Vietnam SME Market Data: How to Source and Use It
Small and medium enterprises (SMEs) make up the overwhelming majority of businesses in Vietnam, yet reliable data, organized cleanly in one place, is surprisingly hard to find. Companies entering the market, sales teams hunting for local partners, and decision-makers weighing investments all hit the same wall. In a market like Korea, well-maintained registries and commercial business-information services make "where do we buy data" the question; in Vietnam, the problem is that usable data is not gathered in one place to begin with. This article lays out, from a practitioner's view, what Vietnamese SME data actually is, where and how to source it, what it is used for in practice, and the limits and cautions you must understand before working with it.
What Vietnamese SME data is
Vietnamese SME data is a structured body of information on the identity, size, sector, location, and activity signals of small and micro businesses. It includes business-registration details, industry classification, headcount, address, online presence, and activity indicators — not a single statistic, but a dataset identifiable at the level of each individual business.
The key here is the distinction between "aggregate statistics" and "business-level data." Figures like the SME share of the economy or the number of establishments by region, published by the statistics office, are useful for sizing the market, but they cannot identify a specific business or turn it into a target. Business-level data, by contrast, lets you reach each establishment — something like "an active food-and-beverage retailer with fewer than ten employees in Dong Da district, Hanoi." Statistics answer "is this market big," while business-level data answers "so who do I actually contact." When you move from planning to operating, you need the latter.
The very definition of an SME differs in Vietnam. The country classifies micro, small, and medium scale by headcount together with revenue or capital, and the thresholds vary by sector — agriculture and fisheries, industry and construction, trade and services. So when someone says "SME data," you should clarify upfront whether it includes one-person micro-businesses or only registered legal entities above a certain size, so you don't misread the nature of the data. Crucially, alongside formally registered companies (doanh nghiệp), Vietnam has a vast population of household businesses (hộ kinh doanh). These household businesses — neighborhood eateries, cafés, small retailers, salons — form the capillaries of the market, yet they often do not appear in corporate-registration data. If your mental image of a "Vietnamese SME" is only the registered company, you may be missing more than half the market.
The forms of the data vary too. The base layer is structured data from business registration, with fields like business name, registration number, address, industry code, and founding date. On top of that sit online activity signals. Digital traces — a Facebook page, a Zalo business account, a Google Business profile, presence on an e-commerce marketplace — help gauge whether a business is "actually alive and operating." Vietnam has an unusually high share of social commerce and messenger-based transactions, so online signals fill in the real activity that registration data alone cannot show. In Korea you might check a homepage or a maps listing first, but in Vietnam a Facebook page and a Zalo account effectively serve as a business's "storefront" and "phone book." Some shops run ordering, payment, and delivery inquiries entirely through a single Facebook page, with no real website. So if you apply Korea-style channel weighting when reading digital signals, you will underestimate actual activity.
Look closer at the fields a record carries and you see layers. Primary identity (name, address, sector) says "who is this business," size information (headcount, estimated revenue band) says "how large," activity signals (recent posts, reviews, traces of operation) say "is it alive right now," and contact information (public reachable channels) says "how can I reach it." Good SME data ties these four layers together within a single record so they reinforce one another. Data with only one layer — say, a directory of names and addresses with no activity signal — is an "address book," not an "actionable list."
In short, Vietnamese SME data is material that combines static registration-based information with dynamic online signals so that individual businesses can be identified, classified, and assessed. If statistics describe the market's "size," SME data describes the market's "members." And in Vietnam, a large share of those members live outside the formal registry, in a zone where digital and offline are blended.
How to source it
There are four broad routes to source Vietnamese SME data: public registration information, collection of public online signals, field- and network-based collection, and the refinement work that combines all three. No single route produces a usable dataset on its own; in practice you have to cross-reference multiple sources for accuracy to climb.
