How to oversee a marketing agency in Vietnam

Most foreign companies entering Vietnam choose a capable agency. Proposals are compared, references are checked, a decision gets made, and the decision is usually reasonable.

What goes wrong tends to be decided earlier and more quietly, in the first few weeks, in choices that look administrative at the time and structural a year later.

Three decisions that carry most of the weight

Who the ad accounts belong to

The accounts should be registered to your company, with the agency granted access. Not the reverse.

This is not about trust. Historical performance data, custom audiences and conversion tracking are company assets, and their value comes from continuity. An account that has been learning for eighteen months is worth considerably more than a fresh one, and that value evaporates if the account cannot move when the relationship does.

In Vietnam, accounts held in the agency’s name are common enough that nobody flags it as unusual. Ask in the first week. Unwinding it later is possible but costs months of history.

There is a practical complication for entrants: the local legal entity often does not exist yet when the first campaigns run, so the accounts get opened under whoever can open them. That is a reasonable interim step and a poor permanent one. Write the transfer into the contract at the point the entity is registered, with a date, rather than leaving it as an intention.

Which system the reports draw from

A report built from the advertising interface and a report built from the revenue system will not agree. Both are accurate within their own boundaries, and the two sides can argue indefinitely without either being wrong.

Name one system as the decision source in the contract. For most companies that is whatever records actual revenue. The others stay useful for daily optimisation — operators need them — but budget decisions come from one place.

Which metric the contract measures

An agency improves at the metric it is held to. That is not a flaw; it is how any organisation with a target behaves. The problem appears only when that metric is not the one the business decides on.

No row is wrong. Each one pulls in a different direction, and it is worth knowing which direction you have chosen.

What the scope actually covers

Agency proposals in Vietnam are often broader than their equivalents elsewhere: media, creative production, marketplace operations, creator and affiliate management, sometimes customer service on chat. Breadth is genuinely useful for an entrant with no local team. It also makes the boundary between capability and subcontracting hard to see from a proposal.

Three questions settle it, and all three have short answers:

• Which of these lines does your own team deliver, and which do you subcontract? Subcontracting is not a problem; not knowing about it is.
• Who specifically handles marketplace operations and live selling? These are the parts that most often sit with a partner, and the parts a regional reviewer is least equipped to assess.
• If we removed one line from this scope, which would you recommend removing? An agency that answers this candidly is telling you where it is strongest, which is more informative than the proposal.

There is a related question about concentration on your side. A single partner covering media, creative and marketplace operations is efficient and leaves no one in the room who can check the others’ numbers. That is a workable arrangement as long as somebody outside it reads the results — and on the marketplace side, reading them means understanding how marketplace fees and promotions work before judging any number that comes out of that channel.

What is genuinely different here

The marketplace channel is a separate discipline. Shopee, Lazada and TikTok Shop run their own advertising systems with their own rules and their own reporting. Strength in Meta and Google search does not transfer automatically, yet the two are often quoted as a single line in a proposal.

Chat is a sales channel. A large share of selling happens in Zalo and Messenger. Ask how the agency measures it. If they cannot, an entire channel is absent from every report you will receive, and the channels that feed it will look weaker than they are.

Cash on delivery separates the order from the money. An order counted in the campaign report is not revenue until the parcel is accepted. That gap varies by category and by acquisition source, and it belongs in the reporting definition rather than in a footnote. Whether the underlying data can be trusted at all is a fair question, and it has a specific answer under marketing measurement.

A regional agency of record meets a local market

Companies with a global or regional agency of record usually extend that contract to Vietnam by default. It is the path of least resistance and it is often right for media buying on the large international platforms.

It is less often right for the rest. Marketplace operations, creator and affiliate programmes, live selling and chat response are local disciplines with local labour requirements, and a regional contract rarely prices them, staffs them or measures them. What happens next is predictable: those channels either go unserved, or get subcontracted locally without appearing in the reporting line the head office reads.

The workable arrangement is a deliberate carve-out rather than an exception discovered later. Three points make it hold:

• Name which channels sit outside the regional contract, in writing, before launch rather than after the first quarter.
• Insist both parties report into one definitions page. Two partners with two definitions of “conversion” produce a combined report that cannot be added up.
• Decide who consolidates. Someone has to own the single view, and it should not be one of the two partners being compared in it.

