Eight in ten advertisers already run marketing mix models and brand lift studies. Only 15% say effectiveness evidence is the primary driver of how budgets actually get set (Ebiquity/WFA, 2026). The measurement exists. It just isn’t deciding anything.

That gap is where the attribution debate goes wrong. The argument in the room is always about the model. Multi-touch or mixed-media, incrementality tests or self-reported attribution, which vendor and which weighting window. It’s a real debate, and it’s the second one to have.

We audit enterprise paid media accounts for a living, and the failure is almost never the model. It’s the signal going into it. Attribution breaks upstream of the model, in the instrumentation, and the dashboard renders a confident answer on top of the break without ever flagging it.

The reports come out clean. That is the part that makes this hard to see.

Why is B2B marketing attribution broken?

Only 15% of advertisers say effectiveness evidence is the primary driver of how budgets get set, even though eight in ten already run marketing mix models and brand lift studies (Ebiquity/WFA, 2026). The tooling is not the constraint. Attribution breaks upstream of the model, in the instrumentation feeding it.

We audited two accounts in the same quarter with the same symptom. Reports looked fine. Business results didn’t follow. In the first, an outside agency was running strategy and execution, and their reporting showed steady performance while CPA targets stayed out of reach no matter how much spend went in. Under the hood, 10% of Meta budget was pointed at an audience segment with no relationship to the buyer. Google Search delivered zero conversions across an entire quarter, and the investment continued the whole time. Campaign structure was fragmented badly enough that budget was trapped in underperformers while the top ad sets ran underfunded.

Nothing in that account was a modeling problem. Every number in the report was accurate. The report was measuring a system that was pointed in the wrong direction, and measuring it correctly.

This is the failure mode that survives a model change. You can move from last-touch to multi-touch, buy an incrementality vendor, rebuild the weighting, and inherit the same broken feed. You end up with a more sophisticated model producing the same wrong answer.

What sits underneath is remarkably consistent from account to account.

What do enterprise marketing audits actually find?

In one account we audited, 98% of paid search budget was going to brand keywords. The company was paying to recapture traffic it already owned, and ROAS looked strong the whole time, because demand you already own converts at a far higher rate. The failures audits surface are almost always mechanical.

In the second account we audited that quarter, ROAS was healthy and revenue growth had stalled, and nobody could say why. Performance Max wasn’t negating brand terms, so most of what the platform reported as new customer conversions were existing customers coming back. Frequency against existing audiences had climbed to 37x, which pushed CPMs up and made the media look more expensive without making it more effective. Tracking parameters were inconsistent across platforms, so cross-channel analysis was unreliable in a way no dashboard announced.

We have seen the same pattern across three accounts in different verticals, run by three different agencies of record. In one, 69% of the new customer purchases that Performance Max claimed had actually come from brand search. The campaigns were built to prospect. In practice they were harvesting demand the brand had already created.

The pattern generalizes past search. A direct-to-consumer brand that scaled globally ended up live in 189 countries, with 10% of budget spread across markets producing zero conversions while its top markets ran underfunded. Global looked like progress on the slide. The budget was following reach, not conversion signal.

Here is what that looks like side by side.

None of these accounts were run by careless people. Competent teams, real agencies, accurate reports. The instrumentation was never audited, so the reporting layer kept doing its job on top of a foundation nobody had checked.

The confidence this produces is measurable. 84% of global marketers say they are extremely or very confident in their ROI measurement capabilities, up from 69% the year before. Only 38% say they evaluate ROI across traditional and digital together (Nielsen, 2024). Confidence is rising faster than practice.

A good share of the distance between those two numbers gets filed under a name that has become very convenient.

What is the dark funnel and how do you measure it?

The dark funnel is real. Buying committees research in Slack groups, podcasts, and private communities you cannot instrument, and that demand shows up later as direct traffic with no history. It has also become an alibi. Before you file a gap under dark social, rule out the tracking you broke yourself.

In practice the two are easy to confuse and expensive to mix up. When tracking parameters are inconsistent across platforms, a meaningful share of attributable demand arrives looking exactly like dark social: unattributed, unexplained, filed under “brand awareness working.” That demand is sitting in your data under the wrong label.

The honest version of dark funnel measurement starts with self-reported attribution, a “how did you hear about us” field that a human reads, and pairs it with a clean baseline of what you can actually track. You will still have a gap. The gap will be smaller than you think, and you will know which part of it is genuinely unmeasurable rather than merely unmeasured.

