Digital team structure: organise around disciplines, not channels

Ask a head of digital to sketch their team and, more often than not, you will get a row of boxes with channel names in them: paid search, paid social, programmatic, and perhaps SEO and CRM squeezed in at the end. Beneath each box sits a small team that plans, buys, optimises and reports on its own channel.

That shape made perfect sense when the channels really were separate trades. Buying keywords in 2012 had very little in common with buying display through a trading desk; the tools, the people and the reporting were all different.

Rather a lot has changed since then. Performance Max and Advantage+ now decide where your ads appear across inventory types, largely on your behalf. Measurement has moved away from last-click attribution towards experiments and modelling. And the things that separate a good team from an average one today, such as feeding the algorithms clean data, testing properly, producing enough creative variation and keeping the tech stack in order, are not specific to any channel at all.

The channel org chart, in other words, answers the wrong question. It is organised around where the money goes, whereas we would argue it should be organised around what the team needs to be good at.

Channels and disciplines are different things

A channel is where media runs. A discipline is a body of expertise, with its own standards and craft, that applies across every channel. Measurement is a discipline, as is producing creative for performance. Search is a channel.

Organise by channel and each discipline ends up being rebuilt inside each channel team. The search team builds its own dashboards, social builds a different set, and programmatic reports on viewability while everyone else is talking about cost per acquisition. Nobody is wrong as such, but the numbers never quite reconcile, and the leadership team spends a good part of Monday morning deciding which of three versions of last month to believe.

In our work with brand and agency teams we use five core disciplines. You may well divide them slightly differently, which is fine, provided each one has a single, named owner.

1. Strategy and planning. Audiences, budget allocation across channels, forecasting and the media plan itself. This is where the trade-offs between channels are made, which is precisely why it should not sit inside any one of them.
2. Activation. The hands-on buying and optimisation across search, social, programmatic, retail media and whatever comes along next. Platform specialism still matters most here, so this is where channel expertise belongs, as a skill people bring to their team.
3. Measurement and analytics. Tracking, attribution, incrementality testing, marketing mix modelling and reporting, with one owner and one version of the truth.
4. Content and creative for performance. Asset production, variation, testing and product feeds. The platforms reward creative volume and variety, so this has become as much a question of capacity as of craft.
5. Marketing technology and data operations. Tagging, consent, server-side tracking, first-party data pipelines, feeds, automation and, increasingly, AI tooling.

What the research says

Very little of this is new to organisational design. Marketing has just been slow to borrow from it.

Melvin Conway set out the underlying principle in 1968, observing that organisations which design systems tend to produce designs that mirror their own communication structures. Software engineers have been quoting "Conway's law" ever since, and it applies to a media plan just as neatly. If your search and social teams rarely speak to one another, you will end up with a search plan and a social plan, rather than a single plan that happens to use both.

The evidence in favour of cross-functional working is reasonably consistent. MIT Sloan Management Review and Deloitte surveyed more than 4,800 executives across 129 countries and found that 44% of digitally maturing organisations used cross-functional teams for digital initiatives, compared with 16% of those at an early stage. The gap on strategy was wider still: 81% of maturing companies said they had a clear and coherent digital strategy, against 15% of early-stage ones.

BCG and Google's research into digital marketing maturity, which covered more than 200 brands, points in the same direction. Eighty per cent of the companies surveyed described their cross-functional coordination as inadequate. The most mature marketers reported cost savings of up to 30% and revenue gains of up to 20%, and high organisational maturity was associated with 2.5 times the revenue impact of less mature peers. These are correlations, of course, and the most mature firms no doubt differ in plenty of other respects. Even so, it would be difficult to read that study and conclude that structure is a minor detail.

Team size is a matter of arithmetic

Each person added to a team brings additional communication overhead. The number of one-to-one links between n people is n(n-1)/2, so a team of five has ten links, while a team of twelve has 66.

J. Richard Hackman, the Harvard psychologist who spent decades studying teams, pointed to exactly this growth in links when explaining why larger teams tend to underperform; his rule of thumb was to avoid double figures. McKinsey's guidance on agile marketing arrives at a similar answer, capping each team at eight to twelve people. We tend to aim for five to nine.

Cognitive load is the less obvious constraint

Matthew Skelton and Manuel Pais wrote Team Topologies with technology organisations in mind, but one of its central ideas translates well: there is a limit to how much any team can reasonably hold in its head. Ask a single team to run search, social and programmatic, own the measurement framework and look after the tag manager as well, and it will do all five of them badly. Their answer is to keep "stream-aligned" teams focused on a single flow of value, supported by platform teams that provide shared tooling as a service, and by enabling teams that build a capability and then step away.

That vocabulary transfers to a marketing function with very little adjustment, as the model below shows.

A note of caution on the Spotify model

Anyone who has read about squads, tribes and chapters will have come across the Spotify model, and it is worth approaching with some care. Jeremiah Lee, who worked at Spotify, wrote in 2020 that the company never fully operated the model as published, quoting co-author Anders Ivarsson: "Even at the time we wrote it, we weren't doing it". Lee's two main criticisms were that chapter leads managed people while having little say over delivery, and that teams were given autonomy without sufficient alignment. Both problems can be fixed, and the model we recommend below is designed with them in mind.

