I sat in a quarterly review meeting last year where an agency presented their client with a dashboard showing nine charts. Reach was up 34 percent. Engagement was up 12 percent. Follower growth was up 8 percent. Impressions had crossed two million for the first time. Sentiment analysis showed 72 percent positive mentions. The client was visibly pleased. The agency was visibly relieved. The meeting ran short, everyone shook hands, and the retainer continued.
I was in the room as a third-party observer for an unrelated reason. What I noticed, looking around at the dashboard, was that not a single chart on the screen actually showed whether the client's business had improved. Revenue from the relevant channels was not on the slide. Customer acquisition costs were not on the slide. Inbound inquiry quality was not on the slide. The metrics that would have told the room whether the marketing was working — in the sense that the client was getting business outcomes proportional to the spend — were absent entirely. The metrics that were present looked impressive and predicted nothing.
This is the structural failure mode at the heart of most marketing reporting, and almost nobody publishes about it honestly because doing so threatens the business model of every agency that participates in the theater. The dashboards, the monthly reports, the engagement metrics, the sentiment analyses — most of it is performance. It is designed to make the client feel that something productive is happening. It is not designed to actually tell anyone whether the work is working. The two purposes look similar from outside the industry. They are deeply different from inside it.
This article is the argument I have been making to clients privately for the last year. The metrics that matter are mostly not the metrics on the dashboard. The signals that actually predict whether content is working are smaller, harder to chart, easier to overlook, and structurally inconvenient for the agency-client reporting relationship. Recognizing the difference is the most useful piece of measurement thinking any operator can do.
Key Takeaways
- Most published marketing reporting is theater. Reach, engagement, follower growth, sentiment analysis, and impression counts are easy to chart and almost never predict whether the work is actually generating business outcomes.
- The metrics that matter are smaller and harder to report. Inbound inquiry quality, the specific language audiences use when they convert, retention curves beyond the first week, and direct attribution conversations are far more predictive than any topline dashboard metric.
- Vanity metrics, health metrics, and outcome metrics are three different categories. Confusing them is the root cause of most measurement failure. Vanity metrics measure visibility. Health metrics measure relationship strength. Outcome metrics measure business impact. They are not interchangeable.
- The agencies that report theater are mostly not malicious. They are reporting what is easy to measure, easy to chart, and easy to defend. The misalignment is structural rather than intentional, which is why it persists.
- Honest reporting requires reporting uncertainty. The signals that actually predict outcomes often take 6 to 12 months to interpret, and any report that promises certainty in 30 days is performing competence rather than demonstrating it.
Why So Much Of It Is Theater
To understand why marketing reporting drifts toward theater, you have to understand the incentives operating on both sides of the relationship.
For the agency, reporting needs to justify the monthly invoice. The simplest way to do this is to show numbers going up. Reach, engagement, followers, impressions — all of these can almost always be made to go up with some combination of effort and time, and "going up" looks like progress on a dashboard. The agency is rewarded for the appearance of progress, which means the metrics that get prioritized are the ones easiest to make appear to progress.
For the client, evaluating marketing performance is genuinely difficult. Most clients do not have the in-house expertise to know which metrics predict outcomes and which are vanity. The dashboard the agency provides becomes the evaluation framework, by default, because the alternative — building their own measurement model — is too much work for most clients to undertake. The client trusts the agency's framing, and the agency's framing is built around metrics that defend the retainer.
The result is a reporting equilibrium that serves neither party's actual interests. The agency spends time producing dashboards that do not help anyone make better decisions. The client receives reports that feel reassuring but predict nothing. The actual question — is this working in a way that will affect the business? — almost never gets answered because almost nobody in the meeting is structurally incentivized to ask it. [BACKLINK PLACEHOLDER → external: a credible critique of marketing measurement, e.g. from the Ehrenberg-Bass Institute, Marketing Week, or HBR's coverage of marketing attribution. Aligns with the $5–10 CPC on analytics tools and measurement platforms.]
The Taxonomy: Vanity, Health, Outcome
The most useful distinction I have arrived at after several years of looking at this honestly is the separation of metrics into three categories. Most published measurement advice conflates them. Most agency dashboards present them as if they were equivalent. They are not.
Vanity metrics measure visibility without measuring anything about whether the visibility translates into outcomes. Reach, impressions, followers, total likes, sentiment percentages. These metrics describe how many eyeballs your content reached and how many of those eyeballs performed a low-cost positive gesture. They tell you almost nothing about whether the people who saw the content care, will remember it, or will ever act on it. Vanity metrics are the simplest to chart because they are mechanically easy to measure, which is exactly why they dominate dashboards.
Health metrics measure relationship strength — the depth and quality of the connection between the audience and the brand, regardless of total scale. Save rate, share rate, average watch time, repeat engagement from the same accounts, direct messages, comments containing specific language, email reply rates if you have a newsletter. These metrics are harder to chart cleanly but predict outcomes much more reliably than vanity metrics, because they describe whether the audience cares enough to do something that requires real cognitive investment.
Outcome metrics measure business impact directly. Inbound inquiries from the relevant channel. Conversion rate from inquiry to engagement. Customer LTV by acquisition source. Sales velocity. Margin per acquired customer. These metrics are the ones that actually answer "is this working" — and they are the metrics most often missing from dashboard reporting, because they are harder to attribute, slower to materialize, and frequently uncomfortable when they reveal that vanity metrics and outcomes are not correlated.
