Multi-touch attribution has had a strange decade. The mid-2010s hype cycle promised a definitive accounting of every marketing touch. The post-IDFA period exposed the methodological limits. The 2024-2025 period saw a rationalisation. By mid-2026 the practitioner community has settled into a more honest position on what multi-touch attribution can and cannot do, and the working stack looks different from the 2019 stack.

What practitioners agree on in May 2026:

Last-click attribution is still the simplest and most reliable reporting view, even if it is not the most accurate. The teams that report on last-click and use it as the default for tactical optimisation are not embarrassed about doing so. The reporting bias is well understood and the operational utility is high.

Data-driven attribution within an advertising platform’s walled garden (Google, Meta) is useful for in-platform optimisation but is not directly comparable across platforms. The teams that try to roll up Google’s data-driven attribution and Meta’s data-driven attribution into a single cross-channel view are producing a number that is methodologically incoherent. The teams that use each platform’s attribution for in-platform decisions and a separate cross-channel methodology for strategic decisions are running cleanly.

Media mix modelling has returned as the serious cross-channel methodology. The Bayesian media mix modelling tools — Google’s Meridian, the Robyn open-source library, and the commercial offerings from the major analytics consultancies — are doing the cross-channel attribution work that the deterministic multi-touch models tried and failed to do. The MMM outputs are at a strategic time-scale rather than tactical, but the methodology is sound and the decision support is real.

Incrementality testing has earned its place. The teams running structured geographic holdouts, intent-to-treat designs, and synthetic-control evaluations are getting the answer to “did the campaign work” with the rigour that attribution alone cannot deliver. The incrementality test is more expensive than reading an attribution dashboard but it is the only way to answer the causal question.

What has been quietly retired:

The deterministic multi-touch attribution model that traced every user across every touch and assigned weighted credit. The post-IDFA cookie loss made the underlying data too sparse to support the methodology. The vendors who sold this in 2019 are mostly out of the business or have repositioned around MMM and incrementality.

The “single source of truth” attribution dashboard that promised to reconcile all advertising platform data into a unified view. The walled-garden data limitations made this impossible to do well, and the teams that tried have walked back to a more layered measurement stack.

What the working 2026 stack looks like:

A platform-level data-driven attribution for in-platform optimisation, accepting that the platforms are doing their own attribution and that this is fine for tactical purposes.

A cross-platform last-click reporting view for tactical decision support, with clear understanding that it under-credits upper-funnel work.

A regular media mix modelling exercise — quarterly or monthly — for strategic budget allocation, run either by an in-house data science team or by a specialist vendor.

A planned programme of incrementality tests across the major channels, with the tests sized for statistical reliability and the cadence sized for ongoing operational utility.

A clean data foundation — server-side tagging, first-party customer identity, consent governance — that supports all of the above.

For Australian marketing teams operating in May 2026, the read is that the attribution conversation is in a healthier place than it was three years ago. The teams that match the methodology to the question are getting clean answers. The teams that demand a single number from a single methodology are working from a 2019 expectation that has not been met by the 2026 measurement environment.

The 2026 best-practice attribution programme is multi-method, layered, and honest about uncertainty. The teams that have got there are making better budget allocation decisions than the teams that have not.