Selling on Vietnamese marketplaces: what the commission table hides

A regional team evaluating Vietnam usually starts with the commission rate. It is published, it is comparable across platforms, and it fits neatly into a model.

It is also the layer least likely to be the reason the channel does or does not make money.

Seven deductions, one order

Six of the seven are absent from the report most teams look at.

None of them is a surprise in isolation. Every seller knows the platform takes a commission and that affiliates are paid. The surprise is in the total, and in how the mix shifts month to month depending on which promotional programme is running.

There is also a timing difference that a first model rarely allows for. The sales dashboard records an order on the day it is placed; the settlement report records it weeks later, after the delivery and return windows close. Matching one month of advertising spend against one calendar month of settlements therefore compares two different populations of orders — a distortion that stays invisible while volumes are flat and becomes severe in exactly the months when spending is being scaled.

Who funded the discount is the question

This is the layer that causes the most misreading, because to the shopper every voucher looks the same.

To the seller they are not. A platform-funded voucher does not touch your margin. A seller-funded one does. A shipping subsidy may be split. Two orders with identical final prices can return materially different amounts depending on who paid for the discount — and the settlement report separates those columns while the sales dashboard does not.

From list price to money received

Written as percentages, the stack is easier to carry into a regional model than any single fee figure. The proportions below are illustrative, to show how the layers compound — the real ones come from your own settlement report, and they differ by platform, category and month.

The number at the bottom is the one to model. Everything above it is arithmetic that the sales dashboard performs somewhere else, in a report nobody opened.

Two consequences for a regional plan. A reported return of four times advertising spend, measured on list price, is roughly two and a half times measured on money received — and those two figures often sit on opposite sides of a break-even threshold. And because the middle rows move with the promotional calendar, the ratio is not a constant: it should be recalculated each quarter, and after any change in promotional policy.

Why in-platform ROAS reads high

Advertising inside a marketplace reaches people already in a buying state. Conversion rates are higher and reported ROAS looks stronger than off-platform advertising. That part is real.

There is a second reason the figure flatters, and it is arithmetic rather than behaviour: in-platform reporting is generally built on order value at list or post-voucher price, while the money that arrives is the figure at the bottom of the stack above. The metric and the bank account are measuring different points on the same order.

It also competes directly on price with every other seller in the same category, on the same platform, in the same hour. Winning that competition is frequently paid for with deeper discounting. So high in-platform ROAS and thin in-platform margin tend to travel together, and looking at one metric does not reveal the other.

Who operates the store is part of the economics

For a company already selling in Vietnam, the store is theirs and the fee stack ends at the platform. For a company entering, there is usually another layer, and it belongs in the model from the first version.

The fourth row is the one most often underestimated, because its cost does not arrive as an invoice. A longer delivery promise raises the cancellation rate, and cancellations were already the largest single deduction in the stack above. The effect lands in marketing performance figures rather than in logistics, which is where it will be looked for.

The distributor row carries a different kind of cost: when the distributor holds the store, the customer data and the advertising history are theirs. That may be the right trade for a first year, but it should be a decision taken deliberately, with a stated review date — the same principle that applies to choosing and overseeing a local agency.

Traffic you send into the marketplace is a measurement dead end

One structural problem deserves stating on its own, because entrants meet it in the first month and there is no tidy solution.

Advertising run on social platforms that sends people into a marketplace store creates a gap that neither side closes. The marketplace attributes the order to itself, since that is where the session ended. The advertising platform reports a click and never learns what happened next. The same spend is therefore either double-counted or invisible, depending on which report is being read.

The partial remedies are worth knowing, and worth knowing the limits of:

• Store-level campaign codes or dedicated landing sections, so at least the arrival is identifiable.
• Campaign-specific voucher codes, which capture the orders that use them and miss the ones that do not.
• Holdout periods — turning the off-platform spend off for a defined window and reading the change in total marketplace revenue. Blunt, slow, and the most honest of the three.

