How to set up a content strategy that closes the loop

To set up a content strategy that actually earns its keep, build it as a loop rather than a calendar: produce a page aimed at one buyer and one reason to search, measure what it does at the level of enquiries, decide the next page from that evidence, and adjust the plan every month. A calendar tells you what to publish; a loop tells you what to publish next, and that difference is the whole game.

Content strategy and SEO: one system, not two tasks

A content strategy decides what you publish and why; SEO decides whether anyone finds it. Treat them as separate workstreams and you get one of two failure modes: brilliant content nobody searches for, or keyword-stuffed pages nobody trusts. The fix is to build SEO into the strategy from the first step, not bolt it on after publishing.

Read it as a content strategy for SEO: the plan starts from what your buyers actually type, not from what the business feels like saying. Every page on the calendar exists because a real query, with real demand, needs an answer your site is positioned to give. That single shift, from topics to queries, is what separates a strategy from a publishing habit, and it only holds when SEO and content strategy are planned as one system instead of handed off between people.

In practice, one system means the search work happens before anything is drafted: the page's target question is named, the demand behind it is checked, and the number that will decide the page's fate is agreed in advance. Publishing then executes a decision already made, a different job from publishing and hoping.

Why most content strategies fail at search

The direct answer: they start with topics instead of queries. A team sits in a room, lists themes that feel on-brand, and schedules them. Some of that content ranks by accident. Most of it doesn't, because "what we want to say" and "what people search for" were never reconciled.

Three patterns show up in almost every failed content strategy we audit:

• ✓No feedback loop. The plan runs for a quarter regardless of results. Winners aren't doubled, losers aren't killed.
• ✓Topic-first planning. No keyword or query data behind the calendar, so there is no way to estimate demand before investing in a piece.
• ✓No owner for measurement. Traffic gets reported, but rankings, impressions and conversions per page don't. Nobody can say which posts earned their cost.

The fix is a principle, not a checklist: plan strategy and SEO as one system from the first step, with one owner page per query and a monthly measure wired in before publishing starts. The rest of this guide runs that system as a working loop, produce, measure, decide, adjust, and the sections below walk the loop stage by stage, in the order it has to run.

A strategy run this way compounds: every page you publish either earns its place or tells you why it didn't. If you'd rather skip the trial-and-error phase, get a free audit and we'll show you exactly where your content strategy stands today, and which queries are within reach in the next 90 days, and which ones aren't worth chasing yet.

Start from the decision the content has to serve

Before a single topic goes on a list, write down the business decision the strategy exists to influence, because everything downstream is measured against it. For most established businesses that decision is an enquiry from someone who could actually buy, not a subscriber, not a pageview, not a share. Naming the decision up front stops the strategy drifting into vanity work, and it gives you the one number you'll use later to decide whether any page was worth publishing.

Produce: one page, one cohort, one reason to search

Every page should target exactly one cohort at one moment of intent, and if you can't name both in a sentence, the page isn't ready to write. A cohort is narrower than an audience: the operations manager comparing suppliers before a renewal is a different person, asking a different question, than the owner pricing a first purchase. When AI can draft competent copy in seconds, the scarce input is no longer words, it's the judgement about who a page is for and why they'd be searching. Put that judgement into a tight brief and the production itself, the research, the drafting, the internal links, becomes the cheap part.

Guard against a subtler failure here: two pages aimed at the same query. When a cluster grows, it's easy to publish a second article that competes with one you already have, splitting the clicks and confusing the engines about which page owns the topic. Assign one owner page per query before you brief the next piece, and expand the owner rather than writing a rival.

The SEO content brief: the one-page template that makes every page worth writing

A content brief is a one-page document that names who a page is for, what they already believe, what question the page must own, and what the reader should do next — written before a word of the page is drafted. When drafting became cheap, the brief became the scarce input: anyone can produce competent prose now, and the difference between a page that earns enquiries and one that fills a calendar is decided entirely in the brief. This section gives you the six-line template, the pre-flight checklist, and a worked example you can lift directly.

