The agency growth stack: how thought leadership, trade press, AI search and awards compound into pipeline

Most agencies treat marketing the way they treat new business pitches: in spikes. A burst of LinkedIn posts when the founder remembers. A press release when there is a win. An awards submission when the deadline arrives. A site refresh every three years when the existing one starts to embarrass the team. The result is that the agency is famous for one week at a time and invisible for the other fifty-one.

What works instead is a stack. Four visibility surfaces, each with a different mechanic and a different decay curve, sharing one editorial line. The surfaces feed each other. The line is the founder's actual opinion about how their corner of the industry works.

This is the operating system. It is also, openly, what They Said sells. We are describing it here because the principle does not stop working when you describe it, and because most agencies will read this, agree, and still not implement it - the way most agencies read about awards submissions and still wait until two weeks before the deadline.

The four surfaces

One. Founder LinkedIn. The fastest surface. A post lives for forty-eight hours and dies. The signal it gives a buyer is "this person has opinions about the work I am paying for". The cadence is two posts per week from the founder, ideally with one comment thread that turns into its own piece by the end of the week. See agency-linkedin-cadence for the structure.

Two. Trade press. The slowest surface. A byline in Campaign, PRWeek or The Drum lives indefinitely. The signal it gives a buyer is "this person is a known authority - the kind editors call for context". The realistic cadence is one byline per quarter from a senior person, with two or three expert-comment placements per month. See how-long-to-get-into-campaign for the timeline.

Three. AI search. The newest surface. AI engines cite agencies whose names already appear, in the right contexts, across enough third-party publications. The signal it gives a buyer is "when I asked ChatGPT for agencies in this space, this one came up". The cadence is invisible day-to-day but compounds quarterly. See ai-entity-audit for what to fix first.

Four. Industry awards. The deepest surface. An award win lives for years, becomes a sentence in every pitch deck, and gives every other surface something to reference. The signal it gives a buyer is "this agency does work judged best in class by people whose opinions matter". The cadence is two to four entries per year, written to jury standard, in categories chosen on purpose. See category-strategy-awards and awards-coverage-vs-juries.

Why they compound

Each surface produces an asset the next surface can use.

A founder LinkedIn post that earns engagement becomes a paragraph in a trade-press byline. The byline becomes a quote in an industry roundup. The roundup gets cited by an LLM. The LLM citation surfaces the founder's name when a buyer asks for "agencies that do X in London". The buyer reads the LinkedIn post. The loop closes.

That loop is invisible to the agency. It is visible to the buyer, who arrives at the website having read three of those touch-points without remembering which one came first.

The compounding only happens if all four surfaces use the same editorial line. If LinkedIn says one thing, the trade-press byline says another, and the awards entry positions a third, the buyer sees a confused agency. The model sees three different entities. The award jury sees a fourth. Nothing compounds.

The single editorial line

Every agency we have worked with has a real opinion about their corner of the industry. Most have buried it. The exercise is to find it and name it.

The test: write three sentences that contain a specific opinion, a specific reason and a specific consequence. Read them aloud. If you cannot defend the opinion against a sceptic, the opinion is not sharp enough yet.

For They Said, the line is: Most agencies have brilliant opinions that never leave the building. We are the company built to get them out. Everything else - the four surfaces, the cadence, the formats - is structure around that line.

If your agency does not have a line, the four surfaces will compete for content instead of reinforcing each other, and your output will look like the LinkedIn-press-AI-awards mush of every other agency.

The cadence that holds it together

This is what a working stack looks like as a calendar.

