Ask buyers who own an attribution tool what they'd tell their past self and the theme isn't features — it's trust. Whether the number holds up in front of the CFO, whether anyone can explain how it was calculated, whether the data underneath was ever clean. The methods differ; the regrets converge on believability. Here they are, in the order buyers hit them — for the market shift behind them, see the flagship read on why nobody believes their attribution anymore, and for a DTC head-to-head, Triple Whale vs. Northbeam.
The seven lessons
Multi-touch numbers stopped being believed — that's the whole crisis
The starting point for every other lesson. Privacy changes — iOS signal loss, cookie deprecation — broke the tracking that click-path multi-touch attribution depends on, and buyers describe the fallout bluntly. One apparel brand deemed its MTA numbers unreliable to the point of excluding them from its weekly business review. Another large buyer doubted the accuracy of the results outright and felt under-supported for its size. MTA still has a role for directional, in-platform optimization, but the buyers in the corpus increasingly treat it as one input, not the source of truth. If you're buying a last-touch dashboard expecting a defensible number, you'll be disappointed.
Named in this context: Rockerbox · Triple Whale · Northbeam (MTA & dashboards)
The real test is whether finance believes the number
The clearest divide between the tools buyers love and the ones they tolerate. Buyers rate the incrementality leader highly for a specific reason: its data is clear and persuasive and well-understood by finance teams, to the point of driving incremental dollar investment. The winning measurement isn't the prettiest dashboard or the most granular touchpoint map — it's the number that survives a conversation with the CFO and actually moves budget. When you evaluate a tool, bring finance into the room early and ask the real question: will this hold up when we defend the media plan?
Named in this context: Haus · Recast
A black box you can't explain is a number you can't defend
The same tool praised for rigor is, by other buyers, called a "black box" — its calculations opaque, leaving users skeptical of the results in a space where trust is the entire product. In attribution, methodology is the value: if you can't see how the model reaches its answer, what it assumes, and how the holdout was constructed, you can't stand behind the output when leadership pushes back. Recast earns its reputation partly on transparency for exactly this reason. Every account of a model here is of one already running under contract, and the skepticism is retrospective: no buyer describes auditing the methodology before signing. So the next part is our recommendation rather than theirs: before you sign, make the vendor walk you through the model end to end — and treat "it's proprietary" as a yellow flag, not a reassurance.
Named in this context: Haus · Recast (the transparency counter-example)
The tool can't fix your data — messy inputs sink attribution first
Onboarding pain in the interviews concentrates in one place: messy internal data and slow integration. Attribution and incrementality models amplify whatever you feed them — clean tracking produces a trustworthy read, and broken tagging produces a confident, wrong one. Buyers who rushed a tool onto a shaky data foundation describe cumbersome setups and results they couldn't act on; the ones who did it in the right order fixed tracking, tagging, and data pipes first, then bought the model. The tool is the last mile of measurement, not the cleanup crew.
Named in this context: Haus · Rockerbox
Match the tool to your spend and cadence — incrementality has a minimum viable budget
The leading incrementality tools are rigorous and well-supported — but several buyers on seasonal or sporadic media budgets describe the annual contracts as a poor fit, paying through off-peak periods when they aren't running enough tests to justify the cost. Geo-lift and incrementality methods also need a minimum viable media budget to produce a readable signal; below that threshold, the tests can't conclude and you're paying for measurement you can't use. Price the tool against how many conclusive tests you'll actually run in a year, not the sticker — and if your spend is lumpy, negotiate the cadence, or you'll resent the contract in Q1.
Named in this context: Haus
No single tool is the source of truth — you'll triangulate
The mature buyers stopped hunting for one number. They describe fractured measurement across multiple tools without a single source of truth not as a failure state but as the reality of the category — running multi-touch for in-platform optimization, incrementality to validate what actually caused sales, and marketing-mix modeling for the cross-channel and offline picture. The tools even have channel blind spots that force the stack: buyers note some platforms skew heavily toward paid and aren't built for email or SMS marketers. Budget for a measurement stack where incrementality is the referee, not for a silver bullet that doesn't exist.
Named in this context: Haus · Rockerbox · Northbeam · Triple Whale
Historical data is a switching cost — leaving resets your baseline
One of the quieter traps. A jewelry buyer preferred a competitor and had effectively outgrown its current tool — but stayed, specifically to preserve a year of accumulated historical data. Marketing-mix models and attribution baselines are built over time; the model gets smarter the longer it runs, so switching tools restarts the clock and leaves you with a thin baseline right when you need confidence. It's a real cost, and vendors know it. When you migrate, factor in the lost history and the ramp to a trustworthy read — and if you're early, weigh locking in the tool you'll want to keep before the switching cost compounds.
Named in this context: Rockerbox · Northbeam
How buyers talk about the tools
The landscape sorts by trust, not features. Haus sits at the top on the numbers — buyers believe them and finance does too. Recast earns the same kind of confidence, from too few buyers for us to publish a rating. The knocks are cost-to-spend fit, onboarding on messy data, and, for one, "black box" opacity. Rockerbox is the capable multi-touch incumbent buyers increasingly doubt: praised for its interface and support, flagged for data-accuracy concerns, complexity, and a paid-channel bias. Triple Whale and Northbeam are the DTC dashboards — fast and popular, sharing the same data-trust question that hangs over the whole category. The pattern is consistent: the methods that measure lift earn trust; the ones that assign credit are losing it.
