Marketing as Operating System.

Where marketing stops adding up.

On a Tuesday in early 2026, a marketing director at a €120 million industrial manufacturer in North Rhine-Westphalia opens her quarterly review. She has done nothing wrong. She has hired well, invested in the right tools, kept a strong agency. And yet, three numbers across the table refuse to agree.

The agency reports a blended return on ad spend of 4.7. Her team reports leads up 18 percent year over year. Her finance director, two seats over, reports that customer acquisition cost is up 41 percent and the forecast is missing. All three numbers are defensible. All three are produced by competent people. None of them describe the same business.

She runs fourteen dashboards across six tools. She sits through performance reviews where each channel owner explains why their channel is working. She sits through board meetings where her CFO asks why marketing spend is up and contribution margin is down. The two conversations do not connect, because nothing connects them. There is no place in the company where the marketing system, as a system, actually lives. It exists as people, contracts and reports, stitched together by habit.

This is the central condition of mid-market marketing in 2026. Activity is high. Tools are many. Spend is justified, line by line, channel by channel. But the operating layer that should bind it together, the layer that should make decisions explicable to a CFO, growth steerable across markets, and knowledge transferable across hires, does not exist as a system. It exists as a custom assembly that no one fully owns.

This document is for the people who feel that gap and have not yet found language for it. It describes the gap, names what is breaking, and proposes a different way of building marketing operations, one suited to a market in which capability is cheap, attention to detail is expensive, and the next decade of growth will be won by teams that build infrastructure, not teams that book activity.

Four forces are rewriting the economics of a marketing function.

None of them are new on their own. What is new is that they have arrived together, and that their combined effect is structural, not cyclical. The marketing function that worked in 2019 cannot be patched into a marketing function that works in 2027. It has to be rebuilt.

AI collapses the current marketing system.

That is the story being told. The reality is more specific. AI is taking the existing organisational shape apart, but not by removing the junior layer. It is dissolving the logic of role-by-execution itself. The work that used to define junior roles is now table stakes for any model. The work that used to define middle management is now orchestration across disciplines. The work that used to define leadership is system design, including the decision of what not to automate. The triangle is being replaced by a different shape, with no clean specialist tier in between.

The trap underneath this change is architectural. Pluggable AI platforms, meaning marketing clouds, generative suites and all-in-one CRMs with built-in models, look like the obvious answer. They are reasonable at 100 to 200 € per seat, sometimes free at entry. They are crushing at 500 € or 1,000 € per seat, and by the time pricing escalates the data is locked, the workflows are wired in, and switching costs are prohibitive. The question stops being whether the tool works and becomes whether you can still leave. The marketing functions that outlast this reshaping will be the ones that keep their nodes replaceable: tools as interfaces, data in structures the team owns, AI sitting on top of both rather than underneath.

MarTech sprawl has become its own problem.

The average mid-market marketing team operates around 10 marketing tools, with another 20 to 30 systems across the wider organisation touching marketing data. The realised utilisation of that stack, the percentage of bought capability actually used, has fallen from 58 percent in 2020 to 33 percent in 2025. What was sold as enablement is now overhead.

The CFO is back in the room.

For most of the last decade, marketing budgets were defended by narrative. By 2026 they are defended by numbers, or they are cut. Fifty-eight percent of marketing leaders report rising pressure from CEOs and boards to prove the financial value of marketing. Forty percent of CMOs name attribution and ROI measurement as their single highest performance priority. Three percent of European CMOs can attribute more than half of their budget to a measurable business outcome.

The mismatch is not subtle. Marketing is being asked, often for the first time, to behave like a P&L function. Most marketing functions were not architected to answer that question.

Read the full field note on this

Inhousing is no longer about cost.

Eighty-two percent of brands now operate in-house agency capacity in some form. The first generation of inhousing was about cost and control: pull the agency margin in-house, get faster turnaround. What changed in 2026 is what one internal person can actually do. With sub-agents, a single operator now scopes a campaign, briefs a creative direction, runs a media plan, and reports it back to the board. The headcount logic of the old model no longer holds against what one skilled operator can now cover.