Four common structures, compared

Most digital teams we come across use one of four shapes, or some combination of two.

Hub-and-spoke deserves particular mention because it is so common in regional set-ups, the Middle East included, where a single hub often serves several markets. It is not really an alternative to pods and chapters; the two sit together comfortably, with the hub simply being where the chapters and the platform team are based.

The model we recommend: pods, chapters and a platform

For most teams beyond the early stages of maturity, we would recommend three layers.

1. Outcome pods

A pod is a small, cross-discipline team that owns a commercial outcome rather than a channel: new customer acquisition in the UK, for example, or app growth, or a single product category. In an agency, it might be a major client. Each pod would typically include a planner, two or three activation specialists covering the channels that outcome actually needs, an analyst (sometimes shared between two pods) and a content or creative producer.

The pod lead is accountable for the priorities, the backlog and the target, and the pod itself should number somewhere between five and nine people.

2. Discipline chapters

Chapters run across the pods. The measurement chapter, for instance, includes every analyst in every pod. Its lead sets the standards (naming conventions, the attribution approach, how tests are designed), looks after hiring and training, and owns career paths.

This is where the Spotify version came unstuck, so it pays to be explicit. Everyone should have one line manager of record, and in our model that is the chapter lead, since decisions about development and promotion need a specialist's judgement. Day-to-day priorities, meanwhile, come from the pod lead. It is worth writing that split down, agreeing how disagreements are escalated, and revisiting the arrangement after six months.

3. Platform and enablement

A small team owns the shared machinery: tag management, consent and server-side tracking, data pipelines into the CDP or warehouse, product feeds, reporting infrastructure and automation. In Team Topologies terms this is the platform team, and it treats the pods as its customers.

The same team, or a separate one in a larger organisation, can also take on the enabling role, for instance by rolling out a new bidding approach or AI tool, coaching each pod through it for a few weeks and then moving on.

And above all three

A head of digital, or group lead in an agency, owns the overall strategy, the allocation of budget between pods and the relationship with the wider business. The strategy and planning chapter supports that role directly, which keeps cross-channel trade-offs with the people who are actually in a position to make them.

Where the line falls between in-house and agency

These days, almost every conversation about structure arrives with a second question attached: what should we be doing ourselves?

In-housing stopped being a passing trend some time ago. The Association of National Advertisers (ANA) found that 82% of its members had an in-house agency in 2023, up from 42% in 2008. In programmatic, an IAB study found that 65% of US brands buying programmatically had moved some or all of it in-house by 2018, and by 2020, 84% of advertisers in a World Federation of Advertisers study were using an in-house or hybrid "brand trading desk".

Budgets are pushing in the same direction. In Gartner's 2025 CMO Spend Survey, marketing budgets remained flat at 7.7% of company revenue, 39% of CMOs planned to reduce agency spend, and 22% said generative AI had already lessened their reliance on agencies for creative and strategy work.

None of this means that everything belongs in-house. Our rule of thumb is a simple one: own what compounds, and buy in what scales.

What compounds is knowledge that grows more valuable the longer you hold on to it, such as your first-party data, your measurement framework and test history, ownership of your ad accounts and your planning principles. If those sit with an agency, they tend to leave when the agency does.

What scales is capacity and specialist depth that you would struggle to keep busy, or up to date, on your own payroll: peak-season activation, creative production at volume, niche platforms and new channels that are still being tested.

For those working agency-side, the same model holds on your side of the table: client teams map to pods, practice heads to chapter leads, and the data and technology team is the platform. The difference is largely commercial. Utilisation targets tend to pull agencies towards channel teams, because a channel specialist is easy to time-sheet against a scope of work. The agencies we see doing best have, quite sensibly, stopped letting the timesheet draw the org chart.

Matching the structure to your maturity

There is no need to put all of this in place on day one; a team of six probably should not have chapters at all. Using BCG's four maturity stages, this is how we would phase it.

Making the change

In our experience, restructures fail far more often because of how they are run than because of the design itself. There are a few things we would always do:

• Start from outcomes. List the four or five commercial outcomes the digital team exists to move, then check that every proposed pod owns at least one of them.
• Move measurement first. It is the lowest-risk change with the quickest return, and a single analytics owner tends to put an end to the reconciliation arguments almost immediately.
• Pilot one pod for a quarter. Choose an outcome with clean data, so that you can tell whether it has worked before converting everyone else.
• Write down decision rights. Who sets pod priorities, who can move budget between channels and who owns the tracking plan. A lack of clarity here undid a good many Spotify-style rollouts.
• Protect channel depth. Activation specialists still need deep platform knowledge, so give them time, through the chapter, to share it across pods.
• Review at six months. Some pods will turn out to be too large and some outcomes too vague, and both are easily adjusted.

If you change only one thing

Take measurement out of the channel teams and give it a single owner. Channel-shaped activation is something you can live with for a while; running a sensible digital function on three competing versions of last month is rather harder.