Attribution isn’t just harder to trust technically; it’s failing marketers in the room where budgets get defended.
Somewhere between 30% and 40% of the conversions marketing teams used to be able to track are no longer visible to them. Not lost to bad tagging or a broken pixel — lost structurally, to privacy regulation, cookie restrictions, and app-tracking permissions that most users now decline by default. That number isn’t a forecast. It’s already happened, and most measurement stacks are still built as if it hasn’t.
The consequences are showing up beyond the reporting dashboard. A recent Gartner survey found that nearly half of marketing leaders struggle to prove their value and gain recognition for their contributions internally — a credibility gap that tracks closely with the measurement gap.
This is the crisis underlying many marketing budget conversations right now. Multi-touch attribution — the model most teams have built their reporting around for the last decade — depends on being able to see a user’s full journey: every touchpoint, every device, every channel, stitched together into a single line that ends in a conversion. That assumption is what’s broken. Journeys now span connected TV, retail media, creator platforms, marketplaces, and offline moments that were never especially visible to attribution in the first place, and the identity signals that used to stitch a journey together across devices are disappearing by design, not by accident.
The result is a reporting problem that’s no longer staying within the marketing team. Attribution gaps are now showing up in finance reviews and budget defences, because a number that can’t be trusted doesn’t stay a marketing problem for long — it becomes a business one.
What Actually Broke
It’s worth being specific, because “attribution is broken” has become a vague enough phrase that it’s easy to nod along without registering the mechanics. Four things happened at once:
- Privacy regulations restricted what can legally be tracked.
- Browser- and device-level changes reduced what’s technically visible, even where tracking remains legally allowed.
- Walled gardens increasingly report performance only within their own ecosystems.
- Customer journeys fragmented across more channels than any single tracking system was built to cover.
None of those four pressures is temporary. None of them is waiting on a technical fix that’s still coming. This is the environment now.
The deeper issue isn’t just that less data is available — it’s that attribution was never actually measuring what most marketers assumed it was measuring. It answers “Which touchpoints did the people who converted happen to encounter?” That’s a correlation question. It was always being treated as if it answered a causal one — “Which touchpoints made those people convert?” — and for years, the gap between those two questions was small enough to ignore.
The distinction is easier to see outside marketing. Imagine someone gets a headache, takes an aspirin, then eats a sandwich, then feels better. Ask “what happened before they felt better?” and the honest answer is: the aspirin and the sandwich. Only one of those actually did the work — but a system that just logs sequence, rather than testing cause, has no way of knowing which. Attribution has always worked the same way. It’s very good at logging the aspirin and the sandwich. It was never actually built to tell you which one mattered.
Signal loss didn’t create that gap between correlation and causation — it was always there. What signal loss did was shrink how much of the journey marketers could even see, making a gap that used to be small enough to ignore too large to keep pretending it wasn’t there.
Where Marketing Mix Modelling Fits
Marketing mix modelling (MMM) takes a different approach entirely, and it’s worth understanding why that difference matters rather than treating MMM as simply “the old thing we used before digital tracking existed.” Instead of trying to track individual users across their journey, MMM works at an aggregate level — it looks at spend, revenue, pricing, seasonality, and external factors over time, and statistically estimates how much each input actually contributed to the outcome. No cookies. No device IDs. No user-level identity to lose in the first place.
That’s not a workaround for privacy restrictions — it’s a fundamentally different question, answered at a different altitude. Attribution tries to trace a path. MMM tries to isolate contribution. In a world where the path is increasingly untraceable, contribution becomes the more honest thing to measure.
Legacy attribution methods have also long struggled to quantify offline and brand-building efforts, tending to overweight bottom-of-funnel tactics because those are simply easier to track. That structural blind spot, combined with regulatory pressure on third-party tracking, has accelerated the shift toward MMM well beyond privacy-driven necessity — it’s increasingly seen as the more complete picture, not just the fallback option.
Why This Is a Resurgence, Not Just Nostalgia
MMM isn’t new. Consumer goods companies have used versions of it since the 1960s, long before digital attribution existed at all. What’s changed is why it’s suddenly getting board-level attention rather than being treated as the measurement method teams fall back on when nothing else works. A Gartner survey found that 64% of senior marketing leaders have now adopted MMM solutions — and among B2C leaders with high MMM utilisation, marketers were twice as likely to successfully prove their value and receive internal credit for it, compared with those not using it at all.
