“Yes, we would be interested in a localised version.”
That line can make a research call feel successful. It can also mean: perhaps later; if our legal team approves; only for one internal cohort; if someone else runs the platform; and not at the price or delivery format you have in mind.
The interview did not fail. The note failed if it records only “interested.” A buyer answer becomes useful evidence when its conditions, the speaker’s role, counterevidence and the next test are kept beside the words.
For a China-facing expert product, use a five-field buyer-answer log. It will not manufacture demand. It will make it much harder to confuse courtesy, curiosity and a conditional operational need.
The quick answer
After every exploratory interview, record five separate things: the answer, the conditions attached to it, the speaker’s role, counterevidence or uncertainty, and the smallest next test. Do not promote an answer to a market conclusion until the conditions can be tested by the people who would actually buy, operate, learn from or approve the product.
The key question is not, “Did they say yes?” It is, “What would have to remain true for that yes to mean anything?”
The five fields
| Field | What to write | Why it matters |
|---|---|---|
| Answer | The closest accurate wording, marked as paraphrase or quotation. | Stops the note from becoming more confident than the conversation. |
| Conditions | Timing, audience, existing tools, budget boundary, approvals, exclusions or dependencies named by the person. | Most apparent demand lives or dies here. |
| Speaker role | What this person can observe, decide, buy, operate or veto. | A learner, sponsor and technical operator may all give valid but different answers. |
| Counterevidence | A hesitation, alternative, contradiction, missing fact or reason the answer might not generalise. | Prevents a polite positive response from becoming the only record. |
| Next test | One bounded action that could change the current confidence level. | Converts conversation into a falsifiable decision rather than a CRM decoration. |
These five fields must remain separate. When they are merged into a paragraph, the conditions tend to disappear first.
A concrete scenario: “We need an AI course”
Consider an overseas AI educator speaking with a manager at a China-based team. The manager says, “We need an AI course for our analysts.”
A weak note reads: Strong China demand for an AI analyst course.
A useful log reads differently:
| Field | Record |
|---|---|
| Answer | “The manager described a need for analysts to review AI-generated research summaries.” |
| Conditions | A small pilot is only relevant if it uses the team’s approved source materials; external tools cannot be assumed; the manager expects the internal analyst lead to review outputs. |
| Speaker role | The manager coordinates the team and can describe the workflow, but does not own purchasing or data-access approval. |
| Counterevidence | The team may need an internal workflow note, not a course; no learner group, timeline or purchasing route was confirmed. |
| Next test | Ask the analyst lead to review one bounded exercise brief and identify what a usable learner output would contain. |
Nothing in this record is less useful because it is conditional. It is more useful because the next conversation has a job.
Record the answer before interpreting it
Write the answer as close to the person’s language as possible. Mark whether it is a quote, a paraphrase or your synthesis after the call. Do not write “the buyer needs” when the person said “we may need” or “my team has discussed.”
This is especially important when English is not the speaker’s first language, when interpretation is involved, or when the call shifts between a product discussion and a relationship-building discussion. The record should survive a later rereading by someone who did not attend.
One practical technique is to use two lines:
Observed answer: The participant said the current English material is difficult for new hires to apply.
Editor’s interpretation: A narrow localised practice module may be worth exploring, subject to the learner lead confirming the task.
The first line is evidence. The second is a hypothesis. Keeping them apart protects both.
Conditions are the centre of the log
An unqualified “yes” is usually not the decision. The conditions are.
Suppose a creator says they would consider a China version of a cohort course. Relevant conditions might include whether they can approve edited examples, whether the initial audience is an existing community or a new audience, how a learner receives access, and whether the creator is willing to appear live. These are not implementation details to collect after the commercial conclusion. They determine which offer, if any, exists.
Capture the condition in a form that someone could later test:
- Not: “They care about quality.”
-
Better: “They want the first Chinese exercise to retain the original risk warning and be reviewed before release.”
-
Not: “They need support.”
- Better: “They asked whether one named internal operator can receive an escalation note during a two-week pilot; no learner help desk was discussed.”
The second formulation has edges. It can be accepted, rejected or redesigned. That makes it research rather than a vague persona description.
