Case studies.

A European supply chain group needed to open two markets where nobody knew the category existed, then hold one coherent message across four countries without flattening any of them. The local model built for the smallest market became the standard for the region.

The challenge

The group was strong operationally and close to invisible commercially in the markets it wanted next.

In two of them the category itself was unfamiliar. Buyers were not choosing a competitor. They did not know the service model existed, which is a harder problem than competition and much harder to brief. No search demand to capture, because nobody was searching.

In the two established markets the problem inverted. One faced long-standing alternatives and had gone years without consistent marketing presence, so the brand had slipped quietly out of the consideration set. The other had awareness without positioning.

Four markets, four different problems, one group narrative that had to survive all of them.

What the problem actually was

The brief described a visibility problem. It was not one.

Visibility assumes the market already knows what it is looking at. Running awareness campaigns into a category nobody understood would have bought impressions and taught nobody anything. The real problem was category education under commercial pressure: teach a market what the service is, and prove commercial demand, at the same time, before anyone funds the years of presence that category building normally takes.

That reframe changed the engagement. We did not propose a campaign. We proposed an experiment designed to produce evidence a board could act on, and we said in advance what result would justify stopping.

The second diagnosis mattered more over the year. Four local teams held real market knowledge and had no shared architecture for using it. Every country was solving the same problems separately. The scarce asset was never content. It was a way for four teams to stay coherent without any of them being overruled.

What we did first

A short validation pilot in the two unfamiliar markets, on a deliberately modest budget.

It was scoped as a test rather than a performance campaign, and we were explicit about that with the client. The question was not how many leads can we generate. It was whether the market responds at all, who responds, and what it costs to find out.

One message was tested identically in both markets, so the variable was the market and not the copy. Each platform was given a distinct role so we could see which part of the funnel it was actually doing. And there was a hard stop, so a test could not quietly become a permanent budget line.

What that proved

The markets responded, and they responded almost identically.

Two countries, different languages, different competitive conditions, different maturity, and category search presence within a point of each other inside sixty days. That consistency was the finding, not the level. It suggested the message was doing the work rather than local circumstance, which is precisely what you need to know before committing budget to two more countries.

What it did not prove. A short pilot cannot model a pipeline, the sample was too small for meaningful attribution, and one channel could not be defended on its own numbers. We put all of that in the report. A pilot that produces only good news has not been designed properly.

Scaling it

Phase two went from two markets to four, and from testing to building.

At the centre was a messaging framework run at roughly 60% shared, 40% local. The shared part carried the group’s positioning, the category argument and the non-negotiables. The local part belonged to each country: examples, references, terminology, tone, and the freedom to say a thing differently because the market hears it differently.

That ratio was a decision, not a compromise. Push shared much above it and local teams stop recognising themselves in the output, then quietly stop using it. Drop below it and within six months you have four brands.

On top of that ran a weekly publishing rhythm per market, regular newsletters, search-led articles, and pillar content built for long-term authority rather than short-term traffic.

Working with their team

This is the part most agencies leave out, and it decided whether any of it worked.

We met more than thirty of their colleagues over the year, in person and online: country directors, local marketing, operations, technical experts across all four markets. Not onboarding. The method.

We held strategy and coordination as the central point, content production, search and analytics, social management, newsletters, paid architecture and the editorial calendar. Their central team held final sign-off on messaging, topic priorities and alignment with group positioning. Their local teams held the things we could not: market insight, validation of language and tone, local examples, and the final go-ahead to publish in their own market.

That last one matters. Local teams had a veto, not a review. It slowed publishing down, and it is the only reason the programme did not collapse into four polite translations of a single head-office plan.

The model that travelled

The most useful thing we learned all year came out of the smallest market, and it was not a strategy. It was a format.

We filmed the client’s own people, at their own site, talking about their own operations. No studio, no script, local language, real faces. It was built to solve a narrow problem: prospects in that market were not certain the group had a genuine local presence, and the sales team was spending the first ten minutes of every call proving it did.

It outperformed produced static work by a wide margin, and cost less to make. The engagement rate was not the point though. The commercial effect was. Sales stopped having to argue that the local operation existed, because prospects had already seen it. Verifiable proof replaced an assertion, and that removed friction from the top of every conversation.

The format is now applied across the region. It began as a fix for one market’s credibility gap and became the standard, which is the most durable kind of result: the work outlived the campaign that produced it.

What we would tell you honestly

Lead flow is still not predictable across all four countries. We can generate leads. We cannot yet promise a number per month per market, and anyone promising that during a category build is guessing.

Attribution is the real gap. Limited funnel reporting and no automated nurturing for most of the period. We can prove attention. Proving pipeline needs plumbing that is still being built.

Budget continuity hurt the first phase. Fragmented allocation meant campaigns stopped before they could compound, and the clearest lesson of the year is that consistency of spend mattered more than size of spend.

