Your quarterly business review went fine on paper. The slide said marketing sourced $41 million of pipeline, up 18% on the prior quarter. The CFO nodded and asked one question. How much of that would have closed if we had spent nothing?
You gave a reasonable answer about the multi-touch model. She wrote something down.
Three months later, your 2027 request came back 20% lighter, and nobody connected it to that meeting. Finance rarely announces the discount it applies to your numbers, and it usually takes that discount out of your budget.
What she wrote down was a ledger entry. Your CEO and CFO keep one on every executive who makes forward statements to them, and marketing makes a forward statement in almost every meeting it attends.
Two old models, one marketing ledger
Two well-worn frameworks explain how the ledger works. Stephen Covey's emotional bank account, from The 7 Habits of Highly Effective People, treats trust as a balance. "A kept promise" adds to it, while dishonesty and neglect take away from it (Franklin Planner on Covey's 7 Habits).
The trust equation, from The Trusted Advisor by David Maister, Charles Green and Robert Galford (getAbstract), is more precise. Credibility "has to do with the words we speak." Reliability asks "Do we meet expectations?" Intimacy covers whether people feel safe being vulnerable with you. All three sit over a single denominator, self-orientation, which measures whether your focus is on yourself (Trusted Advisor Associates). Green's firm writes that a high score in self-orientation "destroys" trust (Trusted Advisor Associates).
I call the marketing version the Credibility Ledger. Every forward statement you make to the CEO or CFO is an entry. A forecast, a launch date, a budget case, any sentence that begins "marketing drove." Each entry later clears as a deposit or a withdrawal, and the running balance sets how much of your next claim finance believes.
The entries are not the same size. My two defaults follow from that. Commit to forecasts you would hit four quarters out of five. Put only test-measured lift in the headline of any attribution claim, and label everything else as influenced.
The biggest withdrawal is a credit claim
An overclaimed attribution number is the costliest entry because it hits two terms of the trust equation at once. The words turn out wrong, which costs credibility. They turn out wrong in your favor, which reads as self-orientation. A missed forecast only costs reliability, and the CFO can see you trying.
Finance has good reason to expect the overclaim, and one of the best-known cases came from a CFO. In 2017 Procter & Gamble pulled digital spend that was reaching bots or landing in placements that did not help its brands. CFO Jon Moeller told analysts, "We didn't see a reduction in the growth rate," and said the spending "was largely ineffective" (Wolf Street, quoting P&G's earnings call, July 2017). The $100 million cut in the quarter to June had "little appreciable impact on the business," so P&G cut another $100 million by December (The Drum, March 2018).
Uber's former head of performance marketing, Kevin Frisch, described on the Marketing Today podcast how the company cut two-thirds of its ad budget and saw no change in rider app installs. Installs credited to paid channels "suddenly came through organically" (Daily Maverick, January 2021).
At a B2B software client, our attribution model ranked branded search as the best money in the plan. We went dark on it in a few regions for a quarter, and MQLs and closed-won barely moved.
A 2021 study of TV advertising for 288 consumer brands found effects "substantially smaller" than the published literature, and negative returns at the margin for more than 80% of brands (Shapiro, Hitsch and Tuchman, Econometrica, 2021).
Marketers know it too. In Haus's 2026 survey of 500 paid marketing and finance professionals, 35% said more than 20% of their marketing budget is inefficiently allocated (Haus, 2026). CMOs already sense where the skepticism sits. In Gartner's 2024 survey of 378 senior marketing leaders, only 52% said they were successful at proving marketing's value, and they ranked CFOs (40%) as the executives most skeptical of it (Gartner, September 2024).
The damage spreads past the one number. Accounting researchers studying 600 companies found the market "looks skeptically at forecasts that are higher than analysts' expectations" and "adjusts for the bias in good news forecasts by dampening its response" (Knowledge at Wharton, 2002). Your CFO runs the same adjustment on you. Once she suspects your numbers lean optimistic, she haircuts all of them, including the honest ones.
The default rule is a two-column claim. "Caused" holds revenue that a holdout or geo test measured. "Influenced" holds everything your attribution model assigns. Only the first column goes in the headline. If you have no test yet, headline the influenced number, label it in the same font size, and tell finance when a test will report. The case for testing is in The Uncomfortable Truth About Incrementality, and a starting design is in What Your CEO Says About You When You Leave the Room.
In Partnership with HubSpot
Turn AI into Your Income Engine
The cheapest deposit is a number you set
Reliability is meeting expectations. When you write the expectation yourself, you control what the deposit costs.
The common failure is forecasting at the number you hope for, because the plan rewards ambition in January. By September the gap shows up as a miss. My default is to commit at the level you would reach four quarters out of five, publish the range around it, and report against that commitment every quarter. Deviate upward only when you have at least four quarters of history on the metric and finance has asked for a stretch case. Even then, label the stretch case separately from the commitment.
The accuracy band matters in both directions. Beat your own forecast by more than 10% three quarters running, and finance starts reading your forecast as a floor. That is a withdrawal of its own, because it tells the CFO you were managing her. Aim to land within 10% either side, and re-forecast mid-quarter when you can see you will not.
The researchers in the Wharton piece argue managers need "reputations for truthful forecasting" (Knowledge at Wharton, 2002). Two other deposits build that reputation and cost almost nothing. Flag a likely miss early. My rule is to warn the CFO once the odds of missing pass 50%, and no later than week six of the quarter. The second is giving credit away. When sales and marketing both touched a deal, put it on the shared line, even when your model says marketing sourced it.
The Credibility Ledger
These are the entries I see most often, with the weight I assign each. The weights are my judgment from sitting in these reviews. Nobody has measured them.
Entry | Clears as | Weight | Default rule |
Hit a forecast you set | Deposit | +1 | Commit at the level you hit four quarters in five |
Flag a miss before week six | Deposit | +2 | Warn once the odds of missing pass 50% |
Report a test that failed | Deposit | +2 | Share the result and the spend you cut |
Give credit to sales or product | Deposit | +1 | Shared deals go on the shared line |
Return budget you cannot use well | Deposit | +2 | Return it in the quarter you know |
Miss a forecast you set | Withdrawal | -2 | Own it on the first slide |
Finance finds the miss first | Withdrawal | -3 | Report misses before finance closes the books |
Beat by 10%+ three quarters running | Withdrawal | -1 | Re-forecast upward mid-quarter |
Change a metric definition mid-year | Withdrawal | -3 | Restate history under the new definition |
Claim a test later disproves | Withdrawal | -5 | Headline only test-measured lift |
On these weights, one disproved attribution claim costs five quarters of forecasts hit on the nose. A CMO who forecasts conservatively and claims narrowly can bank a year of deposits with no heroics. A CMO who claims broadly can spend that year in one meeting.
Keep the ledger as an actual document. List each promise with the date, the audience, the number and the actual. Open every quarterly review with last quarter's promises against what happened, before anyone asks. The CFO is already keeping this ledger. Keeping your own version means the two copies match, and building the model with FP&A, as I argued in The CFO Problem, means they start from the same numbers.
Audit your own book before planning closes
Your 2027 plan likely goes to finance within weeks. Before it does, pull every deck you showed the CEO or CFO this year. List each number you stated as marketing's contribution and sort it into the caused or influenced column.
Then restate the ones that fail, in writing, before finance asks. A correction you volunteer is a deposit under every rule above. The same correction extracted by the CFO in a budget meeting is the five-point withdrawal, and it lands in the one meeting where your balance sets your budget. If you are heading into a board cycle after that, the same discipline applies to the one number you bring into the room with the directors.
Send the restatement with your plan, and put the caused column first.
James