Run audience analysis

Build an audience dataset from a collected dataset, then read demographics, interests, brand affinities, creator tier, and an inferred Personality (OCEAN) fingerprint for the people behind the posts.

A collected dataset tells you what people are saying about your e-bike DTC brand. Audience analysis tells you who they are. From the same posts, buzzabout walks the authors and commenters and profiles them — location and language, content niche, interests, brand affinities, creator tier, a marketing-focused summary, and an inferred "Personality (OCEAN)" fingerprint. Instead of guessing your buyer, you get a profile of the real people in the conversation, with directional guidance for who to target and how to message them.

Prerequisites

A dataset with a completed collection run, so there are authors and commenters to profile. Audience analysis is built from a collected dataset — you can't profile an audience from scratch. See Set up your project and Create a mention view.
A Pro plan or higher. Audience analysis unlocks at Pro+.
Credits to spend. Each profile costs 3 credits, and you choose how many profiles to collect — so budget for the run before you start.

Walkthrough

Open a chat against your dataset

Open a chat scoped to the dataset that holds your e-bike mentions — from /chat or from your project's chat. There is no top-nav "Audience" tab: audience analysis lives inside a chat conversation and opens as a canvas overlay, so start from the same place you ask any other research question.

Ask the assistant to profile the audience

Type a plain-language request, such as Profile the audience for this dataset or Who are the people talking about our e-bikes?. There is no "Audience" button — you describe what you want and the assistant prepares an audience run over the dataset's authors and commenters.

Audience analysis

The assistant renders an Audience analysis card in the chat thread. It shows the read-only Source dataset it'll profile and an Audience size picker — four preset buttons: 100, 250, 500, and 1000 profiles. Each button shows its credit cost at 3 credits per profile, so 250 profiles reads 750 credits. The default selection is 250 (your last choice is remembered for next time); higher options can be locked on lower plans, with a Upgrade your plan to analyse this audience size. tooltip.

Collect Audience

Pick the size you want and click Collect Audience. The total credit cost shows next to the button before you commit. The assistant then kicks off the run at the count you chose — there's no separate confirmation step.

Audience research in progress

The card switches to Audience research in progress and works through the pipeline steps. This is asynchronous and can take several minutes — it reads the dataset's deduped posts, walks the authors and commenters, and profiles each one until it reaches your target count. The card updates to Audience research complete when it finishes (or Audience research failed if something went wrong). Use Show settings on the card to see the run's configuration.

View profiles

When the run completes, the card shows a View profiles button labelled with the actual profile count — for example View 248 profiles. Click it to open the audience canvas — a right-panel overlay with a sortable, filterable table of every profiled person.

Sort and filter the audience canvas

The canvas table lists each profile with their computed fields — follower count, engagement rate, and creator tier (civilian, nano, micro, mid, macro, mega). Sort by score, follower count, engagement rate, or authenticity, and filter to narrow the audience: by source platform, verified status, location country, creator tier, content niche, or interest clusters. This is how you isolate, say, the micro creators in your e-bike niche from the civilian commenters.

Click any row to open the profile detail modal. Alongside the person's identity and metrics, you'll see the LLM-derived layer: Content niche, Interests, Brand affinities, a marketing-focused summary covering lifestyle, purchasing habits, and brand affinity, plus per-platform chips (for example Top subreddits on Reddit, Tagged locations and Collaborations on Instagram, or Skills and Education on LinkedIn). These are the fields you use to target: who else they follow, what they care about, and where they fit.

Read the Personality (OCEAN) bars

Many profiles include a Personality (OCEAN) section: five bars for the Big Five personality traits — Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. Buzzabout infers these from each person's public posts and profile via an LLM, then shows them as percentages. Use them as a directional creative signal: a high-Openness segment tends to respond to novel, experimental angles, while a high-Conscientiousness segment leans toward proof, reliability, and spec-led framing — useful when you're deciding how to pitch your e-bike's range, build quality, or design.

Grow the audience later

To add more people to an existing audience, use Collect more profiles. It's a free-numeric input — pick any count from 50 to 5,000 (it defaults to 100), still at 3 credits per profile — and the new profiles are appended to the same audience. Use this when an early 100- or 250-profile pass looks promising and you want a deeper read of the same conversation.

Now you have a built audience dataset. Now analyse it → skills. To roll individual profiles up into something you can present, apply an audience skill in chat: Audience demographic chart summarises the demographic spread, and Audience psychographic chart plots the OCEAN traits across the audience. Both run over the audience dataset you just built.

What OCEAN is — and what it isn't

OCEAN (the "Big Five") is a common framework for describing personality across five dimensions: Openness (curiosity, openness to new experiences), Conscientiousness (organisation, reliability), Extraversion (sociability, energy), Agreeableness (warmth, cooperation), and Neuroticism (emotional sensitivity, stress reactivity). In buzzabout these scores are inferred by an LLM from a person's public posts — a directional AI estimate, not a validated psychometric test. Treat them as a hint about messaging tone, not a clinical assessment, and pair them with the harder targeting signals on each profile: interests, brand affinities, and creator tier.

Gotchas

You choose the audience size — the default is 250. The Audience analysis card offers preset sizes 100 / 250 / 500 / 1000, pre-selected at 250. There is no hidden "850 by default" in the app — 850 is only an internal fallback the assistant uses if you trigger a run purely conversationally without naming a count.
OCEAN is an AI estimate, not a psychometric test. The five traits are inferred by an LLM from public posts. They're a directional signal for creative and messaging, not a measured personality score — don't treat them as ground truth.
OCEAN may be absent for low-signal profiles. When a person hasn't posted enough for a confident read, the model returns no scores and the entire Personality (OCEAN) section is hidden for that profile. Expect plenty of profiles with no OCEAN bars.
The bars read as percentages. In the profile modal each OCEAN trait is shown as a percentage — they are not scored "out of 5".
You need a collected dataset first. Audience analysis profiles the authors and commenters of an existing dataset. Run a collection before you ask the assistant to profile the audience.
It's a Pro+ feature and it costs credits. Each profile is 3 credits, so a 250-profile pass is 750 credits. Higher presets can be locked on lower plans. Confirm your plan and credit balance before kicking off a run.
There's no standalone audience page. Audience analysis is triggered in chat and opens as a canvas overlay — there is no top-nav "Audience" route to navigate to.

Doing this over MCP

If you're driving buzzabout from Claude or another MCP client, the same flow is available as tools: buzzabout__create_audience_dataset to create the audience container, buzzabout__create_audience_dataset_run to run it against your source dataset, poll buzzabout__get_audience_dataset_run until it completes, then buzzabout__list_audience_profiles to read the profiles. Over MCP you pass total_profile_count explicitly (1–2000); when you omit it the agent falls back to its internal default rather than the app's 250 card. See Connect MCP to Claude and the audience profiles endpoints.