The distinction matters because one of these is a strategy problem and the other is a maintenance problem, and only one of them gets fixed by talking about it in a QBR.

The same confusion runs one level up, in the metric most teams still lead with.

Is the MQL dead? What should B2B teams measure instead?

The MQL is not dead, it is demoted from headline metric to diagnostic. A form fill tells you someone read something. It does not tell you a buying committee is moving, and those two stopped correlating years ago. Lead with pipeline created and cost per qualified opportunity instead.

We made the full case for this separately, including what high-performing teams put in its place, in Is the MQL Dead? What High-Performing B2B Teams Measure Instead. The short version for the purposes of this argument: the MQL is a symptom of the same disease. It is a metric that is easy to instrument and easy to hit, which is exactly why it survived long after it stopped predicting revenue.

Swapping the metric without fixing the signal underneath just moves the problem. If your conversion definitions are wrong, “pipeline created” will be wrong too, and it will be wrong with more authority.

That trap sits inside the most popular fix in B2B marketing.

What is revenue marketing, and does renaming the function fix the measurement problem?

Revenue marketing reorganizes marketing around pipeline and revenue rather than leads and MQLs. It is the right target. It is also, on its own, a rename. Only 13% of organizations rate themselves strong on speed from data to insight (Ebiquity/WFA, 2026). A new title on the dashboard does not fix the feed underneath it.

The teams we see get real value from the shift are the ones who treated it as an instrumentation project first and an org chart second. They rebuilt conversion definitions so that the events being optimized toward were the events the business actually cared about. They negated brand terms so prospecting campaigns had to go find strangers. They excluded existing customers from acquisition audiences. Then they renamed the function.

Do it in the other order and you get a revenue marketing team reporting revenue numbers generated by a measurement system nobody has audited. That is where a lot of teams sit, and it is why 60% to 75% of buy-side leaders say advanced measurement falls short on rigor, timeliness, trust, and efficiency (IAB, 2026).

The question worth answering is which numbers survive contact with an audit.

What should you measure instead?

Measure the things that survive an audit. Pipeline created, cost per qualified opportunity, and the share of spend landing on demand you already own. Then verify the signal underneath them: conversion definitions, tracking parameters, brand-term negation, audience exclusions. 46% of organizations sit at the lowest maturity level for integrating data sources (Ebiquity/WFA, 2026).

Verifying the signal underneath the metrics is the part that matters, and it is the least glamorous work in marketing. Nobody gets promoted for auditing a conversion definition. The teams that do it are the ones whose numbers still hold up when the CFO pushes on them, and increasingly, the ones whose automated bidding is not optimizing toward the wrong buyer. We wrote about what happens when it does in Bad Conversion Data Doesn’t Just Mislead You. It Trains the Algorithm Against You and AI Is Making Media Decisions on Data You Haven’t Checked in Years. The stakes went up when the machines started acting on the signal without a human in between.

Rebuilding an instrumentation layer is the unglamorous half of what Moving Parade’s Foundation work actually is. The deliverable is a verified feed underneath the dashboard you already have.

One move: Pull the last 90 days of paid search spend and calculate what share landed on your own brand terms. If it is north of 50%, your prospecting campaign is harvesting demand you already created, and your efficiency metrics are telling you a story that isn’t true.

Frequently asked questions

Why is B2B attribution harder than B2C attribution?

Buying committees. Six to fifteen people research independently, most never identify themselves, and the purchase closes months after the touch that mattered. B2C attribution tracks a person. B2B attribution tries to track a consensus, which no model reliably resolves into fractional credit.

Is multi-touch attribution worth building?

Only after the signal underneath it is verified. A multi-touch model built on mislabeled conversion events produces confident, precise, wrong answers. Audit conversion definitions, tracking parameters, and audience exclusions first. Then decide whether you still need the model.

How do you know if your attribution is broken?

The tell is disagreement between the report and the business. Reports show steady performance while CPA targets stay out of reach, or ROAS looks strong while revenue growth stalls. When the dashboard and the P&L tell different stories, the dashboard is usually the one that’s wrong.

What is the first thing to audit?

The share of paid search spend going to brand terms, and whether Performance Max is negating them. It is the fastest way to find out whether your prospecting is actually prospecting or harvesting demand you already own.

Should marketing and finance use the same numbers?

Yes, and the number that survives that conversation is pipeline created, not leads. Fewer than 3% of advertisers are fully confident they can separate short-term performance from long-term brand building (Ebiquity/WFA, 2026), which is precisely the conversation finance keeps asking marketing to have.