This rewrites the question. Inhousing is no longer a make-or-buy decision against an agency. It is an opportunity to design a marketing OS that scales without proportional headcount, resting on a different mix of human judgement and technical capability than five years ago. Agencies face the inverse pressure: they cannot win on generalist execution any more, because the in-house team has the same tools. They have to specialise into niches deep enough that AI cannot replicate them. Ownership has already moved in-house, but what that ownership is for has not caught up. The old scorecard was outputs and impressions; the one that matters now is outcomes and contribution margin. Most functions have made the first move and not the second.

Capability is cheap. Architecture is rare. The teams that win the next decade will be the ones that built the operating system underneath.

The same mistake, made three different ways.

When marketing teams try to fix the broken middle, they tend to fall into one of three traps. They look like different problems with different remedies, but they are three expressions of the same underlying assumption, that marketing is a set of activities, not a system. Once that assumption is named, the three traps stop being mysterious.

Activity over architecture.

Most marketing functions are run as a to-do list. There are campaigns to ship, posts to schedule, spend to deploy. Reviews are organised by what got done, not by what got decided, and certainly not by why. Activity is visible. Architecture is invisible. Reviewers gravitate toward the visible.

The cost of running marketing as activity rather than architecture is not that nothing happens. Plenty happens. The cost is that nothing accumulates. Each quarter is a fresh assembly of campaigns. The next quarter is another. The architecture that should explain why one campaign performed and another did not, meaning the channel logic, the audience model, the attribution chain, is implicit. It exists in fragments, in the heads of three or four senior people. It is never written down because writing it down feels less urgent than the next campaign.

This is the trap that produces fourteen dashboards that do not agree. The dashboards are not wrong. They are doing what they were built to do, measure activity. They were never built to measure a system, because no one sat down to design the system.

Procurement over construction.

When something does not work, the default response in most marketing functions is to procure a solution. New agency. New tool. New hire. The instinct is reasonable, since marketing teams are usually understaffed and overcommitted, and buying capability is faster than building it.

The trap is that procured capability does not compose. Each agency arrives with its own dashboards, its own attribution logic, its own definition of what a qualified lead is. Each tool arrives with its own data model. Each new hire arrives with their own playbook from their last company. The result is a marketing function that owns many pieces and integrates none of them. Half of the SaaS sits unused. A third of the agencies overlap. The hires churn out before they have stabilised.

Procurement is a fine response to a known requirement. It is a poor response to a system problem. System problems are not solved by adding more components.

People over system.

The third trap is the most expensive, because it pays out slowly. Marketing knowledge, the real working knowledge that decides which channel gets the next euro and why, does not live in documents. It lives in three or four people. When those people leave, the system leaves with them.

Seventy-five percent of first VP-Marketing hires at venture-backed companies fail within eighteen months. Forty-two percent of all CMO hires fail in the same window. The conventional reading is that these are casting problems, the wrong people, hired badly. The structural reading is different: the people are mostly fine. They are stepping into roles where there is no system to step into. Their first six months are spent reverse-engineering the implicit logic of a function that was never made explicit. By the time they understand it, the patience of the board is gone.

This is also why every reorganisation costs three to six months of productivity. Not because the new people are slow, but because the old system was never written down.

All three traps share an assumption: that marketing is a function performed by people, not a system operated by them.

What a marketing studio actually is.

A marketing studio is a temporary co-operator that builds the missing marketing system, or rebuilds the existing one, and hands it over.

The word studio is not a euphemism for agency. It is a different organisational logic. An agency sells hours; a studio sells a working system. A consultancy sells diagnosis; a studio takes responsibility for the build. A freelancer sells capacity; a studio brings its own intellectual property, methods, frameworks and scoring systems, and uses them to construct something that did not previously exist.