Why the brief is the whole game

The failure mode of modern content isn't bad writing; it's pages with no reason to exist. A brief forces every page to pass four gates before production: one cohort, one moment of intent, one owning query, one action. If any gate can't be answered in a sentence, the page isn't ready — and discovering that now, on one page of planning, costs minutes instead of discovering it in three months of rankings data.

The brief is also your cannibalisation defence. A written record of which page owns which query, made before briefs go to production, is what stops two of your own pages competing for the same search and splitting the clicks that would otherwise compound.

The six-line brief template

Every brief, whatever the page, answers these six lines in order. They take the template's three-line brief standard from later in this guide — who it's for, what they already believe, what they do next — and add the three lines that make it enforceable: the query the page owns, the claim it makes, and the evidence behind it. Each line exists to make the next one easier:

1. 1.The cohort. One buyer, at one moment — not an audience, a person mid-decision. Good: "the office manager comparing suppliers three months before a contract renewal." Not good: "business owners."
2. 2.The reason to search. The question this person is actually typing, in their words. This becomes the page's target query and its H1 candidate. One question per page, and the question must not already belong to another page of yours — check the ownership map first.
3. 3.What they already believe. The assumption the page has to meet or overturn. Good: "they believe switching suppliers always means a service gap." This line is what separates a page that sounds like everyone else's from one that gets quoted.
4. 4.The claim the page will make. One sentence, defensible, specific. If you can't state the page's point in one sentence, the page doesn't have one yet. Vague claims produce vague pages, and AI assistants — which now read everything — cite specific ones.
5. 5.The evidence. What makes the claim believable: numbers you can stand behind, a worked example, a named case, first-party data. Never brief a page whose evidence line is empty; that's the line that gets content quoted and ranked.
6. 6.The next step. The single action a convinced reader takes. One per page — a reader offered three choices takes none.

The pre-flight checklist

Before a brief goes to production, run it past this checklist. Any box unticked sends it back:

• ✓✓ The target query returns real search volume, and no other page on the site already owns it.
• ✓✓ Two competitors ranking for the query have been read, and the brief states plainly what this page will say that they don't.
• ✓✓ The claim can survive a sceptical reader asking "says who?" — the evidence line names the answer.
• ✓✓ The page has a planned place in the site's internal links: at least one existing page that will point to it, and one it will point onward to.
• ✓✓ The next step is one action, and the page that receives the traffic is ready for it.

That first box is the content audit discipline in miniature: the ownership check is the same check you run across the whole site each quarter, applied at the moment it's cheapest — before anything is written.

A worked brief, filled in

Using the commercial cleaning example from earlier in this guide, here's the brief for one page:

1. 1.Cohort: the office manager comparing suppliers before a renewal.
2. 2.Reason to search: "how to compare cleaning quotes."
3. 3.Already believes: that the cheapest quote is the risky one, but doesn't know what to check instead.
4. 4.Claim: the difference between quotes is almost never the hourly rate — it's what's excluded from scope.
5. 5.Evidence: a line-item comparison of two anonymised real quotes showing the same building priced two ways, with the exclusions that changed the true cost.
6. 6.Next step: request a line-item quote from both finalists.

Six lines, one page, no ambiguity about who it's for or why it exists. That page can be written by anyone — including an AI assistant — and still end up pointed at the right reader, making the right point, with the proof to back it.

The four ways briefs fail

• ✓✓ The cohort is actually two people. "Small business owners and marketing managers" is two briefs. Split it; each page earns more when it aims at one reader.
• ✓✓ The claim is a topic. "Everything about cleaning contracts" is a subject, not a point. If the claim line reads like a category, the page will read like a Wikipedia stub.
• ✓✓ The evidence line is "our experience." Experience only counts when it's made specific enough to check. A number, an example, a named case — something a reader, or an AI assistant quoting you, can point to.
• ✓✓ The next step is the reader's problem to solve. "Consider your options" is not a next step. Name the action, make it one, and make sure the page it's on converts.