Sales and marketing alignment still varies by country. The relationships are good. Shared processes and common definitions are not there yet.

And four markets at four maturity levels is genuinely hard. We underestimated the coordination cost in the first quarter of the expansion.

How we work

Execution has been commoditised. AI writes, designs and ships at close to zero marginal cost. What has not commoditised is the decision underneath: which market to enter, what the actual problem is when the brief says something else, and which local truth deserves to override the central plan.

Every one of those is made under real uncertainty, and AI does not reduce that uncertainty. It reduces the feeling of it, by returning a confident answer to a question whose honest answer is a range.

So the scarce skill is not producing more. It is judgement. Knowing which outcome you could not survive, which number in the plan is a range in disguise, and when the person in the smallest market is right and the central strategy is wrong.

That is what business knowledge actually is, and it is why we start with commercial reality rather than the creative brief. A good model cannot rescue a bad strategy. It only gets you lost faster, and it does it fluently.

A residential developer with fifteen years of delivery behind it was about to launch a property into a category where nobody was searching yet, while four other projects still needed to sell. The launch scaled more than ninefold in a quarter without the established projects being harvested to pay for it.

The challenge

The developer had a real track record. Fifteen years, a portfolio of delivered projects, thousands of apartments completed, energy efficiency as a standard rather than a selling point. Operationally they had nothing to apologise for.

The problem was that five active projects were being communicated as five separate things, to five different audiences, in a market that had just turned cautious. High costs, fiscal uncertainty, buyers taking longer. And the newest asset was not another apartment block. It was a resort, in a coastal market, aimed at people who did not yet know they wanted it.

So the brief contained two jobs that pull against each other. Launch something into a category with no existing search demand. And do not let the projects currently paying the bills go quiet while you do it.

What the problem actually was

The brief read as a media problem. It was a search demand problem, and the two need different money.

For the established capital-city project, demand already existed. People were searching for apartments in that city, and the work was to capture that intent efficiently and keep the cost per enquiry stable.

For the new coastal launch, demand did not exist in any usable volume. Nobody was searching for the thing by name because the thing had no name yet. Buying media into that is how budgets disappear.

That distinction set the whole first month. Before any campaign ran we built a dedicated keyword architecture for the launch: forty-eight terms, structured into four groups by intent, separating branded search from location search from investment intent from generic category search. Not to capture demand. To find out which door people would come through once demand existed.

The second diagnosis was about the portfolio. Five projects with one developer behind them were competing for attention against each other. The developer brand itself, the thing that should have carried credibility across all five, was the least communicated asset in the group.

What we did first

Two things, in the same month, before spending meaningfully.

A full audit of existing campaign performance and a rebuild of the keyword sets by intent and match type. And a complete launch playbook for the new property: the central message and concept, the full search structure, the social structure split by audience rather than by format, two named audience archetypes, and the creative brief that everything downstream would be built from.

That playbook mattered more than any single campaign. It meant that when the launch went live in month two, nobody was inventing the message in the ad copy.

What that proved

The launch scaled every month.

Month one was deliberately small, a live test of the architecture rather than a launch. Month two put it in market properly. Month three was the strongest month of the quarter, at more than nine times the enquiry volume of the first.

The curve is the point rather than the level. A launch that grows month on month is being tuned against real search behaviour. A launch that spikes and settles was bought.

Meanwhile the established project did the other job.

Budget was reallocated daily toward whatever was working, which in practice meant toward the launch. But the established project was held at a steady cost per result rather than being drained to fund the new one. That was a deliberate constraint, and it is the part of this engagement we would defend hardest. Cannibalising a performing asset to make a launch look good is the easiest way to produce a great quarterly report and a worse year.

What it did not prove. One quarter is not a sales cycle in residential property. Enquiries are not reservations and reservations are not completions. We can show that interest was created and that it grew. Attribution through to signed contracts sits with the developer’s own sales data, and connecting those two systems properly is still ahead of us.

Scaling it

Once the architecture held, the work widened rather than deepened.

An always-on awareness layer ran underneath every performance campaign, across search, social and video, so that the projects were never invisible between campaign bursts. Around one hundred and seventy-five creative variants were produced across the five projects in four months, in every format the platforms take, because a single execution fatigues long before the message does.

And the developer brand finally got its own campaign. Not a project campaign. A campaign about the company behind the projects: the years, the delivered portfolio, the energy standard, the integrated teams from architecture through to handover. In a cautious market, the most useful thing a developer can prove is that it will still be there at completion.

Working with their team

The rhythm was daily, weekly and monthly, and it was the same every month.

Daily: delivery monitoring, budget redistribution and bid adjustments across live campaigns. Weekly: creative production, format variants, and maintenance of the master campaign file. Monthly: reporting, new campaign concepts, and coordination with the client’s team and their internal production people.

