The Deal You Lost Was Never Real

(Was this newsletter forwarded to you? Sign up here.)

In January 2017, Marc Pritchard stood on a stage at the IAB Annual Leadership Meeting and broke an unwritten rule of the marketing industry. The Chief Brand Officer of Procter & Gamble, the largest advertiser in the world, called the digital media supply chain "murky at best, fraudulent at worst." He demanded viewability standards. Third-party verification. An audit trail. He said P&G would review every agency contract in 2017, and that the industry had until the end of the year to clean it up or P&G would invest elsewhere.

That was nine years ago. The supply chain has not gotten cleaner. It has gotten harder to see.

In 2023, Adalytics found that roughly 75% of ads purchased through Google's TrueView skippable in-stream offering were displayed on surfaces that did not meet Google's own definitions of in-stream. Skip buttons were hidden. Ads auto-played behind other content. Brands got refunds. In the same year, the Association of National Advertisers released the Programmatic Media Supply Chain Transparency Study, which concluded that the open web programmatic ecosystem contained as much as $20 billion in annual waste, with made-for-advertising sites alone accounting for 15% of spend and 21% of impressions. Only 36% of the studied spend went to brand-safe, non-MFA, measurable, viewable impressions. The other 64% was, in one form or another, fiction.

Apple's App Tracking Transparency shipped in April 2021, and Meta's CFO told analysts a year later that the change would cost the company roughly $10 billion in 2022 ad revenue. Apple's Mail Privacy Protection, which shipped a few months later, pre-caches inbox content, which means your email open rates have been fiction since September 2021. Google's cookie deprecation has been delayed so many times that most CMOs have stopped tracking the deadline. GA4 replaced Universal Analytics in 2023 and broke every year-over-year comparison that had been running for a decade.

And those are just the problems you can name.

Nobody has ever gotten this right

I have worked inside Series A startups and public companies, private equity rollups and family-owned businesses, mature SaaS and pre-revenue AI. I have seen account structures built by McKinsey and account structures held together with a Google Sheet and a prayer.

I have not, at any point in my career, walked into a company whose measurement stack I would describe as "figured out". Not once. Not close.

The companies that think they have it figured out are the most dangerous. They build strategy with conviction on top of a reporting layer that drifted six months ago. They reward channels that got flattered by a pixel glitch. They fire agencies over field renames nobody noticed. They make seven-figure decisions on data a junior analyst could have invalidated in an afternoon.

The companies that know they don't have it figured out are, counterintuitively, better off. They triangulate. They ask whether a number passes the smell test before acting on it. They sleep with one eye open.

I wrote about one version of this back in November in The Uncomfortable Truth About Incrementality. Attribution was never the truth. It was control. The platforms handed you the illusion of precision because precision made you dependent. This piece is about what happens when the illusion stops holding up at all.

The three fictions

Every CMO is operating inside three fictions simultaneously, and most of them only see one.

The fiction of digital attribution

Pixels fire wrong. UTM parameters get stripped by email clients, LinkedIn redirects, and any app with a reader mode. Dark social is invisible by definition. Rand Fishkin's research at SparkToro found that 76% of SparkToro's own web traffic is logged as "direct" because the referring platform quietly stripped the source. Most of what your analytics tool calls direct traffic is actually traffic from somewhere you can't see. Your data-driven attribution model is data-driven right up until the day Google quietly changes how it weights touchpoints, which it does, and never announces.

Byron Sharp and the Ehrenberg-Bass Institute have spent decades arguing that the things most marketers credit for their results are uncorrelated with what actually moves market share. Les Binet and Peter Field's 60/40 rule, built on 996 IPA effectiveness case studies across 700 brands over 30 years, made the case that short-term activation metrics systematically miss the long-term brand-building effect that accounts for the majority of marketing's revenue contribution. Bob Hoffman has been saying the same thing in cruder language for a decade at The Ad Contrarian: more than half of digital ad spend is wasted, the industry knows it, and the industry can't admit it.

These are not fringe voices. They are some of the loudest voices in the room. Most CMOs cannot hear them because admitting the scoreboard is fiction is professionally terrifying. If the scoreboard is fiction, the CFO has no basis for funding you.

The fiction of internal attribution

Marketing says "we sourced this account." Sales says "I closed it cold." Both can point to CRM evidence. Neither is wrong.

The credit assignment determines the budget assignment, which means the attribution debate is never actually a debate about attribution. It is a budget fight in a lab coat. Whichever function sets the rules wins the budget. Whoever loses the budget loses the headcount. The function that lost writes a new framework the following year to explain why the rules should change.

This fiction is worse than the first one because everyone inside the company participates in it. Marketing ops tunes the model to flatter marketing. RevOps tunes the model to flatter sales. Finance picks whichever version the CFO is inclined to believe this quarter. The "data-driven" decision in most boardrooms is just the version of the data that the loudest person in the room brought with them.