None of them recovers order-level attribution. The practical answer is to stop trying: judge off-platform spend on its effect on total marketplace revenue over a period, and accept that the per-order figure does not exist. A team that keeps asking for that figure will keep receiving an invented one.

What to model before committing budget

Not a data project. One table, built once and refreshed monthly:

• Export the settlement report, not the sales report.
• Join in-platform advertising cost for the same period.
• Add affiliate and creator commissions — acquisition cost that sits in another report.
• Split by product line, never as a store average. The average conceals the line that is losing money.
• Apply the delivery acceptance rate by acquisition source.

The output answers the only question that matters for allocation: if we put one more dollar into this line, how much of it comes back and stays.

Two habits keep the table honest. Refresh it from the settlement report for the same month every time, rather than mixing a fresh sales report with last month’s fees, because fees and campaign terms change during the year and the table is only as current as its oldest column. And give it one owner who is not the person running the advertising, so that a line which has quietly stopped paying is reported as a finding rather than explained as a phase.

Four things a regional model usually gets wrong

Models built outside the market tend to fail in the same four places. None of them is a modelling error; all four are assumptions inherited from somewhere else.

• Using GMV where money received belongs. Gross merchandise value is a platform’s unit of account, not a seller’s. It is the figure most often quoted in a board pack and the least useful for deciding whether to spend more.
• Importing fee assumptions from another market. Commission is comparable across markets; the promotional stack on top of it is not. A model calibrated elsewhere can be accurate on the one published number and materially wrong on the total.
• Annualising a campaign month. Double-date campaigns and year-end peaks produce revenue at margins that do not hold for the rest of the year. Twelve times a peak month is not an annual plan.
• Treating the acceptance rate as a logistics metric. It changes by acquisition source, and it belongs in the marketing model because it decides how much of the reported revenue ever becomes money.

Questions for the platform’s own commercial team

Platform account managers are useful and will answer directly, but they are measured on your spending, not your margin. Three questions that produce specific answers:

• For a category like ours, which fee lines change with the programme we join, and which are fixed?
• In the campaigns you are proposing, what share of the discount is funded by the platform and what share by us?
• What is the typical gap, in weeks, between an order and its appearance in the settlement report?

The third answer determines how your finance team reconciles the channel, and it is the one least often asked.

Concentration is a number, not a feeling

When most revenue comes through one platform, every change to fees, ranking or promotional policy lands directly on your results, and you have no say in any of them.

That is not an argument for leaving. Marketplaces are where the buying happens in most consumer categories, and abandoning real revenue to avoid a hypothetical risk is a poor trade.

It is an argument for putting two figures in the monthly report next to revenue: what share of profit depends on a single platform, and whether any customer relationship exists outside it. Neither figure demands a decision. They make sure that when a decision is needed, you know where you are standing.

Once a year, add a stress test to those two figures: if the main platform’s traffic fell by a fifth for a quarter, or its fees rose by a few points, would the remaining contribution still cover the fixed costs of the Vietnam operation? The answer does not call for action on its own. It tells head office how much room there is before a platform decision becomes a business problem.

What “profitable” should mean in the board pack

Most disagreements about a marketplace channel are disagreements about a definition, discovered late. Agreeing the line items once, in writing, removes the argument for a year.

The last line is what MWY can reconstruct and stand behind. It is not net profit, and it should not be labelled as such in a board pack — cost of goods and inventory sit outside it. Reported honestly, it still answers the question the board is actually asking: is the next dollar into this channel worth spending.

Where this work stops

MWY reconstructs profit after advertising and acquisition cost — the amount actually received per the settlement report, less advertising spend, less affiliate and creator commissions.

MWY does not reconstruct true profit, which requires cost of goods and inventory and sits outside independent marketing oversight. On marketplaces the scope covers advertising, promotions, live selling, affiliate and store conversion. It does not cover cost of goods, inventory, fulfilment or returns handling.

Sizing the category and building the model before committing budget both belong to Go-to-Market Strategy. Once spending is live, keeping the table honest month after month is Vietnam Marketing Advisory & Oversight. How the same deductions distort measurement across every channel is covered under marketing measurement.