Suppose an agency launches the same acquisition program in three countries with a consistent account structure, clean conversion tracking, and comparable budgets. A month later, Germany is producing clicks, but few leads; Spain is producing leads at an acceptable cost, but sales complain they are poor quality; and the UK campaign is converting fine on the original creative. Someone reading the dashboards would probably conclude that the German landing page is weak and the Spanish targeting is loose. Someone who knows those markets might conclude that the offer sounds odd in German and that the Spanish lead form invites the wrong kind of inquiry. Only one of those readings is available from headquarters.
In December 2024, the World Federation of Advertisers asked 17 CMOs from its CMO Forum whether their global marketing organizations were centralized, decentralized or hybrid, and where headcount sat between global, regional and local. The answers are member-only, but the fact that a benchmark among the world’s largest advertisers still has to ask which model a company runs suggests the question is not settled at any scale. Rather than choosing a model, an agency does better to sort the work itself according to which decisions improve when they stay consistent and which ones deteriorate the further they move from the customer.
Measurement gets worse when every country reinvents it
Conversion definitions are the clearest example of work that needs to stay consistent. If one market reports a lead when somebody submits a form, another waits until the sales management tool accepts the lead, and a third counts booked meetings, then the three markets are reporting different things under one name, and every cross-market comparison is built on that mismatch. The same applies to attribution settings and to the duller parts of account governance such as naming conventions, access rights, and report definitions. Local teams should feed information into all of these, but they should not be able to rebuild the measurement model because the country dashboard looks inconvenient.
Centralizing measurement is not the same thing as buying a shared reporting stack, and research WFA published with Ebiquity in July 2026 shows large advertisers stuck at exactly that distinction. Among 71 senior leaders surveyed and 25 interviewed, collectively responsible for around $40 billion in annual ad spend, eight in ten already run marketing mix modeling and brand-lift studies. Only 13% rate themselves strong on moving quickly from data to insight, and only 15% say effectiveness evidence is the primary driver of how budgets get set. WFA’s global media lead, Tom Ashby, attributes the difference between leaders and laggards to governance and decision rules rather than to the tools. For an agency, the central measurement role is worth having only if it owns the definitions and the rules for acting on them. A central team that maintains the dashboards while each market argues about what a lead is has the administration without the authority.
A location setting is not local knowledge
Google Ads makes geography easy to configure. Campaigns support targeting countries, regions, cities, or a radius, and Google’s own documentation describes location targeting as a best effort built from user settings, devices, and behavior, with no guarantee of accuracy. None of this tells the agency what buyers in a market expect to hear. A central media buyer sees that a keyword produces an expensive CPA in France. A French marketer is more likely to notice that the phrase belongs to an earlier stage of the buying process, or that a local competitor is running a promotion the global team has never seen.
Language is about to become a sharper version of this problem. On 13 August 2026, Google’s Ads Developer Blog announced that from late September the campaign-level language targeting setting will be removed from Search and AI Max for Search campaigns, and will stop applying to the Search inventory inside Performance Max. Matching will instead run on the language of the ads and landing pages, combined with what Google already infers about the languages a user understands. Google had announced the same removal in August 2025 with a year-end deadline and then let it slip by nine months, so the September date is a plan rather than a certainty.
Google’s help page illustrates the new behavior with a user who searches in Spanish and, because her other activity suggests she reads English, may be shown an English ad. Once the campaign setting is gone, the language of the ad copy determines which queries the ad is eligible for, and what many agencies treat as a translation step at the end of production becomes part of targeting. The change also creates a problem the local marketer cannot solve alone. In a Swiss, Belgian, Canadian, or Spanish account, where many users read more than one language, ads from one campaign sometimes become eligible for queries the agency previously fenced off for another, and how the account handles that overlap has to be decided once, centrally.
Creative production and creative judgment are different jobs
Agencies centralize creative because production gets cheaper and faster when designers, copywriters, and brand assets sit inside one system. The difficulty is that central production tends to turn into central judgment without anyone deciding that it should. Suppose a global team builds six paid-social concepts and ships translated versions into four markets, all on brand. One local team can see immediately that the featured testimonial comes from a company nobody in its market recognizes. Another notices that the offer in three of the six concepts is weaker than what local competitors already advertise. If fixing either requires a request back to headquarters, the local team’s knowledge is being paid for and rarely used.