The closest analogue is an architecture practice that also runs site supervision. The architects do not just hand over plans. They are present while the building is constructed. They make decisions in the field. They sign off on the building when it is finished. And then they leave, because their role was never to live in the building.

What it is.

A marketing studio takes ownership of a system, not a deliverable. The mandate is not to produce campaign X but to make marketing function Y operate as a steerable system. The studio brings methodology to the work, works inside the live operation rather than from a deck, and treats the eventual handover as the design constraint that shapes everything else.

What it is not.

A studio is sensible in one specific moment: when a system needs to be built or rebuilt. Outside that moment, an agency, a consultancy or an internal team will all be a better fit. This is not a model that competes for permanent residency in a marketing function. It competes for the transition.

This applies as much to the function that has grown into incoherence as to the function that has not yet been built. A mid-market brand discovering that fourteen dashboards no longer agree, and a Series A founder discovering that the next marketing hire has nothing to step into, both face the same architectural question, just from opposite sides.

That self-imposed boundary, temporary by design, is what makes the model work. It changes the incentive of the work itself. A studio that wants to be paid next year has to make sure its system runs without it this year.

What this implies for the larger Worqshop.

The studio is the primary vehicle, but not the only one. Worqshop also develops products that emerge from recurring system patterns: ratio, built from our experience with network agencies and global corporations, for marketers and C-level executives without ad-account access but with questions about the campaigns, and Syn, synthetic target-group testing as a pre-layer to qualitative research. And Worqshop invests selectively as operator-angels, small cheques paired with build sprints, into companies where studio capability and capital reinforce each other. All three vehicles share the same logic: build the operating system underneath.

See the products

How a studio operates inside a live business.

Four principles describe the way a marketing studio works. They are not values in the soft sense. They are operating constraints. Each one rules out a class of behaviour that would otherwise be tempting.

The metric we care about.

The formula is deliberately analogous to ROAS, but with time as the unit, not money. It measures how much manual work one hour in the new system replaces, and it has adoption built in as a constraint, not as a separate qualitative question.

Why adoption sits inside the formula. T_adopted is actual usage time, not theoretical ideal. If the team does not use the system, T_adopted stays close to T_baseline and ROA stays close to 1. A system that was built but never adopted scores the same as no system at all. This is the inversion the model forces: adoption is the denominator, not a downstream KPI.

The same logic applies where capital is involved, not just time. When Worqshop invests as operator-angel, the question is not how fast revenue grows, it is how fast the operating system the team built becomes load-bearing without external input. Adoption is the leading indicator of system change in every vehicle, not just the studio.

Four phases. Fixed scope per phase. Ownership transferred at each one.

The principles describe the posture. The method describes the sequence. Every service runs through the same four phases, in the same order. Each phase has a fixed budget, a fixed deliverable and a defined exit point. You own what comes out of each phase regardless of whether the next phase is commissioned.

The shape of the method is deliberate. Most consulting engagements start in the deck and end before contact with reality. Most agency engagements start in execution and never produce a written architecture. The method places the system design in the middle, with operational work on both sides, built up to it and run out from it.

We start where most engagements skip: inside your live operation. We end where most engagements try to stay: at the door, with everything documented.

The same method, applied to five different system problems.

The Worqshop Method is a single architecture. It produces different named patterns depending on where the work begins and what kind of system needs to be built. Five named patterns recur.

The Inhousing Path

Global travel retail group, multi-brand portfolio

Migration of a fragmented agency-led media operation into an internal Media Operations function. Phase 0 mapped agency dependencies, Phase 1 designed the in-house architecture, Phase 2 ran media in parallel, Phase 3 transferred ownership. Outcome: blended ROAS up 39 percent, agency cost reduced, time to internal ownership nine months.

Cold Start in 90

From zero infrastructure to validated launch

Build of a complete marketing system from zero infrastructure to validated launch within about 90 days. Phase 0 starts with C-level shadowing of the founders, Phase 1 designs the media system, Phase 2 runs smoke tests and pre-warming, Phase 3 closes with a measurement framework the board could read.