The stories behind the lessons
The counter-current: the tool is rarely the whole problem
The buyers who trust their measurement look different before the contract, not after it. They cleaned their tracking and tagging first; they picked methods that measure lift and validated the dashboards against holdouts; they demanded transparency into the model and brought finance in to pressure-test it. The buyers still fighting their numbers — dirty inputs, a black-box they can't defend, one tool asked to be the whole truth — tend to carry those problems to the next vendor. The lessons above are cheap insurance; re-instrumenting measurement after you've lost leadership's trust is not.
How buyers rate the tools
| Tool | Buyer rating | What buyers say |
|---|---|---|
| Haus | 8.0 | Top of the numbers here. Buyers believe the results, and finance does too. |
| Triple Whale | 7.8 | A DTC dashboard — fast and popular, and carrying the same data-trust question that hangs over the category. |
| Northbeam | 7.7 | The other DTC dashboard, with the same profile: quick to read, same trust question. |
| Rockerbox | 6.7 | The capable multi-touch incumbent buyers increasingly doubt — praised for interface and support, flagged for data-accuracy concerns, complexity and a paid-channel bias. |
| Recast | not rated | Too few buyers rated it to publish a number. Earns the same kind of confidence as Haus from the buyers who do use it. |
Want this read against your own stack?
Get my read →What this means for your evaluation
Four checks before you sign. First, decide what question you're actually answering — credit (MTA), causal lift (incrementality/geo), or contribution (MMM) — and don't buy one expecting the others. Second, bring finance into the evaluation and ask whether the output will survive a budget defense; if it won't, the rigor is wasted. Third, make the vendor walk you through the model — assumptions, holdout construction, and where the numbers come from — and treat opacity as a risk, not a moat. Fourth, audit your own tracking and data before you onboard, and price the tool against the conclusive tests you'll really run. For the market shift behind these lessons, see why nobody believes their attribution anymore; for buying marketing-mix modeling specifically — buy-vs-build, the legacy-vs-modern split, and the months-long ramp — what MMM buyers wish they'd known; for the same complaints graded and mapped to which tool earns which, top complaints about attribution & measurement tools; and for the DTC dashboards, Triple Whale vs. Northbeam.
Common questions
Is multi-touch attribution (MTA) still worth it?
For most buyers in Alium's corpus, less than it used to be. Post-privacy signal loss — iOS changes, cookie deprecation — has eroded what click-path multi-touch attribution can actually see, and buyers describe the numbers accordingly: one brand pulled its MTA figures out of its weekly business review because they were unreliable, and others doubt the results outright. MTA still has a place for directional, in-platform optimization, but buyers increasingly treat it as one input, not the source of truth. The trust and the budget are moving toward incrementality testing, geo-lift, and marketing-mix modeling — methods that measure lift rather than assign credit to touchpoints. If you're buying MTA expecting a defensible number, the interviews suggest you'll be disappointed.
What's the difference between MTA, incrementality testing, and MMM?
They answer different questions, and buyers increasingly run all three. Multi-touch attribution (MTA) assigns credit across the touchpoints in a converting journey — useful but degraded by signal loss and prone to over-crediting trackable channels. Incrementality testing (including geo-lift) measures causal lift by holding out or varying spend and comparing outcomes — the method buyers trust most for did this spend actually cause sales? Marketing-mix modeling (MMM) uses statistical models over historical spend and outcomes to estimate each channel's contribution, strong for cross-channel and offline but hungry for data history and time. The mature setup buyers describe isn't one of these; it's a stack where incrementality validates what the dashboards and models claim.
Why don't marketers trust their attribution numbers?
Three reasons recur in buyer interviews. First, signal loss: privacy changes broke the tracking that click-path attribution depends on, so the numbers no longer reconcile with reality — one brand excluded them from its business review entirely. Second, opacity: several buyers describe their measurement tool as a 'black box' whose calculations they can't see or explain, which makes the output impossible to defend to leadership. Third, dirty inputs: attribution models amplify whatever data you feed them, and messy internal tracking and tagging produce confident but wrong answers. The buyers who regained trust did it by validating with holdout and geo-lift tests, demanding transparency into the model, and fixing their data foundation first.
How much does an incrementality testing tool cost?
Enough that it only makes sense above a certain spend and cadence. Buyers praise the leading incrementality tools for rigor and support, but several on seasonal or sporadic media budgets describe the annual contracts as a poor fit — they pay through off-peak periods when they aren't running enough tests to justify the cost. Geo-lift and incrementality methods also need a minimum viable media budget to produce a readable signal; below that threshold, the tests can't conclude and you're paying for measurement you can't act on. The buyers who got value matched the tool to their spend level and testing cadence, rather than buying the rigor and discovering they couldn't feed it. Price it against how many conclusive tests you'll actually run in a year.
Get this research made for your stack
Choosing an attribution or measurement tool right now? Do a 15-minute interview about your own spend and stack and get this personalized — what peers with your channel mix chose, where each tool's rough edges are, and which of these lessons apply to your shortlist.
Get my personalized read — 15-min interviewNo password · your interview is anonymized before it ever informs a page like this one.