Frameworks & methodology for building high-performance marketing teams

You have 34 marketers, four product lines and a CEO who wants a new org chart signed off by the end of the quarter. Three shapes are realistically on the table: keep the functional department and add a planning layer on top, split it into cross-functional squads that own audiences or products, or push profit and loss accountability into small units that price and pay themselves. Whichever you pick, you own it for at least a year, because a reorg resets reporting lines, tooling, budget codes and every informal relationship people were using to get work done. That deserves a method rather than a preference.

The three models on the table

Squads and tribes come from Spotify's 2012 internal paper by Henrik Kniberg and Anders Ivarsson: small autonomous teams (squads) grouped into tribes capped near 100 people, with chapters holding craft standards across squads and guilds carrying shared interests across the company. Agile transformation, as ING ran it from 2015, applies that shape to a whole commercial organisation in one move, dissolving marketing as a standalone department and merging it with product, channel and IT. Micro-enterprise units, Haier's Rendanheyi model, go further: each unit carries its own P&L, sets its own targets, contracts internally with other units and can replace its own leader.

The axis underneath all three is how much autonomy you grant and what you attach to it so autonomy does not become twelve competing marketing strategies. Spotify attached craft chapters. ING attached a quarterly planning ritual. Haier attached money.

Key sub-concepts

1. squads and tribes: the coupling matters more than the squad

A squad is six to twelve people with everything needed to ship (writer, designer, analyst, channel owner), owning one segment or product area and running its own cadence. Most companies copy that part and skip the machinery that holds it together. Chapters group people of the same craft across squads, with a chapter lead acting as line manager and standard setter; guilds are voluntary communities of practice. Without chapters, four squads produce four visual languages, four naming conventions and four attribution models within two quarters, and you spend the following year re-centralising.

The second cost is duplication. Every squad wants its own analyst. If you have two analysts and five squads, you either dilute them into part-time slices or you accept that three squads fly blind. A usable threshold: cross-functional squads pay off when a squad can complete roughly four out of five pieces of work without waiting on anyone outside it. Below that you have built a matrix with extra standups.

2. agile transformation: an org event with a re-hiring step

ING moved around 3,500 people at its Amsterdam head office into roughly 350 squads of nine, grouped into 13 tribes, merging marketing, product management, channel management and IT development. Every employee had to reapply for a role in the new structure, and a substantial share ended up in a different job than the one they held. COO Bart Schlatmann, in the 2017 McKinsey interview about the change, was direct that culture rather than process was the hard part. Release cadence went from a handful of times a year to every two to three weeks.

The failure mode is skipping the reapplication step. Keep the same managers with the same approval habits and the sprint board becomes a status meeting with sticky notes. The other predictable consequence: senior people whose authority came from owning a discipline lose their team and gain a chapter, and some of them leave. Budget for the replacement hiring before you announce.

3. micro-enterprise units: autonomy priced in money

Haier broke itself into thousands of micro-enterprises of roughly ten to fifteen people and removed on the order of 10,000 middle management positions in the mid-2010s, documented by Gary Hamel and Michele Zanini in HBR (2018). User-facing units buy design, manufacturing and marketing services from platform units at negotiated internal prices, and can buy outside if the internal offer is worse. Pay follows value created, not headcount managed.

For a marketing function that has one sharp implication: internal marketing becomes a supplier that has to win the work, which is the hardest version of the in-house and agency question the agency relationships lesson works through. The model needs revenue attributable to each unit. If demand is created centrally by a brand campaign and cashed in by fifteen units, the internal market prices nothing and the strongest negotiator wins instead of the best work. Haier extended the model to GE Appliances after acquiring it in 2016, so it does travel, but it travelled into a business with product lines that each have their own customers and margins.

4. choosing: four tests before you draw boxes

• AttributionAttributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition →: can you draw a revenue or margin line around the proposed unit? If not, use squads with shared goals rather than micro-enterprises with P&Ls.
• Interdependence: what share of a unit's work can finish without approval from outside it? Measure it on last quarter's real deliverables, not on the ideal.
• Craft depth: can each discipline afford to be spread thin? One analyst does not distribute across five squads.
• Change cadence: two-week cycles pay off where the work changes weekly (performance, lifecycle, content, community). Brand platforms, sponsorship and pricing architecture run on annual cycles and get worse when forced into sprints.

Autonomy also needs a shared scoreboard. Two to four measurable key results per squad per quarter, set at squad level before anything cascades to individuals, and kept out of compensation decisions so nobody games the target down.

Real-world cases

Spotify: the model its own authors warned about

The people who wrote and lived the 2012 paper have said publicly that it described an ambition at one moment, was never uniformly implemented across the company, and carried known weaknesses: chapter leads overloaded as both managers and specialists, and squads that turned out to depend on other squads far more than the diagram suggested. Treat it as a vocabulary, not a blueprint. What transfers is small full-stack units, one clear owner per area, and standards held in chapters rather than in a review committee. Wrapped, in its current annual form since 2016, is put together by a comparatively small cross-functional group rather than by the whole marketing organisation, which is the point: the structure lets a small group ship something enormous.

ING: keeping the model after the noise

Tribe leads at ING reallocate people and budget in quarterly business reviews, which is what stops 350 squads from drifting. The bank kept the structure and rolled it beyond the Netherlands, so this is not a two-year experiment that quietly reverted. Worth noting what it did not fix: an org model changes who decides, not who you hired. Persistent skill gaps stay exactly where they were, in the roles and culture contract the foundations lesson defines.

Haier: when internal competition duplicates spend

Micro-enterprises negotiating for themselves will each buy their own tooling, their own research and sometimes their own agency, and the aggregate bill lands on the group. Haier's answer is the platform layer: shared capability sold at an internal price, so duplication becomes visible as a cost someone has to justify. If you copy the autonomy without the platform layer and the internal pricing, you get five units each paying full retail for the same data subscription.

CMO action items

• Run the four tests on paper against your current team before you draw a single box. Write the interdependence number down; it is usually far worse than leadership assumes.
• Decide the coupling first. Name chapter leads and the standards they own (brand, data definitions, media buying rules) in the same week you name squad leads, not three months later.
• Pilot one squad in your highest-change area, with real budget authority and a named end date for review. A squad that has to ask permission to spend is theatre.
• Publish the internal price of central capability if you are heading anywhere near the Haier model, even as a rough day rate. Unpriced shared services get treated as free and consumed accordingly.

Common mistakes that kill results

• Copying squads and skipping chapters. Autonomy without craft governance fragments the brand faster than any budget cut.
• Reorganising to fix a hiring problem. If the work is average because the skills are average, squads will produce average work in two-week increments.
• Giving units a P&L when revenue cannot be attributed to them. The arguments then move from strategy to accounting.
• Running the transformation without changing who manages whom. ING's reapplication step was the uncomfortable part and also the part that made the rest work.