OS Architecture

Berlin production studio, about 18 people, three internal divisions

Reconstruction of the operating system for an entire media organisation, not just marketing. 18 tools in use, eleven documented processes, six systems in the finance flow alone, manually stitched. Our mandate: design the operating system, recommend the stack, supervise rollout, build the technical instruments needed.

Phase 0 mapped people and tools with the co-founder and strategic lead. Phase 1 ran five leadership shadowings, coded 24 friction points weighted by people, process and tool reach, and produced a blueprint with thirteen design principles and a prioritised lever list. Phase 2 supervised rollout. Four domains under one roof: project steering, finance as its own foundation, knowledge and standards, onboarding and access.

Success measured per lever, not per project: hours before against hours after, evaluated at four, eight and twelve weeks. Below factor 2, smaller tweaks would have done the same. Above factor 5, this is substitution rather than optimisation.

The same pattern now carries the conversational layer. That engagement rebuilt an operating system and supervised the rollout. The current work is narrower: a single conversational interface over the stack a company already runs, on the client's own infrastructure, with Skills that carry their processes and terms into every conversation, so the model speaks their CRM and not a generic one. Tools are not replaced and data is not migrated. Measurement stays where it was, in ROA, which is also where this pattern is most exposed. An interface nobody talks to scores the same as no interface. The architectural questions are unchanged: what the system should look like, who owns it, how it runs without us.

Service Productisation

Internal capability turned product

Conversion of an internal performance capability into a sellable, fixed-scope product. Typical for agencies and in-house teams sitting on real expertise that cannot be sold without a name, a price and a defensible delivery shape. Phase 0 shadows how the work is actually done, Phase 1 designs the framework covering scope boundaries, pricing logic, delivery playbook and sales argumentation, Phase 2 pilots with a real client, Phase 3 closes with a productisation playbook the team can extend without us.

Embedded Operator

Leading EU medical cannabis producer, regulated category, and a leading German digital agency, DACH

Temporary senior-operator responsibility at C-level or senior-operator level, with a defined end. The format for situations where a function needs an operator presence rather than a system rebuild: interim heads, fractional C-level, fractional senior operator seats inside an existing team, interim paid media capacity with its own lead and plan, focused operator sprints, funnel recoveries, measurement work without a full system rebuild.

The end condition is written down before the first month is invoiced: the internal role gets filled, the permanent team arrives, or the structural problem gets scoped as its own mandate. That is what separates this from a staffing arrangement. In the agency engagement it took twelve monthly extensions to reach it, and the handover happened when the internal team was refilled. Same posture as the other patterns, smaller surface, same handover discipline.

In the regulated-category engagement, an always-on performance-led creative testing model produced a registration rate up 8 percent. In the agency engagement, senior operator responsibility carried a major-account performance mandate with a defined handover.

Each of these is the same method. The pattern names exist so that conversations about new work have a reference point, so that a CMO can say "we need a Cold Start" or a CFO can say "we need an Inhousing Path," and both sides know what is being commissioned.

Browse the services

Three ways to start. None of them is a sales process.

If anything in this document maps to a problem you are sitting on right now, there are three entry points. Each one is designed to be useful even if it goes nowhere.

A note on what we are not selling.

We do not sell open-ended retainers. Every service carries an end condition, written down before the first invoice: the internal role gets filled, the permanent team arrives, or the structural fix gets scoped as its own mandate. Paid media included. We will run your channels while you are short-handed, and we will hand them back documented. What we will not do is become your permanent external media function, or produce content at scale. If that is what you need, an agency is the better structure, and we will name one.

We are the right partner for one specific moment: the moment a marketing function needs to be rebuilt as a system, because growth has stalled at the architecture layer, because a new owner has arrived, because a CFO is asking questions the function cannot answer, or because the next hire is about to walk into a job where there is no system to walk into.

If that is the moment, we are useful. If it is not, we are an expensive distraction, and we will say so.