Approvals ran through a small, named group on the client side, with their internal producer working alongside our creative output rather than after it. Every creative deliverable shipped with a live view link to the working file, so nobody had to ask what the current version was.

That last detail sounds administrative. It is the reason four months of work across five projects did not produce a version-control argument.

The model that travelled

The most transferable thing was not a campaign. It was treating reputation as a channel with its own workload.

In the fourth month the effort shifted deliberately from campaigns to positioning. We wrote and placed a full editorial interview with the chief executive in a sector trade publication, coordinated end to end with the editorial team, and secured it in the annual editorial plan without paying for placement. The seven themes were chosen to answer what the market was actually anxious about: pricing, cost pressure, regulatory change, and whether developers would deliver what they promised.

In parallel we structured five award submissions for the resort, including three individual categories. That meant building a narrative for each executive from raw material: taking thirty years of hospitality experience, or a chef’s international record, and turning it into a case a jury could act on.

Neither of those is media buying. Both are the same underlying job as the campaigns: making a credible operator legible to people who have not met them. In a market where buyers are asking whether a developer will still exist at handover, third-party endorsement does work that no amount of paid reach can.

What we would tell you honestly

The connection between enquiries and sales is still not closed. We can prove interest and its growth. Tying that to signed contracts needs the developer’s sales data joined to the campaign data, and that work is outstanding.

One quarter is a short window in a category where buying decisions take months. The launch curve is genuinely strong and it is still three data points.

Creative volume was high, and not all of it earned its production cost. Around a hundred and seventy-five variants across five projects in four months is a lot of output, and a tighter test-and-kill discipline would have got to the winners faster with less made.

Awards nominations are not awards. We coordinated the submissions and built the narratives. The outcome was not ours to control, and any agency presenting a nomination as a result is overstating.

And a portfolio of five projects at five different stages is genuinely difficult to keep coherent. Some months the developer brand campaign got less attention than the projects that needed to sell that quarter, which is commercially reasonable and strategically a cost.

How we work

Execution has been commoditised. AI writes, designs and ships at close to zero marginal cost. What has not commoditised is the decision underneath: whether the problem is media or demand, which asset to protect while another one scales, and what a cautious market actually needs to hear before it commits.

Every one of those is made under real uncertainty, and AI does not reduce that uncertainty. It reduces the feeling of it, by returning a confident answer to a question whose honest answer is a range.

So the scarce skill is not producing more. It is judgement. Knowing which campaign is worth starving, which month of a launch is a signal and which is noise, and when the right move is to stop buying attention and go and earn it in print instead.

That is what business knowledge actually is, and it is why we start with commercial reality rather than the creative brief. A good model cannot rescue a bad strategy. It only gets you lost faster, and it does it fluently.

A specialist physician had spent years building the case that medicine works better when nobody sits between the doctor and the patient. She had the audience, the authority and the argument. What she did not have was the machine. Adopters have grown around 50% a month for three consecutive months, with no paid acquisition behind it.

The mission

Somewhere between a doctor and a patient, a third party appeared. It decides how long the appointment is, which treatment is permitted, how much of it, and when. Neither of the two people in the room chose this arrangement and neither can easily leave it.

The client’s argument is simple and, in the profession, quietly radical: the physician’s only obligation is to the patient. Not to an insurer. Not to a hospital. Not to a quota.

There is already a way to practise like that. Specialists can work directly with the people they treat, price transparently, and spend their time on medicine instead of prior authorisation. What there was not, was an institution: somewhere to find those physicians, learn how to become one, and belong to something once you had.

That is the work. Not marketing a membership. Turning a conviction that already had believers into something with a front door.

The problem: an audience is not an institution

The founder came to us with something most clients never have. Years of showing up had produced genuine reach, real authority in her specialty, and an audience that trusted her.

She also had the clearest possible diagnosis of her own situation, and she said it out loud in the first conversations: building an audience and building a business are two different things.

Everything on the left of that gap had taken years and was not in doubt. Everything on the right had to exist, work together, and keep working on the days she is seeing patients. Membership tiers that make sense and do not overlap. A directory that is verified, because on a medical platform an unverified listing is a liability rather than a lead. Payments and renewals. Journeys that run without anyone pressing send. A content system. Presence in search and in AI answers, because that is increasingly where a physician’s first question goes. Press. Speaking. Partnerships.

Each of those is a discipline. The reason we were chosen was not that we do any one of them best. It was that all of them had to be decided together by someone holding the commercial picture, and that is a strategy problem before it is an execution problem.

Pick the membership tiers without knowing the content plan and you promise things you cannot deliver. Build the automation before the category argument and you send beautifully sequenced emails about something nobody understands yet. Optimise for search before the directory is verified and you drive physicians to a page that damages trust.

