MMM aggregates, MTA traces individual paths
Marketing Mix Modeling (MMM) treats the entire dataset as a time series, asking how overall spend across all channels correlates with revenue week by week. It finds the forest: TV spend elasticity, search spend elasticity, and the baseline revenue that happens regardless. Multi-Touch Attribution (MTA) disaggregates, tracing individual user journeys across touches: the email that started the conversation, the search ad that closed it, the social post in between. MMM answers 'what would revenue be if we paused TV?' MTA answers 'who should get credit for this conversion, and should we spend more to reach similar people?' MMM is historical and macro. MTA is journeys and micro.
Complementary blind spots
MTA struggles with offline channels (TV, radio, billboards) because it has no individual event to attach credit to. MMM can model them but often can't explain which specific audiences. MTA assumes all touches are tracked, which is rarely true; untracked word-of-mouth and brand awareness leave gaps. MMM handles those gaps implicitly in the 'unexplained variance' but can't point to them. Most mature marketing organizations run both: MMM to set channel budget allocation at the portfolio level, MTA to optimize bidding and audience targeting within each channel. Using only one leaves half the signal behind.
Model choices matter within each approach
MMM models can be naive (simple correlation) or sophisticated (adstock with saturation curves, control variables for seasonality and external events). Naive MMM overestimates high-spend channels and underestimates channels that compound (brand building). MTA models range from first-touch and last-touch (discredited but still used) to time-decay and algorithmic (ML-based). First-touch MTA rewards awareness but ignores conversion pressure. Last-touch MTA credits the final touchpoint but ignores all upstream work. The model choice matters more than the framework; a well-built time-decay MMM and a naive algorithmic MTA can give opposite recommendations for the same business.