Keep briefs where the loop can see them

A brief is not paperwork; it's the unit the measurement loop runs on. When the monthly review finds a page that produced enquiries, you don't just build its neighbour — you pull the brief that worked and copy its shape. When a page pulled traffic but no enquiries, the first place to look is the brief: wrong cohort, wrong query, or a next step nobody took. The brief template above is how you find out which, in one page instead of one post-mortem.

If you'd rather skip the trial-and-error, the free audit shows which of your existing pages are worth briefing neighbours for, where two pages are splitting the same query, and which queries your buyers ask that no page of yours owns yet — top-10 fixes, competitor analysis and keyword opportunities, run from the homepage.

Measure: enquiries and citations, not pageviews

Close the measurement loop at the level of enquiries, and, increasingly, at the level of who the AI assistants name, because those are the two things that move the business. Traffic is a leading indicator at best and a distraction at worst: a page can triple its visitors while attracting exactly the wrong people. Trace each enquiry back to the page and cohort that produced it, and separately track whether your business gets cited when buyers ask assistants the questions in your market. That second signal barely existed two years ago and is still thinly contested, which is precisely why it's worth instrumenting now.

Decide: let the data name the next page

The next month's topics should be chosen by what the last month's data revealed, not by a plan written in advance. A page that produced enquiries tells you to build its neighbours, the adjacent questions the same cohort asks. A page that pulled traffic but no enquiries tells you the intent was wrong or the capture was missing. A question where a competitor gets cited and you don't tells you exactly where the next page aimed at AI-assistant visibility should point. Deciding from evidence is what separates a strategy that compounds from a backlog that merely grows.

Adjust: cull, double down, repeat

Once a quarter, prune as deliberately as you publish, because dead weight costs you crawl budget and dilutes the topical authority the engines read. Cull or merge pages that attract the wrong people as ruthlessly as ones that attract nobody. Double down where enquiries actually came from. Then run the loop again. The businesses that win at content aren't the ones that publish the most; they're the ones whose fourth month is informed by their first three.

A sample content strategy, worked end to end

A content strategy is a loop with a calendar attached, so the clearest way to show one is to run the loop for a real-shaped business and write down every decision. The example below is illustrative, a composite of the kind of established B2B service business we work with most, but every step is exactly what we'd run in the first ninety days. You can lift the structure wholesale.

The business and the decision it has to serve

The example business is a ten-person commercial cleaning company turning over mid-seven figures, selling contracts to office managers and operations directors in one metro area. The content strategy exists to serve one decision: a qualified enquiry for a contract of twenty desks or more. Not brand awareness, not recruitment, not pageviews. One enquiry type, named in a sentence, so every later call, publish, expand, cull, has a single yardstick.

The first ninety days: produce one page per cohort and intent

The first quarter publishes eight pages, each aimed at one cohort at one moment of intent. Nothing goes on the list until both are nameable:

Notice what the list is not: it is not "eight blog posts about cleaning." Pages 1–6 each own one question the buyer genuinely types, page 7 carries the proof, and page 8 captures a cohort the competitors ignore. One owner page per query, no two pages competing for the same search.

What gets measured, monthly

At the end of each month, two numbers get written down per page: enquiries traced to it, and whether the business is named when a buyer asks an AI assistant the questions on this list. Traffic is recorded but never decides anything on its own, a page can triple its visitors and still be worthless if the visitors were students and window-shoppers. The enquiry count decides whether a page lives; the citation check decides where the next page points.

The decision rules for month two and beyond

The strategy never plans more than one month ahead, because each month's plan is written by last month's evidence:

• ✓A page that produced an enquiry gets a neighbour: the adjacent question its cohort also asks. If page 2 brought a comparison enquiry, month two adds "questions to ask a cleaning company before you sign."
• ✓A page that pulled traffic but no enquiries gets fixed, not expanded: the intent was wrong or the capture is missing. Check who actually ranks for the query and what they promise.
• ✓A question where a competitor is cited and you are not becomes the next page aimed at AI-assistant visibility, full stop. That gap is the strategy telling you where demand exists and no one owns it.
• ✓A page that did nothing in ninety days, no enquiries, no impressions worth the name, goes on the cull list for the quarterly review.

