# Social listening: what it is and how to use conversations > Find questions, objections and problems that can inform your content. Record signals, check their context and decide what to do with them. **Social listening is the analysis of conversations about a brand, product or category to understand questions, needs and perceptions that can inform decisions.** It requires a defined question, a clear account of the sources you can observe and a way to check an interpretation before acting. You can start with comments on your posts, public mentions and conversations you are authorised to review. If your question involves activity outside your accounts, you will need additional sources and access. None of these samples automatically represents everyone who might buy. ## Listening, responding and measuring do different jobs Responding to a comment addresses a particular situation. Monitoring mentions reveals that a conversation exists. Social listening compares and contextualises observations to inform a decision: which explanation is missing, which problem needs attention or which hypothesis deserves investigation. Post statistics help you observe distribution and interaction. Comment text adds information about what people understood or wanted to ask. An engagement rate alone does not explain those reasons. Use the definition of [social media engagement](/en/guides/what-is-social-media-engagement/) to keep the measurements distinct. The American Marketing Association includes interpreting conversations without direct tags and prioritising by impact alongside frequency in its social-listening training topics. That describes the professional scope of the subject, rather than proving results for a particular brand. [AMA training programme](https://www.ama.org/event-agenda/virtual-training/smarter-social-listening-customer-insights-that-drive-business-decisions/). ## Begin with a question you can investigate “What does the internet think of us?” needs coverage that is difficult to define. “Which delivery questions appeared in our comments this week?” identifies a source, period and subject that your team can review. Define the unit before counting. A thread may contain several replies from one person and several questions. If you count conversations, keep an identifier per thread and avoid counting its replies as new conversations. If you tag a conversation under more than one theme, state that the groups overlap; their percentages do not have to add up to one hundred. Record what the sample includes: accounts, languages, dates and sources. Note exclusions such as unavailable messages, deleted comments or conversations on other platforms. These limits make findings more usable; a larger chart will not remove them. ## Classify observations by the work they require Read the full message and the content it responds to. An isolated word may mean something different in context. “Great, still no reply” can express a complaint despite the positive word. Automated labels need review when that distinction affects a decision. Separate information questions, pre-purchase objections, incidents requiring attention and suggestions. Add a specific theme such as delivery, use, price or conditions. A serious incident may require immediate attention even if it appears once. Frequency describes the sample; priority also depends on impact and who can resolve the issue. | Observation | What to check | Possible action | | --- | --- | --- | | Will it arrive before Friday? | Destination, date and actual delivery conditions. | Answer the case and explain timing in relevant content. | | How do I clean this item? | Material, care instructions and their source. | Prepare a demonstration that answers the question. | | My order arrived broken. | The case through an authorised channel, without requesting private details publicly. | Refer it to support and record the problem for internal review. | | I wish there were another size. | Whether this is an isolated request and which variants exist. | Keep it as a product signal and check against other sources. | ## An example that keeps comments distinct from the market In the fictional **South Objects** case, the team reviews twelve separate threads between 21 and 27 September 2026. The [social-listening worksheet](/blogs/research-next-four/listening-en.csv) contains twelve invented observations: eight about delivery, three about care and one incident. Each thread has one main theme so the exercise can be checked. The useful observation is that eight of the twelve reviewed threads ask delivery questions. You cannot conclude that two thirds of customers worry about delivery: the sample includes people who chose to comment and a selection of sources. You also do not know whether another explanation would have prevented those questions. The team confirms conditions with operations, answers open cases and prepares a post explaining how to check timing before buying. In the next period, it reviews new questions of the same kind and records how many conversations it could observe. Changes in channel, exposure or selection are documented before comparison. ## Record the signal and next action The worksheet keeps the period, identifier, source, paraphrase, theme, missing evidence and next action. Avoid placing names, email addresses or order details in a shared editorial file. Case references belong in the appropriate system with the appropriate access. For each finding, write the observation and interpretation separately. “There are eight delivery questions” describes this exercise. “A visible explanation of timing is missing” is a hypothesis that requires checking existing content and service conditions. Assign an owner to the next step and agree when to revisit it. The [WHO and UNICEF listening methodology](https://www.who.int/news/item/06-07-2023-introducing-rapid-social-listening-and-infodemic-insights-for-action-who-and-unicef-launch-manual-on-6-steps-to-build-an-infodemic-insights-report) also organises analysis around a question, sources and actions. Its field is public health. This article’s worksheet is our own editorial approach to brand content; it does not transfer health findings to marketing. ## What you can analyse through HeyMark The [HeyMark MCP](/en/mcp/) can read stored comments, conversations and statistics within your brand’s permissions and available data. Those records can help you review questions on your accounts and relate them to particular posts. That scope does not establish coverage of all external mentions, access to private groups or tracking of arbitrary internet profiles. To investigate an entire category, check which additional sources are required and what your chosen tool can access. An AI summary should retain references and identify what it could not observe. #### Analyse questions in a defined sample ```text Brand: [name]. Question: [what we want to understand]. Period: [dates]. Authorised accounts and sources: [list]. Review available conversations and their post context. Define the counting unit, avoid duplicates, and distinguish questions, objections and incidents. Summarise themes with examples without personal data, checkable references and coverage limits. Separate observations, hypotheses and actions. Do not invent conversations or extrapolate to the market. Do not reply or publish. ``` ## What should change after listening Choose a specific action: answer a question, correct information, refer an incident or investigate a request. Connect the finding to the [strategy plan](/en/guides/social-media-strategy-businesses/) and record how the resulting content will be evaluated. Listening should lead to a decision you can revisit, rather than a list of isolated words. The examples, categories and CSV are HeyMark’s editorial work. South Objects and its twelve threads are fictional, with no actual customer results. The cover is an AI-generated illustration. ## Plan your next post in HeyMark. Keep the idea, review the draft with your team, and see how it performed in the accounts you connected. [Start free](https://app.heymark.ai) · [See how it works](https://heymark.ai/en/#product) ## Developer and agent resources - [HeyMark MCP documentation](https://heymark.ai/en/mcp/) - [llms.txt](https://heymark.ai/llms.txt) - [Full site content for language models](https://heymark.ai/llms-full.txt) - MCP protocol endpoint: `POST https://mcp.heymark.ai` - OAuth protected-resource metadata: [/.well-known/oauth-protected-resource](https://mcp.heymark.ai/.well-known/oauth-protected-resource) - MCP server card: [/server-card](https://mcp.heymark.ai/server-card)