1. Monday: a take. A short reactive post, on the website and LinkedIn, reacting to a signal from the previous week. Trade press call this an op-ed; an agency can call it a take. Either way, it is one opinion with one peg.
2. Wednesday: an answer. A question-formatted piece for AI search. The first sentence directly answers the question. The rest is mechanic. Buyers find it through ChatGPT, Perplexity, or Google's AI overview, and the agency starts being named as a source.
3. Friday: a cornerstone. A 1500-to-2500-word piece on one of the four pillars of the agency's editorial line. Slow, evergreen, internally linked. The hub the rest of the week's content connects to.
4. Every two weeks: a working session. Forty minutes with a senior person inside the agency, recorded, transcribed, mined for the next week's takes. See fortnightly-working-session.
5. Monthly: an AI snapshot. Run twenty buyer queries through ChatGPT, Claude, Perplexity, Gemini. Log mention rate. Watch it move quarter over quarter.
6. Quarterly: a trade-press cluster. One byline, three to five expert comments. The cluster is what trade editors register as "this agency keeps showing up". See wins-week-reading for how editors process the weekly trade-press flow.
7. Annually: two to four award entries. Each written by the same person who has been writing the takes, the answers and the cornerstones. The voice is the same. The category strategy comes from the editorial line. See campaign-grade-pitch for the pitching standard.

What this is not

This is not a content-marketing programme. Content marketing teaches you to write blog posts your buyers might Google. The stack assumes the buyer is not Googling - they are reading trade press, asking ChatGPT, scrolling LinkedIn, and watching which agencies the award juries promote.

It is not a PR retainer. A PR retainer pitches news. The stack pitches an opinion that has been pre-validated across four channels before a publication even hears about it. The PR pitch closes the loop, not opens it.

It is not a thought-leadership package. Most thought-leadership packages produce a single white paper that nobody reads. The stack produces continuous output across four surfaces with one voice, which is the only way thought leadership does what it is meant to.

The honest measurement

The metric most agencies use for marketing - leads per month - is the wrong metric for the stack. The stack does not produce leads at month two. It produces a small number of disproportionately good leads from month six onwards, because the buyers who arrive have already absorbed four touch-points and arrive pre-convinced.

The metric to watch in the first six months is reply rate. If your bylines are getting replies from inside target buyers, the stack is working. The leads arrive on their own clock.

By month twelve, the unfair advantage compounds. The agency starts getting cited in trade press it did not pitch. Award juries pre-recognise the name. Buyers describe the agency in language the founder wrote, which is the cleanest signal that the editorial line has reached the people who hire.

Three common ways the stack fails

When agencies set this up and it does not work, the failure is almost always one of these three.

One. The line is fashionable, not defensible. "We do brand-led performance" reads well in a deck and means nothing on a call. The opinion has to be specific enough that a competitor could write the opposing opinion and not look stupid. If both versions are defensible, you have a real line. If only yours is defensible, you have a slogan.

Two. Two surfaces are louder than the other two. The most common shape is LinkedIn and awards on, trade press and AI search off. The result is high domestic visibility - your team feels the agency is famous - and no movement in the surfaces that actually drive buyer recall. The fix is not to do more of the loud surfaces. It is to fund the quiet ones for two consecutive quarters.

Three. The output sounds like the agency, not like the founder. A stack written by a marketing team for an agency brand is harder to find a voice in. A stack written through one named human, whose opinion is sometimes wrong on purpose, is what readers remember. Buyers do not hire brands. They hire the person they read.

Where to start

If you are starting from scratch, do these things in this order:

1. Name the editorial line in three sentences. Defend them aloud against a sceptic.
2. Audit the four surfaces against the line. Where are they consistent? Where are they not? Fix the inconsistencies before you add anything new.
3. Pick one surface to invest in first. We usually recommend trade press, because it is the slowest to start and the hardest to fake.
4. Add the second surface sixty days later. Add the third another sixty days after that.

The reason this takes a year is that the surfaces have to settle. Trade-press relationships take quarters to build. AI citations follow trade-press citations by months. Awards require category positioning that takes a full cycle to test.

The reason it is worth the year is that the compounding does not stop. The agency that has spent twelve months on the stack has a year of compounding ahead of every agency that has not started.

If you want to see the engine assembled for your agency, get in touch. We do this for a living.