The refresh cadence

Every page carries an expiry date, because content decays: prices move, regulations change, and a page that was right eighteen months ago quietly becomes the reason a buyer stops reading. Once a quarter, each live page gets a twenty-minute review: is every claim still true, does the number in the headline still match reality, is there a newer question this page should now answer? Pages that still earn enquiries get refreshed in place, same URL, updated body, date honestly bumped. Pages past saving get merged into their nearest neighbour with a redirect, not left to rot and dilute the site's authority. Refreshing is not admitting the old work failed; it is the loop doing exactly what it was built to do.

Steal this structure

Stripped to its skeleton, the whole strategy fits on one card:

1. 1.Decision: name the one enquiry the strategy exists to produce.
2. 2.Cohorts: list the two or three buyers, each with the moment they go looking.
3. 3.First eight pages: one per cohort-and-intent pair, one owner page per query.
4. 4.Monthly measure: enquiries traced per page, plus an AI-visibility check on the buyer questions.
5. 5.Next month: expand what produced enquiries, fix what pulled the wrong people, fill the citation gaps.
6. 6.Quarterly adjust: refresh what earns, cull or merge what doesn't.

That card is the template. The specifics will differ by business, but the loop is identical, and it beats a publishing calendar every time for one reason: in month four you are publishing what month one proved, and a calendar never lets you do that.

If you want this loop run against your own pages, which ones produce enquiries today, where the gaps are, and whether AI assistants cite you for the questions your buyers ask, the free audit covers exactly that: top-10 fixes, competitor analysis and keyword opportunities, run from the homepage.

The content strategy template (fill this in, then delete every field you can't defend)

A content strategy template is a one-page document with eight fields: the decision the content exists to serve, the cohorts and their reasons to search, a query-ownership map, the first quarter's page list, a brief standard, the two numbers that get measured, the decision rules for what happens next, and the date of the next quarterly review. If a field can't be filled in with a sentence, the strategy isn't ready — and a template that can't say no is just a publishing calendar with better formatting.

This is the template the worked example above was built from. Fill it in in this order; each field exists to make the next one easier.

Field 1: The decision

One sentence: the single enquiry or action this strategy exists to produce. Not a mission, not an audience — a decision someone takes that the business can count.

• ✓Good: "A qualified enquiry for a contract of twenty desks or more."
• ✓Not good: "Raise awareness of our brand." (Nothing downstream can be measured against it.)

If you write two decisions, you have two strategies. Pick one; the other gets its own template later.

Field 2: Cohorts and their reasons to search

Two or three rows, each naming one buyer and the moment they go looking. A cohort is narrower than an audience: not "office managers" but "the office manager comparing suppliers three months before a contract renewal."

• ✓Good: "Ops director, mid-renewal, unhappy with service — searching how to switch without gaps."
• ✓Not good: "Anyone interested in facilities." (You can't write a page for them.)

Every page you brief later must trace back to a row here. A page with no parent row is vanity work wearing a brief.

Field 3: Query-ownership map

A list of the questions each cohort actually types, with one owning page assigned to each question — before anything is written. This field exists to prevent the most expensive failure in content: two of your own pages competing for the same query, splitting the clicks and confusing the engines about which one owns the topic.

• ✓Good: "how to switch cleaning companies → /switching-cleaning-companies (owner). Nothing else may target this query."
• ✓Not good: a topic list. Topics invite overlap; questions force ownership.

Field 4: The first-quarter page list

Eight pages, one per cohort-and-intent pair from Field 2 and Field 3, each with a working title and the reason it exists. Two strong pages a month beat eight thin ones, because only published, measured pages feed the loop.

If the list runs past eight, cut it. The pages you cut aren't lost — they're the second quarter, chosen by evidence instead of enthusiasm.

Field 5: The brief standard

Three lines every brief must answer before production starts: who is this for (one cohort, one moment), what do they already believe, and what should they do next. When drafting is cheap, the brief is the scarce input — a page written without these three lines is a lottery ticket.

