buzzabout docsBrand research
Collect the public posts, comments, and community chatter about your brand into a dataset you can question — then read the earned media that shapes how people see you.
Brand research reads your earned media — the unprompted posts, comments, reviews, influencer shout-outs, and community threads (a subreddit, a YouTube comment section, an X reply) where people talk about your brand without you asking. Instead of guessing from your own campaign metrics, you collect those mentions into a dataset and ask what people actually love, complain about, and compare you to. Throughout this tutorial we'll use a cold-brew coffee brand as the running example: you type the brand as a keyword, preview how much conversation exists, then collect it into a dataset for analysis.
Prerequisites
- Enough conversation to analyse. Meaningful brand analysis needs roughly 100+ mentions. A large, well-known brand clears this easily; a small or new brand may simply have too little earned media to draw conclusions from yet. The preview card tells you how much is out there before you spend anything — if the estimate is thin, widen the timeframe to
Last yearbefore deciding. - Credits to cover the depth you choose (1 mention ≈ 1 credit).
- A Pro or Business plan for multi-source research. Starter collects a single source — but you pick any one of the six platforms, not a fixed one.
Walkthrough
Search a topic or paste a URL
On the home screen, find the composer under the Discover what people buzz about heading. The input placeholder reads Search a topic or paste a URL. Type your brand name as a keyword — for example, cold brew coffee — to read what the open web says about you, or paste a profile URL (a Reddit, X, Instagram, TikTok, YouTube, or LinkedIn account) to analyse the posts on that account.
You can also click a suggested-question chip under a category such as Trends or Competitors to pre-fill a starting query.
Adjust the inline filters
Before sending, you can tune the inline filter chips. The source picker covers Reddit, X, Instagram, TikTok, YouTube, and LinkedIn — multi-select on Pro and Business, single-source on Starter (you still choose which one of the six). The time chip offers Last week, Last 3 months, Last year, and Custom. There are also language and country chips; the language picker has a Find language... search field.
Defaults are: time Last year, language All, and country US. For a smaller brand with thin volume, keep the timeframe at Last year (or widen a narrower one) to gather enough mentions to be worth analysing.
Send message
Press Enter or click the Send message (arrow) button. A new chat is created, you land on the chat page, and your message auto-sends to the assistant.
Keyword query
For a typed topic, the assistant returns a research preview card titled Keyword query, showing an estimated mention count and a volume badge. For our cold brew coffee example, this is where you confirm there's enough conversation to research before spending any credits — if the estimate is well under ~100, widen the timeframe to Last year or broaden the keywords before collecting.
Edit the keyword chips directly to shape what you collect. The add-input reads Enter a keyword query (or Add keyword…), and you press Enter or comma to add a chip. This is where boolean operators come in.
Boolean operators in the keyword query. Operators are AND, OR, NOT plus parentheses for grouping and "double quotes" for exact phrases. The default join between two keywords is OR (it broadens reach) — add AND to require co-occurrence, NOT to exclude. Operators are case-insensitive (they display uppercase). Examples:
"cold brew" NOT decaf— posts mentioning the exact phrasecold brewbut notdecaf."cold brew" AND (subscription OR refill)— cold-brew posts that also mention a subscription or refill.stumptown chameleon "high brew"— any post mentioning any of the three competitors (implicit OR).
Sanity-check complex queries against the live mention-count estimate on the card.
URL query
If you pasted a profile URL instead, the card is titled URL query and reports profiles found rather than a mention estimate — it sizes the work by the accounts you're collecting, not a mention count. There is no Research depth picker on a URL query and no credit badge; the source picker is locked because sources are derived from the URLs you pasted. Use this when you want to read the posts of a specific influencer, community account, or your own brand profile.
Research depth
For keyword queries, refine the preview in place: edit the keyword chips, and change the source, time, language, or country chips — each edit re-generates the estimate. Then pick a Research depth. The options are Overview (100 mentions), Landscape screening (200 mentions), Qualitative research (500 mentions), and Quantitative analysis (1,000 mentions). The helper text reads How many mentions to collect when you run research.
