Multi-Touch Attribution · The Promise and the Problem
MTA models distribute credit across the touchpoints leading up to a conversion. The math, the modeling choices (linear, time decay, U-shaped, data-driven), and why most teams are quietly retiring MTA in favor of incrementality and MMM.
Attribution. Multi-touch attribution has roots in econometric and marketing science work going back decades. Google's data-driven attribution (DDA, originally Shapley-value-based) and Facebook's attribution products popularized MTA in the digital era. This article reviews the techniques and the modern critique.
What MTA tries to do
Multi-touch attribution (MTA) distributes credit for a conversion across the touchpoints the customer encountered before converting. Instead of giving all the credit to the last click (or the first), MTA splits credit based on a model — linear, time-decay, U-shaped (W-shaped), or data-driven.
The promise: a fair view of which marketing investments deserve credit. The reality has been more complicated.
The common MTA models
First-click. 100% credit to the first touchpoint. Over-credits acquisition channels.
Last-click. 100% credit to the last touchpoint. Over-credits closing channels (often paid search and retargeting).
Linear. Equal credit to every touchpoint. Treats them all as equally important, which they're not.
Time-decay. More credit to touchpoints closer to conversion. Reasonable for short cycles, less for long.
Position-based (U-shaped). 40% to first, 40% to last, 20% spread across the middle. A compromise.
W-shaped. Same idea with a third weighted touchpoint at lead creation. B2B variant.
Data-driven. Statistical model (often Shapley values) learns credit from observed paths. Most sophisticated but most opaque.
Why MTA is quietly being retired
Three structural problems have eroded MTA's credibility:
Privacy changes broke the data. iOS ATT, ITP, ETP, and Privacy Sandbox all reduce the user-level tracking MTA depends on. Most MTA today is operating on degraded data and producing increasingly fictional models.
Dark social and offline are invisible. MTA only credits trackable touchpoints. A buyer who heard about you on a podcast, saw a billboard, and got a recommendation in a Slack DM appears to have arrived "direct" — and MTA invisibly gives all credit to whatever was clickable.
Correlation isn't causation. MTA assumes the touchpoints in a converter's path caused the conversion. They may have just been present. Incrementality testing reveals the difference.
The 2026 measurement triad is replacing MTA. Media Mix Modeling for cross-channel allocation. Incrementality tests for causal validation of specific bets. Brand lift studies for upper-funnel impact. MTA remains useful for short cycles in highly trackable environments — but it's no longer the default attribution methodology for sophisticated teams.
When MTA still works
For short conversion cycles (under 7 days) in heavily-tracked digital-only journeys (e.g., a DTC purchase after seeing 3 paid social ads and an email), MTA still produces reasonable, useful credit allocation. The same model fails for B2B with 90-day cycles or DTC with significant dark-social influence.