Local marketers therefore need the authority to reject a weak translation, swap the proof on a landing page, respond to a competitor offer, and flag creative that misses a local convention, while the central team keeps what travels well: brand assets, production quality, testing documentation, and the record of what has already been tried elsewhere.
Paid media rarely needs to be duplicated
Media buying is different from copy and positioning because most of the craft moves across borders intact. One paid search specialist handles account structure, bidding strategy, budget pacing, and testing across several markets while spend is modest, and sees patterns across markets that a buyer embedded in one country would miss. Hiring a media buyer in every country too early also creates a slower problem: each person develops a preferred way to structure campaigns, and six months later the agency is deciding whether Germany’s structure or France’s becomes the template for the next market.
A dedicated local media role becomes worth the cost when the work itself turns local: when search behavior in the language generates enough complexity to occupy someone every week, when promotions follow a market-specific calendar, and when the central buyer spends more time asking for context than managing media. Until then, a central buyer paired with someone who understands the market is usually enough.
Hire for work that keeps coming back
A common early setup runs on whoever happens to have the missing context: a freelancer checks campaign copy, the client explains local competitors on account calls, and someone on the central team who once lived in the country gets asked whether a headline sounds natural. That works while the questions are occasional and becomes expensive once they arrive every week, and the transition is easy to miss because no single request looks like a hiring signal.
A more reliable test is to take the previous month of campaign work and mark every task that required somebody in the market: copy review, competitor interpretation, client calls at inconvenient hours, creator relationships, market-specific landing-page changes, local promotional planning. If the same cluster keeps returning, the role already exists informally, and the agency has simply not hired for it. An ad-spend threshold, the more common trigger, describes the budget and nothing about the work.
Also Read
The employment decision comes after the role is clear
Once the agency knows the work deserves a permanent local owner, a separate question follows about how to employ that person. An independent contractor suits project work that is actually independent. A local marketer working continuously inside the agency and reporting to its managers is a different arrangement, and because classification rules depend on the country, the structure needs local review.
Setting up an entity is a much larger commitment than one hire in a market the agency is still testing. An employer of record covers the gap between those two options. G-P, which currently supports hiring in more than 180 countries, acts as the legal employer and handles locally compliant contracts, payroll, statutory benefits, and employment administration, while the employee works day-to-day for the agency. G-P’s case study on JANZEN, a Dutch fragrance brand with around 40 employees, describes using the EOR to hire employees in Germany with no local entity and a first hire in place within 30 days. JANZEN is a consumer brand rather than an agency, but the employment mechanics do not change with the job title. What an EOR does not do is decide whether Germany needs a local marketer. That answer comes from the workload, and the EOR becomes relevant only after the agency knows whom it wants to employ and has no employment infrastructure in that country.
Give the local hire something they are allowed to own
A local marketer needs a defined area where approval from headquarters is unnecessary, such as copy changes within brand rules, keyword expansion, competitor responses, or moving a small share of the test budget. A large budget shift or a new claim with legal implications belongs elsewhere. The same principle can be seen in a distributed workforce management platform, where a common system provides central visibility and consistency while allowing teams in different roles and locations to act within defined boundaries. The exact boundary matters less than writing it down, because without one, central teams pull decisions back whenever a market underperforms, local teams make small changes informally because waiting slows the account, and within a few months nobody is sure who owns the result. When WFA’s Rob Dreblow summarized in August 2026 what member CMOs said was actually slowing their transformations, the answers were rarely about technology. They were about habit, opinion overriding evidence, and a desire to hold onto control that ends up slowing everything down.
A review of recent decisions is more revealing than a review of job descriptions. Take the last ten meaningful campaign changes in a market and ask who spotted the issue, who understood why it was happening, who had the authority to act, and how long the decision waited. If headquarters repeatedly had the answer and local execution was inconsistent, the agency has a central process problem. If the central team kept waiting for someone to explain the market, the missing capability belongs closer to the customer. If everyone understood the issue and nothing moved because ownership was unclear, another hire will not fix it, and the next org-chart discussion should start from that finding.
