B2B segmentation in East Asia: choose a customer group to research
Segment B2B audiences in Korea, Japan and China by need, application and buying route, then test which groups are worth researching.
Article language: English

In this article
“Japanese manufacturers” is such a broad group that it is hard to fit a useful first conversation inside it. Start B2B segmentation in Korea, Japan or China with the buying situation: who has a specific problem, and when do they look for change? Add industry, size and country afterwards. That order helps choose research questions and companies. A market label alone cannot explain why someone would buy your product.
You do not need a dozen colourful personas at the start. You need a few distinctions that change the research and subsequent offer. A useful segment lets you say whom you are looking for, why that company might have a need and what excludes it from further analysis. The following process supports that choice. Every example is a project exercise, not a description of an Aeneia client’s results.
Start with the customer’s task, then the industry label
Write one sentence: a company needs a solution when a particular situation occurs and its current way of working causes a specific difficulty. Instead of “we sell industrial sensors”, try “we are looking for plants that need to check a component’s consistency without stopping the whole line”. The first sentence is about you. The second points to an event you can find in process documentation or discuss with a user.
Now define the application’s boundary. The same sensor might help launch a new line, replace worn equipment or address an immediate quality problem. Each situation can mean a different contact, available time and acceptance criterion. Do not automatically combine them. Broad market research should identify which situations occur; segmentation chooses those worth examining more closely. The guide to research in Korea, Japan and China (in Polish) describes how to frame that work.
Separate statistical classification from a sales segment
An industry code helps find and compare data, but does not describe a need. The United Nations Statistics Division presents ISIC as a classification of productive activities used to collect and report statistics. It organises the economy by type of activity. A category alone does not mean every organisation within it buys the same product, in the same way and for the same reason.
Keep separate spreadsheet fields for the classification used by the source, application description, company role in the value chain and indication of a need. Do not copy a code from one national database to another without checking it. Record the classification edition and observation unit. If one table counts establishments and another counts enterprises, matching category names are not enough for comparison. First establish the relationship between the units and the limitation on the result. For import comparisons, add the distinctions in the guide to HS data (in Polish). A classification is a useful drawer. Do not expect the drawer to conduct the sales conversation.
Choose axes that actually change how people buy
Start with four: application, consequence of the problem, implementation and access to the customer. Application describes the work done with the product. Consequence identifies what happens without a solution. Implementation shows whether installation, training or integration is needed. Access asks whether you can speak with the recipient directly, through an integrator or through a distributor. Each axis should justify a different action. For a service, use the guide to user, buyer and payer roles (in Polish) to check whether those roles belong to the same people or organisations. Separating those roles can change whom you interview within one application.
Location may add context to access. The article on Korean industrial complexes (in Polish) lets you consider a plant’s location alongside its service ecosystem. A company’s presence in such an area still does not confirm a need for your specific solution.
In the sensor exercise, a small plant and a large factory may belong to one segment if they measure the same component and accept results similarly. Two factories with similar turnover may need entirely different offers if one buys a complete station and the other only a component. Company size becomes useful when it explains a difference, not merely when it fills a column neatly.
- A shared problem should lead to a similar research question.
- A shared implementation method should make support requirements comparable.
- A shared buying route should point to similar roles and documents.
- An exception that changes the offer should be visible before contact, if it can be checked.
Describe three working segments instead of one huge group
Describe three working segments in a few sentences each. One might cover customers replacing existing equipment. Another might cover companies starting a new process. A third could cover integrators building solutions for their own customers. Add what you do not know yet: replacement frequency, budget, test method or service responsibility. A proposed segment remains a hypothesis, so allow an answer that disproves it.
Do not immediately assign market shares to the groups. A list of a dozen companies is not a denominator for the whole economy. First compare whether you can find sources, real application examples and people to speak to for each candidate. The guide to comparing a first market (in Polish) can help with country selection. Allow this result too: the problem looks interesting, but we cannot yet reach the right users.
Look for indications in profiles, not declarations of demand
If catalogues use different names from your Polish materials, start with the guide to searching in local languages (in Polish). Check a term’s meaning in the process context. That helps find candidates by actual application and avoids building a group around an overly broad translation.
A product sheet, technology description or company profile may help identify an application. In its J-GoodTech FAQ, SMRJ describes searching companies and viewing profiles; it also describes selecting Japanese SMEs using factors including product and technology characteristics and willingness to expand collaboration. The directory has its own scope and participation criteria. It does not automatically represent every company in Japan.
For a specific profile, record only what the description actually supports. Information about processing a material confirms a declared specialisation. It does not confirm a need to buy your device. Add a question to check for every indication. “The company uses this process” plus “the process may require measurement” is a useful direction for a conversation, but remains the researcher’s inference. Sound competitor analysis in Korea requires the same separation of observation and interpretation. The guide to Korean competitor analysis (in Polish) explains the distinction.