Field 6: The two numbers

The only metrics that decide anything: enquiries traced to each page, and whether the business gets cited when a buyer asks an AI assistant the questions from Field 3. Traffic gets recorded but never decides on its own — a page can triple its visitors and still be worthless if the visitors were never going to buy.

Write down, next to the two numbers, how they'll be captured. Uninstrumented numbers are wishes.

Field 7: The decision rules

Four rules, agreed before the first page ships:

• ✓A page that produced an enquiry gets a neighbour — the adjacent question its cohort also asks.
• ✓A page that pulled traffic but no enquiries gets fixed, not expanded: the intent was wrong or the capture is missing.
• ✓A question where a competitor is cited and you are not becomes the next page aimed at AI-assistant visibility.
• ✓A page that did nothing in ninety days goes on the quarterly cull list.

Rules written in advance stop the monthly review becoming a mood.

Field 8: The next review date

One date, calendared, with an owner. Quarterly for the full template; monthly for Fields 6 and 7 alone. A template with no review date is a snapshot; a template with one is a system.

What to leave out

Three things that appear in most content strategy templates and earn their place in none:

• ✓Channel plans. Channels are tactics; they change quarterly and belong in the calendar, not the strategy. The strategy is who you're for and what they ask.
• ✓Persona portraits. Demographic detail feels rigorous but decides nothing. The cohort's moment of intent is the usable signal; their age and hobbies are not.
• ✓Publishing volume targets. "Three posts a week" is an output goal that rewards the wrong behaviour. The volume that matters is measured pages per month, and the template's floor is two.

The three ways templates fail

• ✓It's filled in once and filed. A template that isn't revisited monthly is a one-time exercise; the value is in Field 7 firing every month, not in the document existing.
• ✓Field 6 gets skipped because attribution is hard. Fix attribution first. A strategy that can't trace an enquiry to a page cannot decide anything and quietly reverts to a calendar.
• ✓Field 3 gets skipped because it feels bureaucratic. This is the field that prevents cannibalisation — two owner pages per query is the single most common way content programmes leak authority. Five minutes here saves a quarter of split clicks later.

Steal the fields, keep the loop

Copy the eight fields into whatever tool you already use — a document, a sheet, a project board. The tool is irrelevant; the order isn't. Decision before cohorts, cohorts before queries, ownership before briefs, measurement before volume. Get the order right and the template runs the strategy. Get it wrong and the strategy runs you.

If you'd rather see the loop run against pages that already exist — which ones produce enquiries today, where two pages are splitting the same query, and whether AI assistants cite you for the questions your buyers ask — the free audit covers exactly that: top-10 fixes, competitor analysis and keyword opportunities, run from the homepage.

How long before it pays off

Be honest with yourself about the timeline: a content strategy built as a loop compounds over months, not weeks, and anyone promising faster is selling you the calendar, not the results. The first few pages exist mainly to generate the data that makes the next decisions better; the returns arrive once the loop has run enough cycles that you're consistently building neighbours of pages you already know produce enquiries. That's also why starting matters more than perfecting, an average page published this month, measured properly, beats a perfect page published next quarter, because only the published one feeds the loop.

The loop, in order

• ✓Decision: name the enquiry the strategy exists to produce
• ✓Produce: one page per cohort and intent, one owner page per query
• ✓Measure: enquiries traced to source, plus AI citations in your market
• ✓Decide: pick next topics from what produced enquiries and where you're not cited
• ✓Adjust: cull the wrong-fit pages, expand the winners, repeat monthly

Measure what matters: how to measure content marketing ROI

You measure content marketing ROI by tracking what each piece of content costs to produce, what traffic and leads it generates, and how much of that pipeline turns into revenue — then comparing the revenue back to the cost. If that loop isn't closed, you don't have a content strategy; you have a publishing habit. This section covers the measure → decide half of the loop: the numbers that matter, the calculation, the cadence, and the decision rules that turn data into adjustments.

