Content Marketing Metrics: The Four Numbers Worth Paying For

Content marketing metrics that survive the invoice

Four numbers buy a decision. Everything else on your dashboard is a report line.

Consumption: did anyone actually read it

The exact metric is engaged sessions per URL, restricted to organic search, where engaged means the reader reached the middle of the body rather than the top of the page.

Weekly, one line per URL, no commentary.

What that buys you is one question: does this URL earn the next hour of work, or does the hour go somewhere else? You cannot take that fork on a number that counts arrivals.

Note what this excludes. A session that arrives, reads two lines and leaves is not consumption. A thousand of those is a distribution event, not a content result. Holding the threshold is the whole value of the metric, and setting it takes one afternoon with ten pages you know worked and ten you know did not.

Conversion: did it move anyone closer

The exact metric is assisted conversions credited to content URLs inside a defined window, and the window has to be written down. Ninety days is defensible for a considered purchase. Thirty is defensible for a trial signup. Neither is defensible if you change it mid-year.

Monthly, as a delta only. Not a total.

The budget question it answers is which cluster gets the next quarter of production capacity. That is the largest reallocation a content team makes, and it needs a number that survived contact with a buyer rather than a number that counted a click. Assisted rather than last-touch, because content is structurally first and rarely last.

Cost: what it took to make

Two exact metrics, both uncomfortable. Fully loaded cost per published asset, covering writing, editing, design and the distribution work that actually got it seen. Then cost per engaged session, which is that figure divided by the consumption number above.

Monthly, rolled up by format rather than by piece. The per-piece version produces arguments about individual articles. The per-format version produces a decision.

The decision it authorizes is which format you stop commissioning. A comparison page and a 2,000 word guide do not cost the same to make or return the same thing, and until both sit on one line with a common denominator nobody can say which is subsidizing the other.

This is also where two of the four govern each other. Consumption without cost tells you what worked. Cost without consumption tells you what was expensive. Together they tell you what to stop.

Put two formats on one line and the argument usually flips. A 1,200 word explainer costs $450 and buys 300 engaged sessions, so $1.50 each. A research report costs $2,400 and buys 900, so $2.67 each. The report is five times the investment at nearly twice the unit cost. Whether it earns another depends on the conversion rate behind those sessions, which the cost line cannot supply on its own.

Decay: is it still working

Measure the quarter-over-quarter change in engaged sessions per URL, on pages older than 90 days. Nothing else.

Quarterly. Once. Not monthly, because at a monthly cadence you will be reading noise and calling it a trend.

The decision it authorizes is refresh, consolidate or prune. Those are three different jobs with three different costs, and a page that is flat is none of them.

Decay is the one number here that is cheap to collect and expensive to ignore. An Animalz study tracked organic traffic declining at roughly 1.21% per week, which compounds to a half-life of about 57 weeks. It comes from a single blog, so read the figure as directional. HubSpot's own reporting puts refreshed and republished posts at an average of +106% in monthly organic views. Together, one good refresh roughly buys back a year of decay, which makes an annual cycle the minimum that keeps a page flat. Both figures are compiled secondhand rather than measured here.

What to stop collecting, and the reason on each line

The test is one question. Would a different value of this number, on its own, change what you do next week? If not, it is a report line.

Raw pageviews. It answers a question the consumption metric already answers, and it answers it worse, because arrival is not reading. Collection costs almost nothing, which is the trap: the cost lives in the explanation, and you pay that every month.

Average time on page. In GA4 the nearest equivalent is average engagement time, which counts foreground time with activity. A reader who finds the answer in fifteen seconds and closes the tab scores badly. A reader who leaves the tab open over lunch scores well. No decision survives a metric that rewards the second person.

Social shares. The decision shares would authorize is "make more things like this", but shares measure how a distribution channel felt that week, not whether the content did its job. For most B2B programs the number is also too small to separate from noise.

Keyword count. The number of keywords a page ranks for rises when you add a word to the body. It measures surface area, and no allocation decision should turn on surface area. If you need a search number, take clicks or take position and hold it to the same test.

Pieces published per month. This is the hardest to remove because it is the easiest to hit. It measures effort, and effort is an input you already know. Once it appears in a performance review, a results conversation quietly becomes an activity conversation and never comes back.

Five deletions. If your dashboard had fourteen numbers, that leaves five you have not been asked to defend yet. Name the hours one of them takes at your next review, out loud, then wait. Nobody will argue with the arithmetic, and nobody will volunteer a decision either.

Your year-over-year comparison broke, and it was not your fault

Every number above sits on a denominator that moved in the last two years, and almost nobody has adjusted for it.

Pew Research Center published a study in July 2025 covering 68,879 Google searches by 900 US adults, with data collected in March 2025. Where an AI summary appeared, users clicked a traditional search result in 8% of visits. Where it did not appear, they clicked in 15%. Summaries appeared on 18% of all searches in the sample, and on 53% of searches of ten words or more.

Read that twice before you quote it. Pew describes an association, not causation, and the study covers Google only. It does not say that summaries halved your clicks. It says the click base is not stable across query classes, which is enough to break the comparison you have been running.

So the scorecard gets a rule. Do not compare raw session or click counts year over year without knowing whether both periods carried the same exposure to summaries. Where you can segment by query length and query class, segment. Where you cannot, report the number with no comparison attached and say why in one line.

This is also why consumption is measured as a rate inside a defined population rather than as a raw count. Rates survive a moving base. Counts do not.

Attribution will not hand you one number

Content is usually the first thing a buyer reads and almost never the last thing they touch. Every attribution model therefore makes a choice on your behalf, and last-touch makes the worst one: it credits the final click and treats the two months of reading that made that click possible as if it never happened.

The honest position is that any single-number attribution claim for content is a choice, not a fact. Pick a model, write it on the scorecard, and leave it alone for the year. A model change looks exactly like a performance change in the data.

The mechanics live in multi-touch attribution, and the tooling question lives in marketing attribution software. What matters here is narrower: the conversion line on your scorecard names its model, or it is not a metric, it is a mood.

The one-page scorecard

One page. Four rows. No appendix.

Add an owner column. A metric with no owner is a metric nobody defends, and an undefended metric is the first thing removed when a quarter gets hard.

Then three rules underneath, because the table alone will not survive a meeting.

1. Every comparison states its base. If the base moved, the comparison says so on the same line.
2. Any new metric proposal arrives with its cost and the decision it would change. That includes whatever AI visibility number your team wants next. If you cannot produce the decision, the proposal waits.
3. Anything not on the page does not get reported. This is the rule that does the work. Dashboards expand because adding is easy and removing is awkward, and a scorecard that grows is a dashboard again.

The first month matters more than the design. Delete the five lines in week one and say nothing. Whatever breaks in week two, the person who complains is the owner you were looking for. Build the four rows in week three from data you already have, run the first page badly, fix it afterwards.

Leading indicators that need an expiry condition are a different question, handled in content marketing measurement.

What the page will not do is tell you why a piece worked, and it will not settle a brand argument, which has a longer lag than 90 days. Say so instead of bending the scorecard until it agrees.

The point of cutting

You will not get the hours back immediately. Removing a content marketing metric from a dashboard does not remove it from a conversation. What changes is what the meeting is about. Four defended numbers turn a status update into an allocation decision, and allocation decisions are the ones that move next quarter.

Start with the deletion list. It costs nothing, and within a month it tells you which content marketing metrics you were right to stop defending.