An agency report is an argument, not a record. It was assembled by the party being evaluated, from settings they chose, answering questions they picked.
That’s not an accusation. Most agency reports aren’t dishonest — they’re curated. The agency selected which metrics to show, which time windows to compare, and which definitions to apply. Green arrows are easy to manufacture when you control all three. The report grades the agency’s homework with the agency’s rubric.
If you run a business between $5M and $50M and work with one or more agencies, you’ve likely felt the gap: the marketing agency report says everything is up, and your revenue doesn’t obviously agree. The standard response is to either become a marketing expert yourself or quietly lose confidence in the relationship. Both are bad options.
There’s a third: audit the report. Not adversarially — structurally. You need seven questions, asked consistently every month, that connect every claimed win back to systems you own: your CRM, your bank account, your pipeline. That’s the whole discipline. This post gives you the questions, the tells of a curated report, what a good report actually contains, and how to run the audit without poisoning a relationship worth keeping.
Why does an agency report need auditing at all?
Because the report is produced by the entity it evaluates, and no amount of good faith changes that structural fact. Curated is not the same as dishonest — but curated means chosen metrics, chosen windows, chosen definitions, and every one of those choices was made by someone with an incentive to renew the contract.
Think about what a monthly report actually is. Someone at the agency opened a set of dashboards, decided which numbers to export, decided which comparisons to draw, and composed a narrative around them. If organic traffic dipped, the report can lead with engagement rate. If cost per lead climbed, the window can shift from month-over-month to year-over-year. None of this requires lying. It only requires selection — and selection is the entire job of building a report. (It’s the same structural problem as every attribution dashboard grading its own homework — the settings belong to the party being measured.)
This is a governance problem, not a character problem. You wouldn’t accept a financial statement prepared by a vendor without reconciling it against your own books. Marketing agency accountability works the same way. The report is the agency’s account of its own performance; the audit is how you reconcile that account against reality as your systems record it.
The founders who navigate this well don’t learn attribution modeling or study ad platforms. They apply first-principles thinking to one question: does the claimed result appear anywhere I can independently verify? Everything else follows from that.
What are the seven questions?
Ask these seven questions of every report, every month, in roughly this order. Together they form a lattice that connects the agency’s claims to your systems — and they require zero marketing expertise to ask.
1. Which of these numbers appear in my systems?
This is the anchor question. For every headline metric in the report, ask where the corresponding entry lives in your CRM, your pipeline, or your bank account. Leads claimed should be findable as records you can open. Revenue attributed should reconcile against deals you can name. If a number exists only inside the agency’s dashboard and nowhere in your systems, it’s not yet a fact — it’s a claim awaiting verification. A number that can’t cross that boundary shouldn’t carry weight in the evaluation.
2. What is this metric a proxy for — and did the real thing move?
Every marketing metric stands in for something you actually care about. Impressions are a proxy for attention. Clicks are a proxy for interest. Marketing-qualified leads (MQLs) are a proxy for pipeline. The question is whether the real thing moved. If impressions doubled but qualified conversations didn’t budge, the proxy broke — and a report that celebrates the proxy without checking the destination is measuring the map, not the territory. Make the agency name what each metric is supposed to predict, then check whether the prediction held. (Which metrics actually predict revenue is its own conversation, and worth having with your team first.)
3. What changed because of this — decisions, not activity?
A report full of activity — posts published, campaigns launched, emails sent — describes effort. You’re paying for judgment. Ask what decisions the results drove: what got scaled up, what got cut, what hypothesis got retired. If the answer is “we’ll keep monitoring,” you’re funding observation, not operation. A team that owns its decisions can always tell you the last one it made and why.
4. What did you try that didn’t work?
A report with no failures is a brochure. Real marketing involves tests, and real tests fail — that’s what makes them tests. An agency that never reports a miss is either not experimenting or not disclosing. Neither is what you’re paying for. Ask directly: what did you try this month that underperformed, and what did it teach you? The quality of that answer tells you more about the agency’s process than any green arrow on the page.
5. What would you cut from your own scope if this were your money?
This question inverts the incentive. Agencies expand scope by default; nobody’s monthly report recommends shrinking the retainer. But an agency thinking alongside the business — rather than billing against it — can tell you which line item in their own scope is producing the least. If they can’t name one, they haven’t looked at their work with the scrutiny they expect you to skip.
6. Which of these results would have happened without you?
Some portion of your leads, traffic, and revenue was arriving before the agency did — from referrals, existing brand, sales relationships, seasonal demand. A report that claims the whole number claims your baseline as its own work. Ask the agency to distinguish what it caused from what it inherited. A precise answer may be genuinely hard; an honest attempt at one is not. Watch which you get.
7. What are you measuring next month that you can’t measure yet?
Good marketing operations mature. Attribution improves, tracking gets tighter, definitions get sharper — the measurement system should be evolving alongside the business. An agency that ships the identical report template for eighteen months has stopped improving how it sees. Ask what visibility they’re building next. The answer reveals whether they treat measurement as a deliverable or as infrastructure.