First, public registration information. Vietnam discloses a portion of business registration data, and from the business-registration portal and public tax-related materials you can obtain basic fields — business name, tax code, address, industry. However, this information often does not immediately reflect dissolution, relocation, or dormancy, so a business being in the registry is no guarantee it is currently trading. Treat registration data as a "starting point" and verify activity separately. A trap you hit often in practice is the "ghost record" — alive and well in the registry but in reality shut down years ago. Build a sales list on those records alone and a large share of your outreach lands on dead numbers and old addresses. Another: household businesses are frequently managed outside the corporate-registration portal, so scraping only company data drops a large slice of the market from the very start.
Second, collecting public online signals. On publicly exposed surfaces — Facebook, Zalo, Google Maps, e-commerce platforms — you confirm a business's existence and activity. Signals like the date of the latest post, review activity, exposed contact details, and updated opening hours indicate whether the business is genuinely operating and how much it uses digital channels. Concretely, a Facebook page's last post date and whether the operator replies to comments tell you "is someone tending this page," while the recency of Google Maps reviews and photo uploads tells you "are customers actually visiting." For a seller listed on Shopee or Lazada, sales history and response rate are strong activity signals. What matters at this stage is handling only already-public information and respecting platform terms and personal-data protection principles. A business contact exposed on a public page and a personal mobile number scraped without consent are entirely different in nature, and the latter should be left untouched.
Third, field- and network-based collection. When you need to drill deep into a specific area or sector, non-public or semi-public routes become decisive: direct verification by local staff, directories by association, chamber of commerce, or market, and customer lists held by a partner. In a market with large offline commerce and a sizable informal sector like Vietnam, many active businesses have almost no digital footprint, so field channels fill the data's blind spots. Traditional-market stalls, tenants of an industrial park, or a cluster of same-sector shops along one street barely register online, yet on the ground you can compile a roster in half a day. Adding just one step — a local-language staffer calling to confirm whether the business operates and who the contact is — sharply raises the data's reliability. In the Vietnamese business environment, human verification is not inefficiency; it is an essential complement that reaches what digital alone cannot.
Fourth, refinement and consolidation. Raw data gathered from the three routes above cannot be used as-is. The same business is scattered across sources under different names and addresses; notation, spelling, and administrative-unit names differ everywhere; duplicates and errors are mixed in. The presence or absence of Vietnamese tone marks (dấu), the mixing of abbreviations with full names, and recent frequent mergers of administrative units make it hard to mechanically decide "same business or different." The pipeline that normalizes names and addresses, merges duplicate records, ties registration data and online signals to a single business, and finally grades activity and reliability — that pipeline is what determines the data's real value. With the same raw inputs, the usefulness of the output varies enormously with the care taken at this refinement stage. Unrefined raw data often gives only the illusion that "we have data" while being unusable for an actual decision.
A point of honesty is in order here. Yeowubie has the capability to source and refine SME data in multiple layers while working in the Vietnamese market, but this is not a productized data API that anyone can call and use; it is closer to an operational capability, structured and verified to each project's context. The expectation of "one button produces a nationwide SME list" does not match the reality of Vietnamese data; a meaningful dataset is produced through a purpose-fit process of collection and refinement. Put differently, the value of data comes not from "how much we gathered" but from "how precisely we narrowed it to purpose."
Use cases
Vietnamese SME data is used most directly in four areas: market entry, sales targeting, partner discovery, and market analysis. The common thread is turning a "vague market" into "an accessible list and structure," enabling decisions based on data rather than guesswork.
First, market-entry judgment. A company seeking to bring a new product or service into Vietnam first needs to know "where, how many, and in what form do the SMEs that could become our customers exist." Looking at the distribution of active businesses in a specific sector or region lets you make a first-order call with data: is this a segment worth entering, which city or district should be attacked first. For example, Hanoi and Ho Chi Minh City differ in business density, digital maturity, and average size even within the same sector. Confirming that difference with data first lets you enter where the real opportunity is largest, instead of defaulting to "start with the capital." Conversely, learning in advance from the data that "there are fewer target businesses than expected" — and thereby saving a wasted entry cost — is an equally valuable outcome.
