Frameworks & methodology: how CMOs structure agency relationships that actually deliver

Your roster review lands next quarter and procurement wants one page with a recommendation on it. There are only three structural answers, and they are not variations on a theme: consolidate the work onto a small number of holding company partners, commission a bespoke unit staffed for your business alone, or put the payroll on your own books and run an in-house studio. Most CMOs choose by anecdote, usually whichever model the last conference speaker praised. The choice is mostly arithmetic with two or three judgement calls on top, and unwinding it costs eighteen months, so the reasoning is worth making explicit.

The three structures, and what each one buys

Take as read the roster shapes, scope types and fee models the foundations lesson sets out. What changes here is where the work physically sits and who carries the fixed cost when volume drops.

A consolidated roster keeps everything external and variable. You shrink the number of counterparties, buy on volume, and reduce the number of briefing interfaces your team maintains. A bespoke dedicated unit is still external, but the holding company assembles a named team that works only on your business, usually priced as an annual operating cost rather than project by project. An in-house studio moves the capability and the cost onto your P&L as headcount.

Each one buys a different thing. Consolidation buys price and coherence. A dedicated unit buys institutional memory and turnaround. An in-house studio buys control of high-volume, low-variance output. None of them buys all three, and the structures that fail are usually the ones bought for a benefit they were never going to deliver.

Sub-concept 1: the consolidated roster

The cheapest structural move, because it changes the contract rather than the org chart. Unilever announced in 2017 that it would halve the number of agencies it worked with and cut the number of ads it produced by around 30 percent. Fewer partners, fewer duplicated strategy decks, one lead per brand cluster instead of four.

The cost is one most CMOs discover at renewal: you have destroyed your comparator pricing. When the master services agreement comes up and you have three partners instead of forty, you have no live market price for a 90-second film or a year of always-on social. The counter-move is to keep one small independent on a real retainer, roughly 5 to 10 percent of external creative spend, working on genuine briefs. It is not a hedge against creative failure. It is a price reference you can quote in a negotiation, and it stays honest because the work is real.

Second failure mode: consolidation savings are banked in year one and quietly reverse by year three. Scope creeps back, out-of-scope requests get absorbed, and the fee that looked like a 20 percent reduction is carrying 30 percent more work. Re-baseline scope annually against actual delivered volume, not against last year's contract.

Sub-concept 2: the bespoke dedicated unit

CocaCocaCustomer Acquisition Cost: total sales and marketing spend divided by the number of new customers acquired over the same period.View full definition →-Cola consolidated its global creative business with WPP in December 2021, an account reported at around 4 billion dollars of annual marketing spend, and WPP stood up a dedicated cross-agency team for it rather than routing the work through existing agency brands. Manolo Arroyo, then global CMO, had already been cutting a long tail of agency relationships that ran into the thousands.

A dedicated unit works when your business is complex enough that ramp-up time is the real cost. People who have been on your brand for three years do not need re-briefing on why a particular claim cannot run in Germany. That knowledge is worth more than the hourly rate difference.

Two things go wrong. The first is quiet attrition: the senior names in the pitch rotate off within a year and you are paying dedicated-team rates for a team that is dedicated but no longer senior. Write key-person clauses covering a named core of eight to twelve roles, require a quarterly attrition report on the unit, and put a slice of fee at risk against it. The second is subtler. After eighteen months the unit has absorbed your internal culture and stops arguing with you. You have bought a very expensive extension of your own department. Buy dissent back deliberately: one annual project briefed to an outside shop, or an external creative audit with a named reviewer who does not depend on the account.

Sub-concept 3: the in-house studio

Lego runs a substantial in-house agency across several hubs, handling the volume of product, packaging, retail and social output that would be ruinous to buy by the asset. It did not hand the in-house team its brand platform. "Rebuild the World" was created by BETC Paris and launched in 2019. That split is the model working as designed, not a contradiction of it.

Unilever built its U-Studio network with Oliver, a company whose business is building and staffing in-house studios for clients, so read its published savings claims with that in mind.

