoint? MTA models differ wildly on this
The Commercial Alternative
Instead of trying to credit every touchpoint perfectly, focus on the questions that actually drive budget decisions:
- • "If I turn off Channel X, how much revenue do I actually lose?" - answered by holdout tests
- • "What's my true cost per new customer?" - answered by new-vs-returning segmentation
- • "Is this channel profitable after COGS and returns?" - answered by profit-level reporting
Incrementality Over Attribution
The shift from attribution thinking to incrementality thinking is the most important measurement evolution for ecommerce. Instead of asking "who gets credit?", ask "what's truly incremental?"
Run quarterly incrementality tests on your major channels. Holdout 10-20% of audience from each channel for 2-4 weeks. Compare purchase behaviour. The delta is your true incremental contribution. It's not perfect, but it's honest.
A Practical Framework
- • Daily operations: Use last-click for bid management. It's imperfect but actionable.
- • Weekly reviews: Layer in new-vs-returning analysis to understand acquisition efficiency
- • Monthly strategy: Cross-reference platform data with actual P&L to catch attribution inflation
- • Quarterly validation: Run incrementality tests to validate channel-level investment
Next Steps
Related Reading
More on attribution, measurement, and incrementality.