Second, sales targeting. In B2B, the most expensive cost is "time spent on the wrong target." With a target list filtered by sector, size, region, and digital maturity, a sales team can concentrate resources on high-probability candidates. For example, a group of retailers that run online channels but lack a proper website is a natural candidate for a digital-transformation solution. Conversely, a company selling payment or delivery solutions should prioritize businesses already listed on e-commerce platforms with a steady transaction volume. The point is that the same data is cut differently depending on what you sell. A sales team blindly contacting a thousand businesses fares worse — in both conversion and morale — than one focusing on three hundred filtered by data. In Vietnam the choice of first-contact channel matters especially: some segments respond to a Zalo message, others move only after an in-person visit. If channel-preference signals travel with the data, you can tailor even the mode of approach.
Third, partner and supply-chain discovery. When looking for local distribution, construction, logistics, or content collaborators, business-level data quickly narrows the candidate pool. Screening by size and activity reaches partners you can actually transact with in far fewer attempts than random outreach. A common difficulty foreign firms face in Vietnam is "how do we find a trustworthy local partner," and data systematizes the first stage of that search. Filter by activity signal for "is it actually operating," gauge by size signal "can it handle our volume," and do a first pass on online reputation for "is there baseline trust" — then you spend in-person meeting time only on real candidates. Data does not vouch for a partner's quality, but it pre-screens the clearly unsuitable and raises the density of your meetings.
Fourth, market-structure analysis. Aggregating individual-business data reveals structures that aggregate statistics cannot show. Patterns — which sectors cluster in which regions, which segments are digitizing fast, which areas have brisk new entry — become the foundation for strategy. For instance, the fact that same-sector businesses pack densely into one street or district is itself both a signal that "demand is here" and a warning that "competition is fierce." A segment with rapidly rising new registrations may be in a growth phase; a cluster whose activity signals are cooling may be in decline. Such dynamic patterns are invisible in a single snapshot of statistics; they appear only when you compare business-level data along a time axis.
Real project examples can only be generalized without naming companies. A Korean firm targeting a specific retail sector in northern Vietnam that checked the distribution of active businesses first and then ordered its sales priorities; a company already in the market that narrowed a pool of collaborable small-and-medium suppliers with data — the shared core is that "work that used to begin with guesswork now begins with data." In some cases the data instead revealed that "this segment is smaller than expected or already saturated," prompting a rethink of entry itself. Data itself does not generate revenue, but it raises the quality of judgment about where to spend resources. And sometimes it earns its keep simply by letting you reach a "decision not to" quickly.
Data quality and scope
Data quality is assessed by accuracy, recency, completeness, and the level of activity verification. Scope is determined by which sectors, regions, and sizes are covered, and how densely identifiable businesses sit within that. Quality and scope are in tension; data that hits the peak of both at once barely exists in reality.
Accuracy is the degree to which each field matches reality: whether the name and address are correct, whether the industry classification fits the actual business, whether the contact is still valid. Because of Vietnam's administrative-unit reorganizations and non-standard address notation, address accuracy is a particularly labor-intensive area. The same address floats around under both an old and a new administrative-unit name, and alley-level address notation (street number, alley, ward) differs by source, making it hard to resolve everything to a single point on a map. Industry classification is full of traps too: the registered sector and the actual business often diverge — a business registered as "trading" that in reality runs a café is not rare — so trusting the registered industry code alone misaligns your target. Accuracy verification must therefore include the work of "cross-checking the registered value against the real-world value."
Recency is how well the data reflects the present moment. Registration information is slow to reflect dissolution and relocation; online signals change fast. So data without an explicit "collection date" is hard to trust. A good dataset leaves a last-verified timestamp on each record and re-verifies activity on a cycle. Vietnamese SMEs tend to have short birth-and-death cycles, so the accuracy of year-old data decays faster than we expect. Re-confirming at least the core records right before use is safer than milking a once-built dataset for too long. Without "when was this verified," you cannot even estimate the data's confidence interval.
