A measurement approach that distributes conversion credit across the multiple touchpoints in a customer journey using user-level data.
Multi-touch attribution assigns proportional credit to the several touchpoints along a customer's journey, rather than handing all credit to a single interaction. WhatConverts defines it as a strategy that distributes credit across multiple touchpoints, and notes it requires user-level touchpoint data (whatconverts.com).
There is a family of models. Impact.com lists last-click (all credit to the final interaction), first-touch, linear (credit split equally), time-decay (more credit nearer the conversion), position-based or W-shaped (weighted to first, lead, and last), and data-driven (algorithmic) (impact.com). Each rule produces a different ROAS for the same campaign, which is the core limitation: the answer depends on the model you chose.
MTA is correlational, not causal. It allocates credit among touchpoints that were observed; it cannot tell you whether a conversion would have happened without a given touch. It also depends on durable user-level tracking, which privacy changes have steadily eroded. For causation, pair it with incrementality (holdout tests) and marketing mix modeling (aggregate, privacy-resilient).
Under Blufire's golden rule, MTA, and especially the last-click default in most ecommerce reporting, is framed as reported attribution, not real channel truth. It is a useful map of the journey, not a verdict on what to fund.