The structural failure of most marketing reporting is that it presents vanity metrics with the visual prominence of outcome metrics, while outcome metrics are either absent or buried. The dashboards optimize for what is easy to show, not for what is true. [BACKLINK PLACEHOLDER → external: an analytics tool's documentation on attribution and outcome measurement, e.g. Google Analytics 4, Mixpanel, Amplitude, or PostHog's resources on cohort and outcome analysis.]
What Actually Predicts Whether Content Works
After years of watching the gap between dashboard performance and actual business outcomes, I have come to trust a much smaller set of signals than the published reporting frameworks suggest. None of them are revolutionary. Most of them are unfashionable specifically because they are difficult to chart and slow to materialize. Together, they predict whether marketing is working far more reliably than any topline dashboard.
The specific language inbound leads use when they reach out. When someone messages a brand or fills out a contact form, the language they use often reveals where they encountered the work and which piece of content moved them. "I saw your post about X." "Your video on Y made me think about this." These mentions are gold, because they are the only directly attributable signal of what actually drove the conversation. Tracking them systematically — even just by keeping a note per inbound contact — produces a clearer picture of which content is working than any analytics dashboard.
Save rate and share rate, not like rate. Likes are almost free to give. Saves and shares require the user to value the content enough to do something with it. A piece of content with a 5 percent save rate is dramatically more valuable than one with a 50 percent like rate, even though the like rate looks more impressive. The platforms that surface save and share counts are giving you the signal that actually predicts whether the audience is treating the content as useful versus pleasant.
Retention curves beyond the first week. Initial reach tells you nothing. The piece of content people are still watching a month later, the post still gathering saves three weeks after publication, the video being shared in private DMs long after its main reach cycle — these are the content pieces that compound. The first week of engagement is noise. The signal is what happens after the algorithm has stopped pushing the work and people are choosing to return to it on their own. [BACKLINK PLACEHOLDER → internal: link to article #8 (well-edited video that failed analytics) — both pieces deal with reading retention data honestly.]
Inbound inquiry quality, not quantity. A pipeline of low-quality leads from a high-reach platform is worse than a smaller pipeline of high-quality leads from a more specific one. Quality is measured by the specificity of the inquiry, the fit with the offer, and the probability of conversion. Counting leads without weighting them is a dashboard mistake that hides everything that matters about how the marketing is actually performing.
Repeat engagement from the same audience over time. A person who has engaged with three pieces of content over six months is far more likely to convert than a person who engaged with one piece a single time. Most platform analytics do not surface this signal cleanly, which is why most reporting ignores it. The brands that track it — usually by sampling specific audience members and following their journeys — develop a far sharper understanding of who is actually being reached, and who is being touched by a single impression that means nothing.
Why The Real Signals Are Unfashionable
The signals I have just listed share a set of characteristics that explain why they remain unfashionable.
They are harder to chart cleanly. A "save rate" is a single number, but interpreting it requires context that does not fit on a dashboard tile. Retention curves beyond the first week are slow to materialize, which means a monthly report cannot show them at the cadence agencies are expected to report. Inbound inquiry quality is qualitative, which means it does not aggregate into a green-arrow trend line.
They require honest acknowledgment of uncertainty. A report that says "we saw three inbound leads this month with language suggesting they came from our content, and we cannot prove which specific piece of content drove them" is more truthful than a report that says "engagement is up 14 percent." The first report is harder to defend in a meeting. The second is easier to defend and predicts nothing.
They are slower to interpret. The signals that actually matter often take 6 to 12 months to read clearly. A monthly cadence of reporting forces the conversation onto the metrics that move monthly, which are almost always the vanity metrics. The metrics that matter operate on a slower clock, and the reporting cadence does not accommodate them. [BACKLINK PLACEHOLDER → internal: link to article #4 (fast-approving clients) — both pieces deal with misreading signals that look fine but predict little.]
What Honest Reporting Looks Like
If I were redesigning marketing reports from scratch, with no incentive to perform competence, the structure would look very different from the standard dashboard. It would have three sections.
Section one: what we are doing and at what cadence. A simple description of the work being produced, the platforms it is being deployed on, and the consistency of execution. This section does not need charts. It needs honest documentation of inputs.
Section two: health signals. Save rate, share rate, retention curves, repeat engagement, inbound message quality. These are the metrics that predict whether the work is creating real audience traction. Some of them are quantitative; some are qualitative. All of them should be reported with acknowledgment of what they do and do not tell us.
Section three: outcome signals and uncertainty. Inbound inquiries traced to content, conversion data when available, and explicit acknowledgment of what is not yet possible to attribute. This section should include the phrase "we do not yet have enough data to say" more often than it appears in any current report. Reporting uncertainty honestly is more useful than reporting false certainty confidently.
A report structured this way is shorter, less visually impressive, and significantly more useful than the standard nine-chart dashboard. It is also harder to use as a justification document, which is part of why almost no agency produces it voluntarily. The agencies that do produce it tend to be the agencies whose retainers are easiest to defend, because the reporting itself demonstrates a level of intellectual honesty that the dashboards specifically obscure.