Completeness is how exhaustively the target population is captured. But in Vietnam, "the whole" is effectively impossible, because there is a significant volume of unregistered and informal businesses, businesses with no digital trace, and micro-enterprises that appear and vanish quickly. So completeness is more realistic viewed by the relative measure "how dense within the defined segment" than by the absolute measure "what percentage of the total." For a narrowly defined segment like "cafés in one Hanoi district," field reinforcement can build fairly dense data; for one as broad as "food service across all Vietnam," any method leaves large gaps. So when you promise completeness, you must always say "completeness in which segment" for it to be an honest promise.
On scope, setting honest expectations is important. Data focused on one city or one sector can be deep and accurate, but data claiming to "cover all sectors nationwide" is likely correspondingly shallow or less verified — because breadth and depth are hard to obtain at once for the same effort. So when evaluating data, ask first "how trustworthy is it in the segment that serves my purpose" rather than "how much is there." Rather than being dazzled by a headline count of hundreds of thousands of records, it is more practical to ask how many of those fall in your target segment and whether they have been verified.
Yeowubie does not hide this limit. What we provide is not a "complete nationwide list," but a practical dataset whose accuracy and activity have been verified within a purpose-defined segment. Agreeing on the trade-off between scope and quality at the start of a project ultimately produces better decisions than inflated promises. A conversation that first settles "how broad, and how deep" is the starting point of good data work.
Cautions
When working with Vietnamese SME data, you must be wary of legal and ethical boundaries, data staleness, source bias, and overconfidence. Data is a powerful tool, but ignoring these four makes it easy to spend money on the wrong list, break a regulation, or arrive at a wrong conclusion.
First, legal and ethical boundaries. Vietnam too is tightening personal-data protection rules, and public business information must be handled differently from personal information. Using business information already exposed on public channels and collecting personal contacts to send messages without consent are entirely different matters. Data collection and use must stay within local law, platform terms, and basic ethical standards. In practice, a safe test is "was this information published for business purposes, or did an individual leave it privately." A shop phone number on a Facebook business page leans toward the former; a mobile number scraped from a personal account is the latter. And for marketing outreach, local rules on unsolicited advertising and spam must also be considered. Even if data was gathered legally, breaking a regulation in how you use it leaves the risk fully intact.
Second, data staleness. SMEs are born and die fast. A business active last year may have dissolved this year; a contact may have changed. Reusing a once-built dataset without verification is risky; placing a step to re-confirm core records right before use is safer. The smaller and younger the business, the more frequent the change, so the higher the share of micro-businesses in a dataset, the more conservatively you should assume its rate of decay. You need the operational habit of explicitly setting "how long is this data valid" and re-verifying once that horizon passes.
Third, source bias. Relying only on online signals skews the sample toward digitally capable businesses; relying only on registration data drops the informal sector entirely. No single source fully represents the market, so when you back a decision with data, you must first recognize what bias that data carries. For example, concluding from Facebook-only data that "this sector is well digitized" may be an illusion created because businesses that do not use digital channels are absent from the sample. You cannot eliminate bias, but knowing its direction lets you correct the conclusion accordingly. An honest data report states "what this data captures well and what it misses."
Fourth, guarding against overconfidence. Data is an input that supports judgment, not judgment itself. Even a dataset that looks sophisticated can paint different pictures depending on sample, timing, and definition. It is especially important not to be seduced by phrasing like "proprietary data." Nowhere in the market is there a single dataset that perfectly contains every SME; practical data means data verified to purpose on a foundation that acknowledges its limits. The word "proprietary" is often packaging that hides a lack of verification, so it is safer to ask how that data was gathered and what it includes and excludes. "Proprietary data" that cannot explain its source and method warrants more suspicion, not less, the larger it claims to be.
Finally, data is a means, not an end. Good data starts from a good question. Define first "what problem are we solving, which segment do we need, what level of accuracy is enough," and only then structure the data — that is how you avoid excess cost or a useless list. The most expensive data is not the data that costs the most, but the data gathered without a purpose and ultimately never used.
If you have a concrete purpose around sourcing, refining, and using Vietnamese SME data, reach out to Yeowubie. Working from a real problem — market entry, sales targeting, partner discovery — we will help you define which data to structure, within what scope, and verified how.