Setting up Vietnam measurement a head office can report on

The request from a regional or global head office usually arrives as one line: Vietnam should report like every other market. Same dashboard, same definitions, same monthly cadence.

There are two common ways to meet it, and both fail. The first pours Vietnam's numbers into the group template as they come out of the tools. The report looks consistent, and it means something different from every other market on the page. The second declares Vietnam a special case and lets it report its own way. The numbers are honest, and after two quarters nobody at head office reads them.

A setup that works sits between the two. Every Vietnam number either maps to a group definition that has been written down, or it appears as a declared exception with a stated reason and a stated adjustment. What follows is how that setup is built — in what order, by whom, and where the tools fit. It is not a tagging guide. The configuration work belongs to whoever implements it, and it goes faster once the decisions below have been taken.

Start from the group report, not from the tools

The usual first step is to install analytics and advertising tags, then work out what to report. It is the wrong way round. Tags collect whatever they are configured to collect, and the configuration depends on decisions nobody has taken yet.

Start instead with the report head office already receives from other markets. List every line Vietnam will have to fill, and for each one write down what the line silently assumes.

Every cell in the right-hand column is a definition. None of them is a technical setting. Settle them in writing with finance and head office before anyone opens a tag manager, and the implementation becomes a matter of following a specification rather than a string of judgment calls made by whoever happens to be configuring the tool.

The spine is the order record, not the analytics property

In many markets a team can treat its web analytics as the main record of online sales and reconcile everything else against it. In Vietnam that arrangement leaves out most of the business.

Marketplace orders are completed inside Shopee, Lazada or TikTok Shop and never pass through the company's analytics; what the settlement reports deduct from each of those orders is covered under marketplace advertising. Orders agreed in Zalo or Messenger are keyed in by a salesperson, if they are recorded at all. The website is often where a buyer checks the brand rather than where they pay.

The record that does see every sale is the order system — or, in a smaller operation, the spreadsheet that serves as one. That is where the spine has to be.

Three fields on the order record make the rest of the setup possible:

• A source field that cannot be left blank, with a fixed list of values: website, each marketplace, each chat channel, distributor, other. For chat orders, a second field records which campaign or page started the conversation.
• A customer identifier that works across venues. In Vietnam this is almost always the phone number. It is imperfect — people share numbers and change them — but it is the one identifier that appears on website orders, chat orders and delivery records alike.
• A delivery outcome, updated when the parcel is accepted, refused or returned. Without it, the revenue in the group report is confirmed orders, whatever the column is called.

Web analytics keeps an important job. It reports behaviour on the company's own site and apps, and the source of the orders completed there. It is simply not where the group's revenue figure should come from.

Three layers, and who owns each

A measurement setup has three layers, and most of the trouble comes from assigning them to the wrong people.

The common failure is to fund the first layer, because it is the one that can be bought, and assume the other two will follow. They do not. A well-configured analytics property feeding into no reconciliation produces precise numbers about a fraction of the business.

The second failure is to let the party that runs the media also produce the reconciliation. This is not a question of honesty. The agency does not hold the settlement reports or the delivery outcomes, so whatever it produces is a partial view presented as a total.

What the GA4 setup has to decide

GA4 is still the analytics tool most groups standardise on, and a GA4 setup for Vietnam is usually where a head office request lands first. The configuration itself is routine. The decisions that shape it are not, and when nobody takes them they fall back to defaults that suit neither Vietnam nor the group.

A few of these deserve a sentence more.

Time zone matters more than it looks. Vietnamese marketplaces run their largest sales on double dates — 9.9, 10.10, 11.11, 12.12 — and a heavy share of ordering starts at midnight. A property reporting in a European time zone puts that opening surge on the previous calendar day, which makes day-level comparison with other markets meaningless in exactly the weeks head office watches most closely.

Convert currency once, in one place. Whether collection happens in dong or in the group currency matters less than the rule that conversion is done in a single layer, at a documented rate, and never repeated. A large share of unexplained differences between a country dashboard and a group report turn out to be two conversions at two rates.

Key events should not pretend to know about delivery. GA4 records that an order was placed. It does not know whether the parcel was accepted. Delivery outcomes belong in the reconciliation layer, joined to the order record, rather than being pushed back into the analytics property to make it look complete.