Try classifying real companies
Choose a small set of profiles and ask two people to assign companies to working segments independently. The goal is not statistical representativeness. You are checking whether the definitions can be applied. If one person calls an integrator an end user and another calls it a distributor, clarify the rule. The company may perform several functions and need separate relationship records rather than a forced choice of one label.
Record the reasons for inclusion and exclusion. A profile naming a product is not enough if you do not know whether the company makes it, sells it or merely describes a research project. Do not fill gaps with intuition. Use deduplication rules when organising names and relationships; the guide to Asian company databases (in Polish) explains how to preserve evidence. A segment that appears full of candidates may shrink once repeated entries for the same group disappear.
Compare potential with the cost of reaching and serving customers
A large number of apparently suitable companies is not enough to choose. Ask whether you can deliver a sample, explain specifications, conduct an assessment and handle a post-purchase enquiry. A segment needing a service visit at every installation may be difficult for a team working only from Poland. Another may be smaller but allow a remote material assessment and limited trial.
Do not turn this assessment into a ranking that pretends to be precise. For each candidate, show the argument, evidence and uncertainty. For example: user access looks promising because a relevant industry channel has been found; support cost remains unknown because installation conditions have not been checked. The guide to choosing between a distributor and direct sales (in Polish) explains how the access model can change the segment assessment, so consider the access model alongside the need itself.
Test the segment’s boundary with a conversation that could disprove it
Prepare questions that distinguish cases rather than invite agreement. “When did you last change how you inspect this part?” helps reconstruct a situation. “Are you interested in innovative measurement?” gathers an opinion about an attractive idea. Ask about the previous solution, who initiated the change and the condition without which the purchase cannot proceed. Not every polite reply means a procurement process is open.
In B2B customer interviews in Asia, collect cases that do not fit the segment too. The guide to B2B customer interviews (in Polish) discusses how to prepare such conversations. If every conversation confirms the hypothesis, review your questions and recruitment. You may be speaking only with people referred through one channel. Rejecting a hypothesis can be more useful than an elegant slide about enormous potential. The slide will not pay for later changes to the offer.
When should you split a segment, and when should you stop?
Split a group when a new difference changes the evidence required, route to contact or scope of support. If two subgroups need the same conversation and offer, another division may not help. Do not create a segment because a spreadsheet has a free column. Equally, a shared industry does not justify combining applications where the consequences of an error differ.
A practical test is to draft a short interview invitation for each group. If it needs many qualifications and sentences beginning “depending on the case”, the definition may be too broad. If only the city name changes, the split may be premature. When preparing a brief, use the guide to an Asian market-research brief (in Polish) and record the rule for changing segments so the researcher can improve the hypothesis without pretending the scope was obvious from the start.
Try choosing between two applications of one product
Suppose a Polish company offers a module for measuring a component’s dimensions. In the first application, it goes to a manufacturer that wants to inspect its own line. In the second, an integrator buys it for a station sold onward. The device’s function stays similar, but the assessment changes. The manufacturer will ask about performance in its own environment and error handling. The integrator may need connection design data, update rules and repeatable component supply.
For the first group, prepare an operational-problem hypothesis: the current inspection causes a queue of components or extra work. For the second, test a project hypothesis: the current component makes building the station difficult or needs too many adaptations. These are proposals to test, not established facts. Asking about the last equipment replacement will help in the first conversation. In the second, reconstructing the last project where alternatives were considered may tell you more.
Now add an exclusion condition. Exclude the manufacturer from the current trial if an external company performs the process entirely. Exclude the integrator if it supplies only complete stations from one manufacturer and does not choose components. Those exclusions save time for both sides. The list becomes shorter, but more useful. That is the point: get the right question to the right problem rather than send a catalogue on a tour of random inboxes.
Prepare a segment card someone else can complete
Start the card with a working name and one-sentence definition. Then enter inclusion conditions, exclusion conditions and permitted exceptions. List public indications separately from information requiring a conversation. For the measuring module, a description of the component being made could be a public indication. Whether inspection causes difficulty and whether anyone wants to change it remain questions for a user. Do not move them into the facts column just because they fit the offer.
At the bottom, add three fields: what supports further research, what could block collaboration and what evidence would change your assessment. The product owner should check application boundaries; the sales owner should check access to contacts and the feasibility of the next step. If the two understand the definition differently, fix it before building a large list. An hour’s disagreement about a criterion is better than a series of conversations ending with everyone asking why the meeting happened.
Accept segmentation as a decision about further research
The final result should contain definitions of chosen groups, exclusions, company examples with sources and a plan to test the most important unknowns. Briefly explain why some candidates were deferred. The next team needs to understand the selection, not merely inherit green cells. Keep the assessment date and conditions for revisiting a rejected group, such as the appearance of local service capacity.
To move from general interest in a country to a specific audience, Aeneia Oriens can base its work on defining segments, verifying sources and designing conversations. A product description, an example application and the decision the research should enable are sensible starting materials.










