Nobody believes their attribution anymore

The dashboard still says ROAS 4.2. The leaders who run it tell us they've stopped believing the number. In interview after interview, budgets are leaving click-path attribution for mix models, holdout tests, and clean rooms — measurement that's slower and messier, but believed.

The crisis isn't that attribution tools broke. It's that the people paying for them stopped trusting the output. A telecom marketing director wants attribution "for real-time data analysis" and can't get it. A global consumer-goods giant lists "real-time, multivariable attribution" as a top priority — still unsolved. A real-estate marketplace describes outright "attribution challenges." A used-book retailer is "looking again" for an MTA/MMM vendor, emphasis on again. The complaint is everywhere in the corpus, and it has a consistent shape: the click-path model was precise, real-time, and increasingly fictional.

Then — the click-path story
Impression
Click
Site visit
Conversion
ROAS 4.2 ✓precise · real-time · increasingly fictional
Now — triangulation
MMM where the next dollar goes, modeled across all channels
Holdout test did the channel actually cause anything?
Clean room identity-matched truth, where scale allows
Directional — and believedslower · messier · survives the CFO meeting

The trade buyers describe making: false precision out, defensible direction in.

6.7 → 8.0
avg buyer rating, archetypal MTA platform (Rockerbox) vs. top incrementality platform (Haus)
1 in 2
buyers shopping for an MMM are simultaneously in-market for attribution or incrementality tools
$1M–$100B+
revenue span of companies rethinking measurement — DTC brands to global CPG and payments

The measurement stack reshuffle

Four approaches, four very different trajectories. Ratings are averages from named buyers actively using each tool.

Trajectories and ratings from the Alium interview corpus through July 2026. Ratings are 1–10 averages from named buyers.
ApproachWhat buyers are doingTools named
Multi-touch attributionMTALosing trust Questioning accuracy, citing offline blind spots and post-cookie decay; several describe exits in progress. The exception: Shopify-scale DTC, where click-based tools still score well. Rockerbox 6.7
Northbeam 7.7
Triple Whale 7.8
Marketing mix modelingMMMBudgets moving in The most active shopping category in measurement: mid-market brands running bake-offs, enterprises modernizing or building in-house, everyone asking for faster model refreshes. Recast not rated
Haus 8.0
Measured 7.4
Incrementality testingholdouts & liftThe credibility test The proof layer buyers bolt onto everything else — "did this channel cause anything?" Highest-rated category in measurement, but buyers admit it demands a testing culture they're still building. Haus 8.0
Measured 7.4
Data clean roomsidentity-basedEnterprise build-out Retail, hospitality, and media giants building clean-room stacks on identity infrastructure — for attribution, CTV measurement, and retail-media partnerships. LiveRamp 7.4
Acxiom

The MTA exits

Trust collapseWhen the click path stops describing reality

Buyers leaving multi-touch attribution rarely cite a missing feature. They cite disbelief: numbers that change between pulls, channels the model can't see, and a creeping sense that fractional credit assignment is theater once the journey crosses retail, TV, and long consideration windows. Rockerbox — the archetypal enterprise MTA platform — averages 6.7/10 in our corpus, the lowest of the major measurement vendors, with data accuracy and timeliness the leading complaints. GA4 catches its own share of this frustration in the interviews; that's a story we cover separately.

A heritage watch brand is dropping Rockerbox after concluding it's "better suited for startups focused on performance marketing" — the wrong instrument for a business that sells through department stores and Amazon as much as its own site. The team is exploring Recast and Haus instead, while admitting the prerequisite is cultural: they need to build a testing discipline before the new tools can pay off. Director, Marketing · heritage watch brand · Oct 2025
A wine subscription club describes itself as "highly satisfied" with Recast after years of struggling to trust Rockerbox — the rare interview where a buyer names both the tool they stopped believing and the one that restored confidence. CMO · wine subscription club · 2024–2026 interviews
A Canadian department-store chain told us it needs an MMM and is still working out its measurement strategy ahead of cookie deprecation — shopping for the replacement before the old instrument fully fails. Marketing leader · Canadian department-store chain · prior interviews

MMM and holdouts become the center of gravity

Most momentumThe triangulation stack

Marketing mix modeling — a technique older than the click — is the most actively shopped measurement category in recent interviews, and it almost never travels alone. Roughly one in two MMM shoppers in our corpus is simultaneously in-market for attribution or incrementality tooling: one trust problem, three categories. The bake-offs repeat the same names — Recast vs. Measured at a jewelry retailer, Haus on the shortlist at an online personal-styling company replacing its legacy MMM provider, Northbeam vs. Prescient at DTC brands. At the top end, a global payments network is pushing AI to make its mix models refresh faster, and a major asset manager runs MMM entirely in-house.