The arithmetic is the decision. A twelve-person studio at a fully loaded 120,000 per head is 1.44 million of fixed cost before a single asset ships, plus tooling and management time. Divide that by the number of assets you genuinely produce every year, not the number you wish you produced, and compare it to the per-asset production costs the budget negotiation lesson prices out. In-house beats market rates at high utilisation, somewhere north of 70 percent, and loses badly below it.

Which is why seasonality is the edge case that kills studios. A business with a Q4 concentration idles a permanent team for five months and then blows through its capacity in October anyway, buying the overflow externally at spot rates. Two other things do not in-house well: markets where local language, local casting and local regulatory clearance are the actual work (alcohol, pharma, financial promotions), and any category where your own team becomes the only critic of its own output.

Sub-concept 4: the selection criteria

Score the work, not the company. Five criteria, each on a simple high/medium/low:

• Annual volume and its predictability. High and predictable pushes toward in-house. Lumpy pushes toward external and variable.
• Range of distinct creative problems per year. Wide range needs a roster or a unit with access to specialists; narrow range is studio work.
• Speed requirement measured in hours, not days. Under 24 hours consistently, you need people you can walk to.
• Sensitivity of the data or IP involved. First-party dataFirst-party dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.View full definition → modelling and unreleased product tend to stay inside.
• Fixed-cost tolerance. Ask the CFO how a 15 percent marketing budget cut would land. External scopes pause in a quarter; headcount does not.

Most large advertisers end up hybrid because the scores differ by tier of work. Tier the portfolio first (brand platform, campaign adaptation, always-on volume), then assign a structure to each tier. Coca-Cola runs a dedicated holdco unit and its own Studio X content operation at the same time. The mistake is picking one structure for everything and then complaining that it is bad at the work it was never built for. The organisational-scale version of this arbitration, including what capability you permanently lose and what it costs to reverse, is the playbook lesson's territory.

Real-world cases with results

Unilever made both moves at once: halved the roster and built studio capacity, so the savings from consolidation and the savings from in-housing were reported together. Worth separating when you model your own case, because they behave differently under a downturn. Roster savings are permanent. Studio savings are contingent on volume holding up.

Lego is the useful counter-example to full in-housing. Keeping brand platform work outside, at BETC, while the in-house agency carries the operational load, means the studio is measured on throughput and the external partner on the one thing external partners are actually better at: an outside view of the brand.

Coca-Cola's 2021 consolidation is the cleanest example of the trade-off being made knowingly. Simplify to one holding company, accept the dependency, and offset it with an internal content operation that gives you an alternative route to marketroute to marketThe strategy defining how you'll launch a product: target segments, channels, value proposition and coordinated action plan.View full definition → for fast work.

CMO action items

• Tier your annual output by volume, variance and speed requirement, then put a structure against each tier. If one structure is carrying all three tiers, it is failing at two of them.
• Run the utilisation calculation on any studio you already own. Fully loaded cost divided by assets actually delivered last year, against your external per-asset rate. Do it before the budget round, not during it.
• Put a named-core key-person clause and a quarterly attrition report into any dedicated-unit contract at renewal.
• Appoint a reference partner: one small independent on a live retainer, briefed on real work, kept as your comparator price.

Common mistakes that kill results

• Choosing a structure to solve a talent problem. If the work is weak, consolidating the roster will make it weak more efficiently. Structure fixes cost, speed and coherence; it does not fix a brief nobody believes in.
• Treating in-house as free capacity. Internal teams get requests nobody would ever pay an agency to fulfil, and utilisation looks healthy while the studio produces artefacts nobody uses. Audit output against actual deployment at least twice a year.
• In-housing the interesting work and outsourcing the volume. It reads well internally and it is backwards. The volume is where fixed cost pays back; the low-frequency, high-stakes work is where an outside view is worth paying for.
• Locking a structure for five years to get the discount. Category dynamics change faster than that. Three years with a defined exit and a transition-services obligation costs a little more and saves the reversal.