The four numbers that matter (before any vanity metric)

Everything else — impressions, followers, raw pageviews — is context. These four decide whether your content strategy works:

1. 1.Organic sessions per content cluster. Group your articles by topic, not by individual URL. A cluster is the unit of strategy; one post ranking is noise, five posts compounding is signal.
2. 2.Leads attributed to content. Someone who read a guide and then requested an audit, demo, or trial. If you can't tie content to leads, fix attribution before you write another word.
3. 3.Assisted conversions. Most readers don't convert on the first visit. Look at the paths: how often does a content page appear in the journey before a conversion, even when it isn't the last click?
4. 4.Cost per lead from content vs. other channels. This is the ROI calculation in its rawest form, and it's the number a CFO understands.

Vanity metrics aren't useless — a rising impressions curve is an early-warning system — but they never get to make decisions. Decisions belong to the four numbers above.

How to calculate content marketing ROI

The formula is simple. The discipline is in being honest about the inputs.

ROI = (revenue attributed to content − cost of content) ÷ cost of content × 100

A worked example (figures are illustrative, not a claim about any real campaign):

• ✓A cluster of 12 articles costs £6,000 to produce (writer time, editing, design, tooling).
• ✓Over 12 months those articles generate 40 qualified leads, of which 8 become customers worth £1,500 each in first-year revenue.
• ✓Revenue attributed: 8 × £1,500 = £12,000.
• ✓ROI = (£12,000 − £6,000) ÷ £6,000 × 100 = 100%.

Two honest caveats. First, attribute conservatively: count only leads where content demonstrably touched the journey (first-touch or a recorded assist), and state your attribution model in the report so nobody can accuse you of moving goalposts. Second, content compounds — an article written this quarter often earns most of its leads in months 6–18. Judge a cluster at 6 and 12 months, not at 6 weeks, or you'll kill your best assets right before they start paying back.

A measurement cadence that closes the loop

Measurement without a fixed cadence becomes procrastination. Run this rhythm:

• ✓Weekly (15 minutes): indexation, ranking movement on your target long-tail terms, any sudden traffic drops. You're looking for technical regressions, not trends.
• ✓Monthly (1 hour): the four numbers per cluster, plus CTR on your highest-impression pages. High impressions with low CTR means your title and meta description are the problem, not the content.
• ✓Quarterly (half a day): the full ROI calculation per cluster, plus a cannibalisation check — are two of your own pages splitting clicks for the same query? If so, consolidate or differentiate before producing anything new.

The output of every cadence step is a written decision: keep, improve, consolidate, or stop. "Keep an eye on it" is not a decision.

Content marketing strategy checklist: measurement edition

Before you call a content strategy "set up", you should be able to tick every box:

• ✓Every content cluster has one owning page and a defined primary keyword set — no two pages competing for the same query.
• ✓Each piece has a single primary CTA, so attribution knows what a conversion looks like.
• ✓Analytics records content-assisted conversions, not just last-click.
• ✓You know the production cost of each cluster (time × rate, honestly counted).
• ✓The weekly, monthly, and quarterly cadences above are calendared, with an owner.
• ✓Every quarterly review produces at least one keep / improve / consolidate / stop decision per cluster — in writing.
• ✓ROI is reported against a stated attribution model, with 6- and 12-month check-ins for compounding assets.

If you can't tick the attribution and cadence boxes yet, that's where the next month of effort goes — not into more publishing.

What to do when the numbers disappoint

A cluster that misses is data, not failure — as long as you change something specific. In practice, weak numbers trace back to one of four causes:

• ✓The wrong keyword. You ranked, but for a query with no buying intent. Check what queries actually drive impressions; differentiate the page toward intent that matches your offer.
• ✓The right keyword, weak page. Position 8–15 with decent impressions usually means the content is thin or the title doesn't win the click. Improve the page, don't write a new one.
• ✓Traffic without conversion. The page attracts readers but the CTA is buried or the offer mismatches the reader's stage. Fix the path, not the prose.
• ✓Cannibalisation. Two of your pages split the same query and both underperform. Consolidate them into one stronger page and redirect the loser.

Resist the urge to answer every weak cluster with new content. In most content programmes, improving and consolidating what exists outperforms publishing more — because the asset is already indexed, already has some authority, and only needs to be worth ranking.