An online plant retailer runs Haus for incrementality testing and rates the science highly — while flagging the category's emerging pain: pricing that's hard to justify in the off-season, on an annual contract that can't be paused. Director, Marketing · online plant retailer · May 2026
A chef-founded food brand is hunting for a new incrementality partner after discontinuing its last one — its stated priority is bridging "the attribution gap" between e-commerce and the retail shelf, exactly the gap click-path tools can't see. VP, Marketing · chef-founded food brand · Nov 2025
A children's apparel brand is standing up a data warehouse and incrementality testing in the same motion — centralizing reporting first so the tests have something trustworthy to read from. Director, Marketing · children's apparel brand · Apr 2025

At the enterprise end: clean rooms and identity

Build-outMeasurement as infrastructure

Where there's scale and first-party data, buyers aren't shopping for a dashboard — they're building plumbing. Retailers, hotel groups, and media companies describe clean-room stacks for attribution, CTV measurement, and retail-media partnerships, almost always anchored on identity infrastructure. LiveRamp (7.4/10 average across one of the largest buyer pools in our corpus) is the recurring backbone; the recurring complaint is cost — "cost prohibitive" in one buyer's words. A major automaker is running a year-long MTA test through LiveRamp rather than buying a standalone attribution tool; a warehouse-club retailer is evaluating a clean room for member data; a global money-transfer company lists improving attribution among the drivers of its current platform RFP.

A multi-brand kitchenware retail group builds MTA and MMM in-house, invests in clean rooms with LiveRamp — which it rates a 9 — and is exploring Acxiom for identity. The pattern: at this scale, measurement is an engineering program, not a SaaS purchase. VP, Marketing · multi-brand kitchenware retail group · 2025 interviews
The same group's stated priority a year earlier: "utilize data clean room solutions to improve attribution" while expanding CTV beyond Meta — clean rooms positioned explicitly as the attribution fix. VP, Marketing · multi-brand kitchenware retail group · Feb 2025

The counter-current: the new stack has its own backlash — and MTA isn't dead everywhere

Two honest caveats. First, where the journey is genuinely digital and short, click-based attribution still earns real affection: Triple Whale averages 7.8/10 across one of the biggest buyer pools in our corpus, praised for its Shopify integration — a cannabis multi-state operator rates it highly for exactly that. The two DTC tools go head-to-head in Triple Whale vs. Northbeam — what buyers say. Second, the replacement stack already has a cost problem. Buyers call Haus and Measured expensive and hard to justify off-peak; one brand discontinued its incrementality partner outright; an early Recast adopter flags wide error margins. And the watchmaker's admission generalizes: MMM and holdouts only pay off for teams with a testing culture — which many interviews concede is still under construction.

What this means for your measurement evaluation

If you're mid-decision on measurement, the interviews suggest three checks before you sign anything. Match the instrument to your journey: if most of your revenue converts on-site within days, the Shopify-scale MTA tools still score well — the trust collapse is concentrated in omnichannel, offline-heavy, long-cycle businesses. Buy the culture before the tool: the buyers happiest with MMM and incrementality platforms had a testing discipline first; the ones who didn't describe expensive shelfware on annual contracts they can't pause. Price the triangulation, not the tool: one in two MMM shoppers is also shopping attribution or incrementality — budget for the stack you'll actually end up running, and ask every vendor how their numbers reconcile when (not if) they disagree with the others. For the hard-won lessons from buyers who've already picked, see what attribution-tool buyers wish they'd known; for buying marketing-mix modeling specifically — the three buy-paths and the months-long build — what MMM buyers wish they'd known; and for the complaints graded and mapped to which tool earns which, top complaints about attribution & measurement tools.

Common questions

Why are marketers moving away from multi-touch attribution?

Not because the dashboards broke — because nobody trusts them. Buyers describe accuracy doubts, numbers that shift between pulls, offline and upper-funnel blind spots, and click-path models degrading as cookies disappear. Rockerbox, the archetypal enterprise MTA platform, averages 6.7/10 in our corpus — lowest of the major measurement vendors.

What are buyers replacing MTA with?

A triangulation stack: MMM for budget allocation, incrementality and holdout testing for causal proof, and — at the enterprise end — clean rooms on identity infrastructure like LiveRamp. Roughly one in two MMM shoppers in our interviews is simultaneously in-market for attribution or incrementality tooling.

Which measurement vendors do buyers rate highest?

Through July 2026: Haus leads at 8.0/10 — an MMM/incrementality platform — followed by Triple Whale 7.8, Northbeam 7.7, LiveRamp and Measured 7.4, and Rockerbox 6.7. Recast draws similarly strong reactions to Haus, but from too few buyers for us to publish a number. The testing-and-modeling tools outscore the click-path ones.

Is multi-touch attribution dead?

Not at Shopify scale, where journeys are short and digital — Triple Whale's 7.8 average and praise for its Shopify integration show click-based tools still working there. The collapse is concentrated where journeys cross retail, TV, and long consideration windows. And the new stack has its own backlash: cost, annual contracts, and a testing-culture prerequisite many teams admit they haven't built yet.

This is the aggregate. Your stack is specific.

Do a 15-minute interview about your own measurement stack and get this analysis personalized — what peers rate the tools you run, who's leaving your attribution platform, and what they're triangulating with instead.

Get my personalized brief

No password needed · your interview is anonymized before it ever informs a page like this one.

Methodology. Alium conducts verified interviews with software buyers — the marketing, ecommerce, data, and IT leaders who select and operate these platforms. This page aggregates the measurement-related interviews in that corpus — attribution, MMM, incrementality testing, and data clean rooms — conducted through July 2026. Buyer identities are verified at interview time and anonymized before publication; vendor names and ratings are reported as given. No vendor paid to appear or was able to edit this page.