What are the tells of a curated report?
Certain patterns show up so reliably in curated reporting that they function as a checklist. None is proof of bad faith on its own. Two or three together mean the report is arguing, not recording.
Metric switching between months. Clicks were the headline last month; impressions are the headline this month. When the featured metric rotates, it’s usually rotating toward whatever went up. Hold the report to the same metrics every month and the rotation stops working.
Percentages with no absolute numbers. “Engagement up sharply” means something very different at large volume than at small volume. A percentage without its base is a number wearing a costume. Insist on both, always.
Activity lists dressed as outcomes. Twelve posts, four campaigns, three landing pages. That’s a description of hours, not results. Effort is table stakes; the report should account for what the effort produced.
“Brand awareness” invoked whenever performance dips. Brand is real and worth investing in. But when brand awareness appears as an explanation only in the months when measurable performance falls, it’s functioning as a shield, not a strategy. A genuine brand investment is declared up front, with its own goals — not discovered retroactively in a down month.
Screenshots instead of shared dashboard access. A screenshot is a moment the agency chose, cropped the way the agency chose. Direct access to the underlying dashboards — the ad accounts, the analytics property, the reporting tools — should be standard. If access is difficult to get, ask why. Keep asking.
What should a good report contain?
Flip every tell and you get the specification. A report built for accountability contains five things, and you can request all of them in one conversation.
The same metrics every month, against agreed targets. You and the agency agree on the scorecard once — which metrics, which definitions, which targets — and then it stays fixed. Consistency is what makes month-over-month comparison mean anything. Changes to the scorecard happen by agreement, in advance, not by editorial choice at report time.
Absolute numbers alongside every percentage. The base and the change, together, every time. This one requirement eliminates most of the room for cosmetic reporting.
A failures section. What was tested, what missed, what the miss taught. If your agency reporting doesn’t have a standing place for this, add one. It’s the fastest single upgrade you can make, because it changes what the agency prepares for — not just what it presents.
Spend against the plan. What was budgeted, what was deployed, and where the variance went. Money is the one metric with no proxy problem, which is exactly why it belongs in every report.
A “decisions needed from you” section. A strategic partner surfaces the choices only you can make — budget shifts, positioning calls, approvals that are blocking work. If nothing has needed your decision in six months, one of two things is true: the work isn’t touching anything strategic, or the decisions are being made without you. Either way, you want to know.
How do you run the audit without poisoning the relationship?
Announce it as standing practice, apply it to every vendor equally, and let the response do the judging. The framing matters more than the questions: this is how you review all external partners now, not a referendum on this one.
Say it plainly at the start of an engagement — or at the next monthly review of an existing one: “We reconcile every vendor report against our own systems. Here are the seven questions we ask each month. We’re telling you now so nothing is a surprise.” Announced in advance and applied uniformly, the audit is governance. Sprung selectively on one vendor after a bad quarter, it’s an ambush — and you’ll get defensiveness that tells you nothing.
Then watch the response, because how an agency responds to the audit is the audit.
A good agency will welcome it. Serious operators want clients who look at the numbers, because scrutiny is what separates them from competitors coasting on pretty dashboards. They’ll answer the failure question without flinching. They’ll grant dashboard access before you finish asking. They’ll treat “what would you cut from your own scope” as an interesting question rather than a threat — because they’ve already asked it of themselves.
A defensive response is also information. Not a verdict — some defensiveness is just surprise, and a fair process gives a partner a full cycle to adjust. But an agency that’s still deflecting the failure question three months in, still substituting screenshots for access, still rotating metrics after you’ve fixed the scorecard, has answered a bigger question than any single report could. You didn’t need to become a marketing expert to learn it. You needed seven questions and the discipline to keep asking them.
The deeper point is that the audit isn’t really about catching anyone. It’s about building the feedback loops that let you and your agency operate from the same set of facts — your facts, recorded in your systems, connected to your revenue. Clear decision ownership, consistent measurement, honest accounting of what worked and what didn’t. That’s not adversarial. That’s just how the work should run.
Where this leads
Once you’re asking these seven questions consistently, something shifts. You stop evaluating marketing by how the report makes you feel and start evaluating it by what your own systems confirm. Weak vendors surface quickly. Strong ones get stronger, because they’re finally working for a client who can tell the difference.
But the questions have a ceiling. They tell you whether the current work is real; they don’t tell you whether it’s the right work — whether the strategy behind the reports actually maps to where the business needs to go, whether the budget is composed correctly across channels, whether the whole architecture connects to pipeline the way it should. That’s a different altitude of question, and it usually takes someone who has sat on both sides of the reporting table — who has built these reports, reviewed these proposals, and knows exactly where the curation hides — to answer it holistically, alongside you rather than instead of you. If the audit reveals gaps you can’t close from where you sit, that’s the conversation worth having next.