A CMO we recently onboarded, shared a $2M in pipeline sitting in her data. Invisible in every report her team had run for three years, promptly buried across four disconnected systems.

She asked me: how long has this been here?

The honest answer was: since before we started working together.

This is not an unusual story. A September 2025 survey of 200 CMOs by Adverity found that on average, 45% of the data marketing teams use to drive decisions is incomplete, inaccurate, or outdated. While, almost half the map is wrong, most marketing decisions are being made from it. Even the CMOs making those decisions, don't really trust them.

The tools exist and the data exists - what is still missing in most B2B companies is the infrastructure to connect them.

Here is what I have come to believe after watching this pattern repeat: the technical challenge of attribution is not the model. It is data quality. And data quality is a people problem, not a technology problem.

Yet 76% say less than half of their CRM data is accurate and complete. (Validity, July 2025)

Organisations losing revenue as a direct consequence of poor CRM data quality. (PR Newswire)

These are not small companies with informal operations. These are companies that have invested heavily in their stacks and still cannot see what is happening in their own pipeline.

And it shows up in the first week. Working with so many GTM teams, I can predict it now before we get there.

We recently started working with a DevSecOps company and three weeks into the process, the Head of Marketing Ops said it with the flat tone of someone who had stopped being surprised by bad news.

She was right, she had campaigns labeled by channel, some by region, some by quarter. Some of these campaigns were set up by people who had left years ago and a significant chunk labeled "other."

A company that had grown fast, hired well, and accumulated four years of marketing history that was never designed to be interrogated.

So, we stopped the implementation because one cannot build on a broken foundation.

That month of cleanup is the month that kills most attribution projects, not because it is technically hard, but because it is politically difficult. It requires admitting that the data you have been reporting on for years is less reliable than everyone assumed.

Most teams do not make it through the clean up, to show ROI they start building on the dirty data. The model produces numbers that look authoritative and mean nothing. Six months later, nobody trusts the output. The project dies quietly.

The challenge is not finding a better model. It is fixing what the model is reading.

Campaigns are generally organised by quarter, product line, team, which makes sense when you are running them. But it means nothing analytically. For instance, if you cannot extract "webinar" as a clean category across three years of data because some are labeled "Q3 FY23 Webinar" and others are "content series April," you cannot answer whether webinars drive pipeline.

For more than 50% of senior marketing leaders, using data and analytics to tackle their most important problems has become their second biggest challenge. Buyers now come in through different touch points and take longer to make decisions. Traditional funnels do not reflect how people buy anymore. The taxonomy problem sits underneath all of it.

A website is not a channel, it is a destination. When you treat it as a channel, you lose the ability to see what drove the visit.

A sales rep meets a prospect at a conference and after three follow up conversations, an opportunity created in the CRM with no marketing source attached.

Marketing gets zero attribution credit. But the thing is that the prospect had spent eight months reading content, attending webinars, consuming everything the company produced. None of it was connected to the opportunity because it was not created through a marketing form fill.

In one customer's case, sales-generated pipeline with invisible marketing influence represented 30% to 40% of total pipeline. Marketing had been systematically under-credited for years and more than this being a communication problem. This is a data visibility flaw, the evidence for marketing's involvement was missing.

At a compliance software company, the Head of Growth Marketing told me: "Your last touch is different than our last touch."

He was right. Their attribution logic required two layers to be effective:

The final touchpoint, such as clicking an email, attending a webinar, or visiting a pricing page - that immediately preceded the moment a lead was created or a form was filled.

The traffic or channel source (e.g., paid search, organic, or a specific email campaign) that originally drove the prospect to the platform before that final interaction occurred.

When you build a model on top of that ambiguity, you produce numbers both teams find plausible and neither fully trusts. Every attribution meeting becomes a definitional debate. Everyone is correct, but nobody decides anything.

This is where RevSure comes in. It can help with the infrastructure that enforces these standards at the point of entry and unifies what exists across your CRM, marketing automation platform, ad platforms, and product signals into a single model. For instance, when we run a data audit inside a new customer, we are identifying exactly these five failure points, building the taxonomy logic, and constructing the unification layer that makes the model trustworthy. Not as a one-time project but as a system that holds going forward.

When you bring those four problems into a room, the political pressure almost always runs in one direction: simplify. CMOs tend to pick one model or use last-touch. Give the CFO a number they can understand.

I have watched CMOs capitulate to this. The CFO stops asking questions. Attribution meeting ends in thirty minutes. And then six months later, pipeline is down and everyone blames market conditions.

Validity's 2025 CRM report found that only 19% of CRM users say leaders actually change course when presented with countering data, despite 84% of leaders claiming they do. PR Newswire

This gap between what leaders say they do and what they actually do is the organizational context inside which every attribution conversation happens. The data is not the obstacle. The organizational behavior around the data is.

Avinash Kaushik, who spent years as Google's Digital Marketing Evangelist, put it plainly: "Most companies have data graveyards." What he meant is that the data exists, but the insights are buried in it. The will to act on them is missing.

The CMOs who make it through the cleanup month go in knowing the political conversation is coming. They sit with the CRO before the model is built and agree on definitions. They frame the cleanup not as "we were wrong before" but as "we are building the foundation for decisions we could not make before."

The company that found $2M in hidden pipeline - never knew that this much pipeline existed, built by campaigns they had run and content their buyers had consumed, invisible because four systems had never been unified. When the data connected, account-level journeys became visible.

At another company, the cleanup revealed that one channel was eleven times more efficient than another on cost-per-user. Standard attribution had made them look equal. They had been over-investing in the wrong one for two years. The channel decay was accelerating. Nobody could see it because the measurement was wrong.

One CMO ran an incrementality test on her most-loved campaign. The one her board championed. The test came back with a -2.3x lift. Accounts exposed to the campaign were statistically less likely to convert than those that had not seen it. She asked if we could rerun the analysis.

She killed the campaign.

If any of those is a no, you are not ready to build a model. Poor data quality costs organizations an average of $12.9 million annually according to Gartner estimates. PPC Land

The month you spend fixing the foundation is the cheapest month in the project. The first month is not a detour from the real work, that is the real work.

Reply to this email if you want to walk through what this looks like inside your data. I will share what we are seeing.

Share it with your friends, if someone needs to map their data better.