Consent is not an analytics setting. Vietnam has its own personal data rules, and how consent is collected is a decision for legal counsel before tags go live, not a toggle for whoever configures them.

Attribution: choose a rule you can explain, then hold it

Marketing attribution in Vietnam is usually raised as a question about models: last click, data-driven, position-based. The choice of model matters less than two structural facts.

The first is that site-based attribution can only share out credit for sales that happen on the site. Marketplace and chat sales sit outside it entirely, so any model applied inside the analytics tool is a model of the smaller part of the business.

The second is that every advertising platform attributes to itself. Adding up the conversions each platform reports produces a figure larger than the number of orders, and no model applied afterwards removes the overlap.

A setup head office can use therefore reports two things and labels them clearly:

• Orders by recorded source, from the order system. This covers every venue and counts nothing twice. It is crude — it credits whatever started the last visit or conversation — but it is complete, and the same rule applies to every sale.
• Attribution within owned channels, from the analytics tool, explicitly labelled as covering the website and apps only. This is where a data-driven or position-based model earns its place, because it answers a narrower question honestly.

Platform-reported conversions are kept for what they are good at, which is optimising inside each platform, and left out of any cross-channel total.

Whatever rule is chosen, hold it for at least two quarters. A new attribution rule rewrites every channel's history at once, and a trend that restarts whenever someone prefers a different model is not a trend. If the rule has to change, report both versions side by side for a quarter before switching.

Measuring ROI in a way the group will accept

Measuring ROI in Vietnam is where the gaps above stop being technical and start reaching the board pack. A worked month shows why. The figures are invented for illustration, in US dollars, for a company spending 40,000 on advertising:

The first line is what a platform dashboard shows. The last line is the one that describes the business. Nobody in the chain has misreported anything; each line is accurate for the question it answers.

The group report should carry the last line as the headline and keep the four above it visible, because the distance between the first and the last is itself a figure head office needs. If that distance changes from one quarter to the next, something in the business has changed.

Two conditions make the ratio comparable with other markets. The cost side has to include acquisition costs that sit outside advertising accounts, because in Vietnam creator and affiliate commission can be a large part of what a customer costs — the wider comparison problem is covered under customer acquisition. And the revenue side has to be taken at the same point in every market: if other countries report cash received, Vietnam cannot report orders placed.

The ladder ends at profit after advertising cost. Going further needs cost of goods and operating costs, and the gross margin used for that step should be the figure the company's finance team supplies.

Where Vietnam will not fit the template, and how to say so

Some differences cannot be defined away. The honest response is to list them, state what each one does to the numbers, and attach the list to every report that goes to head office.

The list is short on purpose. An exceptions register that grows every month is a sign that the definitions were never settled. One that stays at five or six lines, with each line's size measured and re-measured, is what lets head office read Vietnam alongside other markets without either forcing the fit or giving up on it.

The order to build it in

• Definitions first. The table in the first section, filled in and agreed with finance and head office.
• Order-record fields next. Source, customer identifier and delivery outcome, required on every order from a fixed date.
• A monthly reconciliation, joining every venue to the order record, owned by finance.
• Then the collection layer, configured to a written specification: the analytics decisions above, channel rules, and naming conventions for campaign links.
• Then the attribution rule, written down with the date it took effect.
• Finally the group mapping and the exceptions register, attached to the first report that goes upstairs.

Built in this order, the tools arrive into a structure that already knows what they are for. Built the other way round — tools first, definitions whenever someone asks — a country ends up with a well-configured analytics property and a revenue figure nobody at head office trusts.

Where this work stops

MWY does not implement tags, configure analytics properties or resell software. The party that writes a specification should not be the party that builds against it, for the same reason the agency running the media should not produce the reconciliation.

What MWY does is take the decisions above — the definitions, the specification for the collection layer, the attribution rule, the exceptions register — write them down, and sign off when the agency or IT team has built them. That is the scope of Measurement, Martech & AI. Checking an existing setup against the same standard, once, is part of a Digital Marketing Audit, and the sample audit report shows what that looks like for a hypothetical company.

The reconciliation stops at profit after advertising cost. MWY does not reconstruct true profit, which needs cost of goods and operating costs outside the scope of independent marketing oversight. Where a margin appears in MWY's work, it is the figure the company supplied.