Why we refused the obvious plan

The obvious plan was activation. There was a warm list, a founding offer, and an obvious set of deadlines and reminders to run against it.

We wrote a version of that. Then we threw it out.

The problem with squeezing a warm list is that it works, briefly, and then it stops, because nothing in it is increasing the number of physicians who know the category exists at all. You get a good quarter and no movement.

So the plan was rebuilt on two speeds running at once, the long and the short. Activation converts the physicians already looking. Category building grows the population who will look later. In a category this under-marketed, being the loudest and most consistent voice is available cheaply, and that share of attention is the actual growth engine.

Two decisions inside that are worth naming, because they are the strategic mindset in practice.

We mapped moments, not personas. The useful question is not who the ideal member is. It is: at what real moment does a specialist first think there has to be another way? Charting at midnight with the family asleep. Another prior authorisation denied. A reimbursement cut that makes the arithmetic impossible. Watching a colleague leave and land well. Every piece of content is built against one of those moments, so the platform surfaces when the thought occurs rather than when we happen to be publishing.

We locked the assets and stopped being clever. One belief line, one rhetorical device, one shape of invitation, the founders’ actual faces, the same visual codes everywhere, every time. With a subject this rich the temptation is a fresh angle every week. Fresh angles build nothing. Repetition builds memory, and memory is what makes a category thinkable.

What that produced

Three consecutive months of adopter growth at roughly 50% month on month, with no paid acquisition behind it. The base has more than doubled.

The rate matters less than the fact it compounds and that none of it is rented. Paid growth stops the day the budget does. Growth that comes from physicians finding the category, recognising themselves in it, and telling colleagues keeps running, and it gets cheaper rather than more expensive as it goes.

That is what category building buys. Not a spike. A base.

Getting the category into the room

Publishing is only half of making something legible. The rest is being where the profession already gathers.

That meant a speaking presence at a sector festival, with the material and outreach built around it rather than bolted on afterwards. Member features published as full profiles with structured data, so individual physicians surface in search and in AI answers and the platform surfaces alongside them. An approach to a national provider as a partner in the argument rather than as an advertiser. And commentary that attached the platform to a live industry debate started by a well-known business figure, using the substance of the argument and never the person.

None of that is media buying. All of it is the same job as the content: making a small, credible organisation legible to people who have not met it yet.

Working with a founder who is also a doctor

She runs a clinical practice alongside this. That single fact designed the whole operating model.

Everything is built to be reviewed in short bursts and to survive her being unreachable for days. Work arrives in batches with the decisions already made and the open questions narrowed to the ones only she can answer. Nothing ships without her clinical and factual sign-off, because on a medical platform the founder is not an approver, she is the fact-check.

The best example of that came from her side, not ours. She caught that a set of our materials was offering founding members benefits that belonged to a higher tier. We had propagated it across a roadmap, several guides and a set of emails. She spotted it, we corrected it everywhere, and the rule was locked so it could not recur.

That is what a good client relationship looks like from the inside. Not frictionless. Checked.

What we would tell you honestly

The first plan was ours, and it was wrong. We wrote an activation-led go-to-market and replaced it weeks later when it became clear it would harvest and stall. That is the process working. It is also weeks.

The founder is the bottleneck by design. Everything routes through one clinically qualified person. That protects accuracy and it limits pace, and in medicine that trade is correct. It is still a constraint, and anyone promising you speed in this category has not understood it.

Attribution is incomplete. We can see the growth, the engagement and what content earns attention. Connecting that cleanly through to revenue across the platform and payment systems is still being built.

Three months is three data points. The rate is real and it has held every month, but compounding percentages flatter small bases and a quarter is not a trend. The rate will come down as the numbers rise. The question worth asking in a year is not the percentage, it is whether the base still grows without paid support.

Category building is slow before it is fast. The compounding only started once there was enough consistent presence for it to compound on. Anyone wanting a quarterly result should not start here.

How we work

Execution has been commoditised. AI writes, designs and ships at close to zero marginal cost. What has not commoditised is the decision underneath: whether to convert the demand in front of you or build the demand that does not exist yet, which promises the membership can actually keep, and what a sceptical, regulated profession needs to see before it trusts anything.

Those decisions cannot be made one discipline at a time. The tier structure is a content decision. The directory is a search decision. The category argument is a product decision. Somebody has to hold all of it at once with the commercial reality in view, and that is the work we were hired for.

Every one of those choices is made under real uncertainty, and AI does not reduce that uncertainty. It reduces the feeling of it, by returning a confident answer to a question whose honest answer is a range.

So the scarce skill is not producing more. It is judgement. Knowing when your own plan is the thing that needs replacing.

A good model cannot rescue a bad strategy. It only gets you lost faster, and it does it fluently.