Several forces are driving that adoption. The cost of entry has collapsed: Google open-sourced Meridian, its Bayesian MMM framework, and paired it with a network of trained measurement partners. Meta did the same with Robyn. Work that used to require a six-figure annual consulting engagement is now something a capable in-house analytics team can actually run.
GenAI is accelerating that shift further. Its integration into MMM platforms is improving insight generation and simplifying the identification of optimal spend scenarios, with AI-driven analysis helping uncover performance drivers hidden across fragmented data views—patterns a human analyst would otherwise have to hunt for manually in disconnected reports.
Cloud-native platforms can now refresh models on something closer to a weekly cadence instead of the quarterly reports MMM used to be known for, and machine learning layered on top of classical regression is starting to catch effects that older models simply flattened out — diminishing returns on a channel, saturation points, creative fatigue setting in before spend gets reallocated.
There’s also a compliance angle that’s arriving faster than most marketing teams have priced in. MMM’s aggregate, non-personal approach sidesteps much of the regulatory exposure that identity-based attribution now carries, at a time when scrutiny of data brokers and ad tech is intensifying. That’s a risk argument a CFO tends to understand immediately, often faster than the measurement argument lands with a marketing team.
The Five Ways Enterprise Players Use MMM
Gartner’s research identifies five primary use cases for MMM solutions, each suited to a different organisational starting point:
- Basic mix modelling: Best for organisations new to MMM, emphasising data management, model latency and adoption enablement for marketers.
- Enterprise mix modelling: Focused on cross-functional adoption, business scenario planning and complex analytics — crucial for estimating the full range of factors affecting ROI.
- Big-budget advertising mix models: Prioritises media optimisation and complex analytics, tailored for advertisers managing substantial media budgets.
- House of brands: Designed for organisations standardising and scaling MMM across multiple brands, with an emphasis on data management and media optimisation.
- Self-service mix models: Built for organisations that want granular control over model specifications, centred on data scientist adoption and complex analytics.
Choosing an MMM Vendor
For organisations evaluating an MMM solution, a few steps make a meaningful difference to how the rollout goes:
- Engage stakeholders across marketing, finance, data management, supply chain and executive leadership early to document data and output requirements and secure enterprise-wide buy-in before signing a contract.
- Begin collecting and auditing at least two years of daily marketing and business conversion data before committing to a vendor — data readiness, not vendor selection, is usually what determines how long a rollout actually takes.
- Evaluate vendors on their capabilities, industry experience and ability to answer the specific organisational questions being asked, rather than on feature lists alone.
- Assess not just current functionality but how each vendor’s roadmap incorporates emerging trends and modelling techniques, including genAI.
The Honest Caveat
MMM is not a replacement for attribution, and treating it as a silver bullet is exactly the kind of overclaim that makes measurement vendors hard to trust. It answers a different question at a different level of granularity and on a different timescale. It won’t tell you which specific ad a specific customer clicked before they bought. It’s not built to.
The measurement programmes that are actually holding up in 2026 aren’t choosing one method — they’re triangulating three: MMM for the portfolio-level view of what’s working, incrementality testing for causal ground truth on specific questions, and platform attribution for tactical, in-flight signals, despite its known limitations. Treated as competitors, these three methods argue with each other. Treated as a layered system, they cover each other’s blind spots.
The Part Nobody Wants to Hear: Fix the Plumbing First
None of this works if the underlying data is a mess. MMM is only as good as the spend and revenue data feeding it, and a lot of organisations trying to stand up a model right now are discovering that their CRM doesn’t reconcile with their ad platform data, their finance systems report revenue on a different cadence than marketing reports spend, and nobody has actually owned the job of reconciling the two. That’s not a modelling problem. It’s an infrastructure problem, and it has to get solved before the model — any model — can be trusted.
The teams treating measurement as a data infrastructure project first, and a modelling project second, are the ones actually getting reliable numbers out the other end. Everyone else is building an impressive-looking model on top of numbers nobody fully trusts, which is a more sophisticated version of the exact problem they were trying to fix.
With genAI integration accelerating and cross-functional collaboration becoming the norm rather than the exception, MMM is on track to become an essential part of the marketing measurement stack — not a replacement for attribution, but the layer that finally tells marketers, and the finance teams they answer to, what their spend actually accomplished.
More articles
Push Notifications Have Become a Retention Risk 08 Apr 2026
Why the Best Product Doesn’t Always Win 31 Mar 2026