The fiction of time

The metrics you trusted last year were calculated against a CRM structure that no longer exists. Salesforce fields got renamed. HubSpot lifecycle stages got restructured. The definition of MQL moved. The definition of opportunity moved. The taxonomy for "channel" moved. Your year-over-year comparison is comparing two slightly different languages. Nobody tells you when a definition changes. The chart keeps drawing itself. It just means something different now.

A 2024 Gartner survey found that only 52% of senior marketing leaders could successfully prove marketing's value and receive credit for its contribution to business outcomes. Put differently, nearly half of CMOs cannot connect what they do to the numbers the CFO cares about, in a way the CFO will accept. That is not primarily a failure of execution. It is a failure of measurement so complete that the person running the function cannot defend their own results.

The pattern I see every month

Last week I was on a call with a CMO whose Q1 trial numbers had fallen off a cliff. Budget had already been cut. Q2 was already off to a slow start one week in. The performance problem had become a budget problem, which had become a staffing problem, which was about to become an agency problem.

Something about the story didn't sit right, so I pulled up two dashboards side by side. Their marketing-wide report showed a steep drop in paid-sourced trials. My paid media dashboard showed a drop too, but a different one. The numbers didn't reconcile. The two systems were telling two versions of the same bad story.

I dug in. The trial form on their site was overwriting UTM parameters on submission. A paid click would land on the trial page, the form would fire, and by the time it reached the backend the source parameters were gone. This was a structural quirk of how the form was built, not a new problem. My paid dashboard handled it one way, reconciling paid-sourced trials through click IDs and landing-page signals. Their marketing-wide dashboard handled it another, crediting trials to paid based on the landing page URL rather than the UTMs. Two systems, two workarounds, each internally consistent.

The actual problem was neither of those. The trial CTA tracking itself had been broken site-wide since mid-February. The trials were still coming in. They were simply not being recorded anywhere. The drop on both dashboards was not a performance decline. It was a measurement outage, visible from two angles in two systems, and confirmed by neither.

The decline and the bug were the same event.

The CMO had already taken real money off the table. Real headcount. Real quarterly guidance. All of it downstream of a JavaScript tag that nobody had audited in two months, layered on top of two reporting systems that nobody had reconciled to each other. By the time the tag got fixed, the smaller budget had already started producing smaller results, which confirmed the smaller budget had been the right call, which made the fiction self-fulfilling.

This is not a rare story. I see some version of it every single month. A CMO makes a real decision on the basis of a number that turns out to have been wrong, silently, since a software update or a field rename or a sync failure nobody was monitoring.

What to trust instead

Trust the things that are too big to be measurement artifacts. Trust revenue, not pipeline. Revenue gets audited by a CFO who has no incentive to flatter your model.

Trust retention and reactivation. Existing customers coming back is nearly impossible to fake in aggregate. If the cohorts are holding, something is working. If they are not, something is not, and no amount of first-touch attribution will save you.

Trust sales cycle length. If it is getting shorter, something is working. If it is getting longer, something is not. A tracker cannot hide that from you for long.

Trust customer interviews. Ten real conversations with buyers will tell you more about your positioning than a year of LinkedIn reporting. Bezos kept an empty chair in the room to represent the customer. Most CMOs keep a Looker dashboard.

Trust holdout tests. A geo holdout, a channel pause, an audience suppression. When you turn something off and revenue moves, you have learned something real. When you turn it off and nothing moves, you have learned something even more real.

Trust twelve-month trends, not twelve-day wiggles. Noise is loud. Signal is slow.

And trust the thing most CMOs were trained to stop trusting: taste. I wrote about this in The Judgment Call. The best CMOs I work with can look at a chart and tell you within thirty seconds whether the number is believable. They built that muscle by being wrong in public and learning from it. They do not trust the scoreboard. They use it as one data point among several, and they weigh it against their own smell test, their own customer conversations, and their own instincts about what is actually happening in the market.

The job

The job is not to defend the scoreboard. The job is to make decisions that would still be right if the scoreboard were off by thirty percent.

That is an unfamiliar muscle for most of us. We were trained inside a version of marketing that promised the scoreboard would get cleaner over time. It has not. It will not. The tools got more sophisticated and the underlying truth got murkier at the same rate.

Pritchard was nine years early. The supply chain is still crappy. The attribution is still fiction. The dashboards still lie, in the ways you can see and the ways you cannot.

The CMO who runs the playbook like the scoreboard is gospel is going to get fired by it.

The CMO who runs the playbook like the scoreboard is a useful but unreliable witness, who triangulates, who trusts taste alongside data, who is willing to say "I don't know" in a room where everyone else is pretending to know, is the one still standing in 2028.

A CMO told me last week they had lost a quarter of pipeline.

They hadn't.

They had lost a JavaScript tag.