The speaker is not the whole buyer
In a course pilot, at least four roles can matter: the sponsor who sees the problem, the operator who configures the route, the learner who performs the task and the reviewer who decides whether an output is adequate. One person may hold more than one role. Do not assume they do.
This is why a positive statement from a sponsor should not be recorded as a complete operational finding. A sponsor may be able to approve exploration but have no view on access. A technical operator may describe a constraint but have no authority to purchase. A learner may articulate the best exercise need but not know the acceptance rule.
Public platform materials can remind a team that content, users, feedback, orders and channels are distinct operating surfaces. Xiaoe-Tech’s merchant instructions show those distinctions on its own interface. The page does not tell us who owns those decisions in a particular buyer organisation. The interview log must do that work.
Counterevidence is not an embarrassing footnote
The counterevidence field is where useful research becomes less theatrical. Add the sentence that made you pause. Record the alternative the buyer already uses. Note a person who was absent from the call. Keep the condition that would reverse the recommendation.
| Answer received | Counterevidence to preserve |
|---|---|
| “A Chinese version would help.” | The group has not named a cohort or agreed that translation, rather than a workflow redesign, is the obstacle. |
| “We want a workshop.” | The request may be for an internal diagnosis and facilitation, not a repeatable learning product. |
| “We would pay for a pilot.” | The participant could not confirm a budget owner, procurement route or deadline. |
| “A live session is useful.” | The team may only want a one-off Q&A after self-paced material, not an ongoing cohort. |
Counterevidence does not cancel the answer. It tells the next researcher where a false positive could be hiding.
Design one next test, not a complete programme
An interview should end with a question that can be made smaller, not with a premature plan for a full China launch.
If the condition is source control, the next test might be an approved one-page exercise source pack. If the condition is learner relevance, it might be a single 30-minute task reviewed by the prospective learner lead. If the condition is delivery, it might be a limited access-path test with a nominated operator. The end-to-end access, payment and delivery test script is relevant only when the question has genuinely reached that operational stage.
The test should be capable of producing a result that changes the plan. “Continue to discuss” is not a test. “Ask three named roles to confirm their respective responsibilities before creating the course card” is.
What public material is still good for
Public material belongs before and after an interview, but it should not overwrite the interview answer. It can help you understand terminology, assemble a neutral background brief or test whether a claimed feature is described on an official page.
For example, China Government Procurement Network guidance explains that objectives, audience, content, frequency or duration and assessment may be specified separately in procurement needs. Zhihu’s older commercial-content preparation guide also distinguishes several context areas such as industry, product, competitors and audience. These are useful prompts to check whether a research note has flattened context.
They cannot validate a private buyer’s claim, establish demand or tell you what will happen in a particular organisation. For that, return to the person, the relevant role and the smallest possible test.
When this method is not necessary
If the only work is documenting an official product field, cite the official source and do not pretend it is interview evidence. If you have a signed, specific buyer requirement with clear scope and an authorised record, you may use the buyer’s own requirement rather than running an exploratory interview log.
The five-field log is for the uncertain middle: the point where a plausible conversation can either become a disciplined test or harden into a story too early.
What the log cannot prove
One or even several conversations cannot prove national demand, buying authority, platform eligibility, a fair price, legal fit, a successful delivery path or a learning outcome. The log also cannot replace permission to quote someone, rights review or a representative research sample.
Its value is narrower. It preserves what a real person said, the conditions under which they said it, what could disconfirm the interpretation and the next action that might improve confidence.
Use When Topic Popularity Is Not Buyer Demand when a visible subject is being mistaken for a buyer signal. If the conversation starts specifying deliverables, compare it with the China training procurement brief guide so a service request is not silently rewritten as a ready-made course.
Sources
- Xiaoe-Tech merchant instructions
- Zhihu commercial-content preparation guide
- China Government Procurement Network guidance on procurement needs
A bounded next step
If you have a real buyer conversation and need to decide whether it justifies a China-facing test, request a private China validation review. The useful output is a bounded next test with named uncertainties—not a market claim based on one encouraging call.