The alternative — a regional contract that nominally covers everything and locally covers half — is the arrangement that produces the most surprised quarterly reviews, because on paper nothing was missing.

The first thirty days, in writing

Most of what goes wrong later is cheap to prevent in the first month and expensive to fix afterwards. Six items, one page:

The fourth row is the one teams skip, and it causes the most damage. “Conversion” means at least four different things across the systems in use here, and a report that never states which one it means will be read as whichever meaning the reader assumes.

Distance makes the review meeting worse than it looks

The structural difficulty in overseeing a Vietnamese agency from abroad is not language or time zone. It is that the review meeting degrades into a status read-out, and a status read-out cannot surface disagreement.

Three patterns, and each has a practical counter:

• Reporting happens in English while the work happens in Vietnamese. Nuance is lost twice — once translating out, once translating back. The counter is to ask for the underlying export alongside the summary, at least quarterly, so the two can be compared.
• The call reviews what happened, never what was decided. The counter is a standing agenda item with one question: what did we change since the last call, and what did we learn from it?
• Nobody disagrees in the meeting. Directness in a review call is not the norm in every working culture, and silence gets read as agreement. The counter is to ask for the objection explicitly — “what in this plan would you do differently if it were your budget” — and to ask it of the person running the account, not the account director.

None of this is about mistrust. It is about the fact that an hour-long call with a slide deck is a poor instrument for finding out that something is not working.

What good reporting looks like

A useful monthly report is shorter than most and ends with a recommendation. Five parts:

• A definitions page that never changes. One page, at the front, stating what each metric means and which system it came from.
• The decision numbers, from the decision source. Three or four, not thirty.
• What changed this month. Budget moves, creative changes, audience changes, with dates — so results can be tied to causes later.
• What we learned. Including tests that failed. A report that only contains successes is a report that has stopped testing.
• What we recommend, including what to stop. A recommendation list with no stop on it in twelve months is not a recommendation list.

If the monthly report already looks like this, quarterly reviews become straightforward. If it does not, no amount of quarterly review will compensate.

A quarterly review that takes an hour

Monthly reviews look at operating numbers. The relationship itself needs a different, rarer conversation:

• Is the metric in the contract still the right one? Businesses change faster than contracts.
• Is the named team still the team on the account?
• Has anything become measurable that was not measurable last quarter? If so, which past conclusions should be revisited?
• What would have to be true for the agency to recommend spending less?

That last question is worth asking plainly. Where an agency is paid a percentage of media spend, recommending a reduction works against its own income. That is a structural fact rather than a character judgement, and the appropriate response is not suspicion — it is making sure the decision to increase or reduce budget has a second perspective on it.

Signs the relationship is drifting

Drift is rarely announced. It shows up as four small things, each easy to explain away on its own:

• The report grows longer while the recommendation section shrinks.
• The named senior person appears only in quarterly calls.
• Requests for raw exports take longer than they used to.
• Every proposal is to spend more, and none is to spend differently.

Any one of these can have an innocent explanation. Three at once usually means the account has been moved down someone’s priority list, and the correct response is a conversation rather than a tender.

That conversation goes better with the evidence on one page: the dates of the last three raw-export requests and when each arrived, the attendance of the named team over two quarters, and the share of recent proposals that changed how money is spent rather than how much. Dates make the point without accusing anyone, and they give the agency something concrete to fix.

Where MWY sits

MWY does not run campaigns and takes no commission from any agency or platform.

What MWY does is the third job: setting the criteria before you choose, shortlisting three vetted agencies against them, reading the proposals, reviewing the contract, and then reading the numbers back each month on the company’s side of the table. Taking money from both sides would remove whatever value that reading has, so MWY takes it from one.

Assessing an existing agency relationship and the numbers it produces is what a Digital Marketing Audit does. The continuing version — monthly reading, challenge and reporting to the head office — is Vietnam Marketing Advisory & Oversight, which opens with that audit as its first month.

The shortlist comes with the criteria that built it, so the choice can be checked rather than taken on trust, and MWY takes no commission for making it. That is the line that keeps a recommendation independent: not declining to name anyone, but being paid by only one side.