Customer conversation market research: using your own conversations

Short answer

Your own customer conversations are a strong but biased source for market research. Strong, because customers state their needs, expectations, comparisons with competitors and price sensitivity in their own words, with no survey form. Biased, because they only show the people who write to or call you: those who know you, are already interested, and often have a problem.

Using them well rests on three rules: draw hypotheses from conversations, not market size; split results into segments and write down whom each segment represents; and triangulate important findings — check them against open sources, a survey or sales data.

What conversations show well

• The customer's own language: how they name the problem and the need — valuable for ads and website copy.
• Decision criteria: what customers ask about and compare when choosing.
• Competitor mentions: which competitors, in which context, with which advantage.
• Unexpected use cases: your product used for purposes you never imagined.
• Early signals: shifts in demand or dissatisfaction appear before sales do.

Sample bias: who writes and who does not

In a conversation base, three groups are over-represented and three under-represented. Over: people with problems (they write to complain), the undecided (they ask many questions) and people used to digital channels. Under: satisfied customers (they have no need to write), people who have never heard of you, and those who prefer another channel (such as walking into a store).

Channel bias

Each channel brings a different audience. The age and interest profile of Instagram messages can differ from WhatsApp or phone calls. If 70% of conversations come from one channel, the whole result reflects that channel's audience. Always split results by channel and state the channel shares at the top of the report.

How to frame the research question

1. 1About whom"Prospective customers who contacted you in the last 3 months" — not the whole market.
2. 2About what"Which criteria do they ask about when choosing?" — a concrete question conversations can answer.
3. 3For which decision"To change the website's main message" — how the result will be used.
4. 4Checked with what"Sales conversion, open competitor data, short interviews with 20 customers."

Segmentation

An overall number is often the average of different groups and describes none of them correctly. Look at least at three cross-tabs: channel, new vs existing customer, and product of interest or price segment. In one segment price may be the top criterion, in another delivery speed. That difference is often a more useful insight than the overall number.

How to build an intent split by segment is covered in customer intent analysis.

Triangulation: checking with three sources

• Internal quantitative data: sales, conversion, average order, return rate — does the view in conversations show up in behaviour?
• Open sources: competitors' websites and prices, official statistics, industry reports — the market context.
• Asking directly: short interviews or a short survey with 15–20 customers — reveals reasons conversations do not show.

If a conversation finding is not confirmed by at least one additional source, present it as a hypothesis.

How to read competitor mentions

When customers mention competitors, that is not accurate information about the competitor but the customer's perception. "It's cheaper there" does not prove the price really is lower. Split the competitor field in two: which competitor, and in what context (price, speed, quality, service). It shows not the competitor itself but how the competitor looks in customers' eyes.

Illustrative example

This is an illustrative example. An online course platform wants to open a new track and sees interest in "data analytics" growing in conversations. The segment split shows the interest comes mainly from existing students looking for their next course, not a new audience. A look at open job postings confirms market demand, but there are also many competing courses.

Decision: instead of an ad campaign for a new audience, a follow-on course is offered to existing students and the hypothesis is tested by sign-ups.

A research summary format

When presenting a market finding from conversations to leadership, write every point in the same structure so readers do not confuse a hypothesis with a fact:

1. 1FindingOne sentence that includes "in conversations": "in conversations, price is the top criterion in the Instagram segment".
2. 2Who is representedPeriod, number of conversations, channel shares, new vs existing customers.
3. 3EvidenceShare, trend and 5–10 conversation links.
4. 4Confirmation statusHypothesis / confirmed by one source / confirmed by two sources.
5. 5Recommended stepA test, further research or a decision.

Ethics and privacy

For market research, aggregate results are enough. When using individual customer quotes in an internal report, remove names, phone numbers and other identifiers. Before using conversations for a purpose different from the one they were collected for, check your privacy policy and the applicable law (in Azerbaijan the Law on Personal Data; GDPR for EU customers) with a lawyer.

Typical mistakes

• Presenting a share of conversations as a market share.
• Applying one channel's result to all customers.
• Putting only interesting quotes into the report.
• Taking what a customer said about a competitor as fact.

What Vexvon provides for research

In Vexvon the company builds fields such as decision criterion, competitor mention and product of interest, and AI picks from the list using the customer's words in each conversation. The panel's AI assistant gives channel × field cross-tabs, period comparisons and the conversation list behind each value; results can be shown as charts and saved as a PDF. Data such as market size and competitor prices is not in Vexvon — it has to come from outside sources. More: Vexvon analytics.

Next step

For your next marketing decision, write one research question using the template above and plan to check the conversation finding against at least one more source. To separate new demand signals, see voice of customer product insights; we can build the research fields together in a demo.

Further reading on this topic: customer demand trends.