The credit cost updates live as you change depth, since 1 mention ≈ 1 credit. Depth options above your plan cap are disabled: Starter caps at 300 mentions per research (so only Overview and Landscape screening are selectable); Pro and Business reach the full 1,000; Enterprise goes up to 10,000 mentions per research.
Preview mentions
Optionally click Preview mentions to inspect a sample of results in the canvas before committing. This is free — it doesn't spend credits or create a dataset. Use it to sanity-check that the mentions are genuinely about your cold-brew brand and not an unrelated topic.
Start research
When the preview looks right, click Start research. This sends Collect this dataset. with your chosen filters and depth, spends the credits, and creates the dataset run. The preview hides and a dataset card takes over the turn while collection runs.
Show dataset
When the run completes, the collected mentions appear as a mentions table. This is where the downstream Sentiment and Sentiment score columns and the sentiment filter live — they're part of the collected data, not the composer. Use Show dataset to open the full table, then ask follow-up questions in the chat against the dataset.
Now analyse it
Collecting the dataset is the starting line, not the finish. Once your cold-brew mentions are in, the dataset is an extraction menu — open the skills panel (the book icon in the chat input) or just ask the assistant in chat, and pull out the answers that matter:
- Pain points — the recurring complaints and frustrations buried in the mentions.
- Feature requests — what people wish your brand offered.
- Positive and negative quotes — the standout praise and criticism in people's own words.
- Share of voice — your portion of total mention volume against rival brands. Ask the assistant to "compare share of voice across these brands in this dataset" and it renders the comparison as horizontal bars (there's no standalone Share-of-Voice button — it's an analysis you request).
- Audience — who is doing the talking, their interests and brand affinities, and an LLM-inferred personality (OCEAN) read.
See the cross-links below for step-by-step playbooks on each.
Gotchas
- Volume matters. Brand analysis is only meaningful at roughly 100+ mentions. Very small or new brands may have too little earned media to draw conclusions from — widen the timeframe to
Last yearand broaden keywords before deciding there's "nothing there". - Keyword vs URL is auto-detected. A typed topic gives a
Keyword querypreview with a mention estimate, volume badge, and aResearch depthpicker. A pasted URL gives aURL querypreview that reportsprofiles found(not mentions), has no depth picker and no credit badge, and a locked source picker — sources are derived from the URLs. - URL search is profiles, not posts. It analyses the accounts you paste (up to 100), not a single permalink. Paste profile/handle links, not one-post URLs.
- Single-source is not Reddit-only. Starter collects one source per research, but you choose any of
Reddit,X,Instagram,TikTok,YouTube, orLinkedIn. An inlineMulti-source research is available on Pro & Businessnudge appears when you try to add a second. - Plan caps differ from picker options. Starter caps at 300 mentions per research (Overview and Landscape screening only); Pro and Business reach 1,000; Enterprise goes up to 10,000 mentions per research.
- Default join between keywords is OR, which broadens reach. Add
ANDto require co-occurrence andNOTto exclude; wrap multi-word phrases in"double quotes". - Preset time and custom date range are mutually exclusive. Picking a preset clears the custom range and vice versa; custom range is capped to the last 5 years.
Preview mentionsis free;Start researchspends credits. OnlyStart researchcreates the dataset run.- Editing any preview field regenerates the estimate. Changing a keyword chip, source, time, language, country, or URL fires a regenerate; buttons lock while a regenerate is in flight or a message is streaming.
- Sentiment is downstream, not in the composer. It appears afterward as a
Sentiment/Sentiment scorecolumn and a filter on the collected mentions table. - Input is capped at 4000 characters (an
N leftcounter appears past 3000).
Next steps
Competitor research
Run audience analysis
Analyse pain points
Use skills
Connect MCP to Claude
Wire the Buzzabout MCP server into Claude using OAuth for interactive use, or an API key for headless agents.
Competitor research
Scope the standard research flow to competitor keywords or profile URLs, group the datasets in a project, then analyse share of voice, audience, and trends with skills.