How do I get you to scan social media for my target users?

The short answer: describe the demographic you want to understand, AI runs 5–10 rounds of public-post scanning over 1–2 days, and you get a structured profile plus 3 custom personas you can immediately interview. The output grounds your research in what real users are actually saying, not what you imagine they might say.

When Scout is worth running

Scout is best when you need a baseline of what's actually happening in a market:

"I want to launch a product for a demographic I don't know yet"
"I want to understand how people in Chengdu talk about coffee differently from people in Beijing"
"I want to see what users complain about with my competitor"

Scout is overkill when you already have customer interviews and you just need to ask your known users more questions — go straight to Interview with personas built from your own data.

The 3-step flow

Step 1 — Describe who you want to scan. Plain language, 1–2 sentences.

Good descriptions name a demographic, a topic, and a goal:

"Scan users discussing sparkling coffee on Xiaohongshu, understand their attitudes toward this product category."

"Scan users in Chengdu discussing coffee consumption, understand differences from Beijing users."

"Scan 25–35 year-old women discussing healthy snacks, understand their consumer psychology and pain points."

Vague descriptions waste Scout rounds:

"Scan young people" — which age, what content?

"Scan users who bought our product" — they're not on social media saying so.

"See how the market is" — toward what question?

Step 2 — AI scans. Runs in the background, typically 1–2 days. You can close the tab.

Scout pulls publicly posted content from Xiaohongshu, Weibo, Douyin, Bilibili, X/Twitter, Instagram, TikTok, depending on your demographic. After ~5 rounds it starts forming hypotheses about your audience. After 5 more, it validates those hypotheses.

Step 3 — View output. Three things:

Demographic profile — age, region, income, attitudes, behaviors, pain points
Verbatim highlights — direct quotes from real posts, attributed by source
3 custom AI personas — built from the scan, ready to use in Interviews or Discussions

What the demographic profile actually contains

It's structured, not narrative:

Demographics — age range, gender split, region, income
Psychological traits — values, attitudes, lifestyle orientation
Behavioral patterns — when they buy, what they read, what influences decisions
Pain points — what frustrates them about current options

A typical profile is dense but short — one screen of structured bullets. The verbatim highlights are where the texture is.

When to use the personas vs throw them away

The 3 personas from a Scout run are built specifically for the topic you scanned. They're useful when:

Your scan surfaced clear segments (e.g., "price-sensitive" + "health-anxious" + "social" all emerged naturally)
You want to validate the scan's findings with a deeper Interview round
You want to test a concept specifically with this audience

Throw them away (or just don't use them) when:

The segments felt forced or arbitrary
You only needed the profile and verbatim highlights
You'll build personas from your own customer data instead

Three realistic Scout scenarios

New product testing. A coffee startup wants to launch a zero-sugar sparkling coffee. Scout scans Xiaohongshu, Weibo, Douyin for "sparkling coffee" discussions over 1–2 days. Output: 25-32 year-old women in tier-1 cities who care about appearance and social sharing, ¥25-28 sweet spot, worried about carbonation being too strong. 3 personas built from these findings can immediately interview-test packaging, pricing, marketing.

Regional expansion. A Beijing coffee brand wants to enter Chengdu. Scout scans Chengdu users. Output: Chengdu users are more price-sensitive, prefer "sit and chat" over takeout, like tea-coffee blends, drink 2–3x/week vs Beijing daily. Recommendation: bigger stores, tea-coffee line, ¥20-25 pricing.

Competitor research. A fitness app wants to understand Keep's users. Scout scans Keep-related posts. Output: Keep's core is fitness beginners (60%+), chose it for "many courses" and "free," complain about course quality, ads, lack of personalization. Differentiation opportunity: target intermediate-advanced users, premium curated courses, personalized plans, ad-free.

What Scout does and doesn't do

Does:

Scan publicly posted content across major Chinese and global social platforms
Find real user voice — actual quotes you can use in reports
Surface segments and themes you didn't expect
Build personas grounded in observed behavior, not generic profiles

Doesn't:

Access private accounts, DMs, or closed groups
Pull historical archives (only recent posts)
Replace statistical surveys (sample sizes are smaller)
Tell you what's statistically true about a population

Optional controls before Scout runs

You can pre-configure Scout with three settings:

Platforms. Default: AI picks based on demographic. Manual: specify Xiaohongshu for young women, Douyin for lower-tier markets, X/Twitter for tech circles, etc.

Rounds. Default: 10 rounds. Fewer (5) for quick understanding; more (15) for niche or complex audiences. Beyond 20 has diminishing returns.

Focus dimensions. Default: broad scan. Specify a focus like "price sensitivity," "usage scenarios," "competitor comparisons," or "pain points" to weight Scout's analysis.

Mid-scan adjustments

You can intervene while Scout runs:

Pause. Progress is saved; resume later.
Add rounds. "Run 5 more, focus on price discussion." Scout continues from where it left off.
Shift focus. "I'm more interested in younger users now." Scout re-weights subsequent rounds.

You can also ignore Scout mid-run and check back when notified.

Re-running Scout over time

Markets shift. Re-run Scout every 3–6 months on the same topic:

Health snack users in October may focus on "low sugar low fat"
Health snack users in January may have moved to "emotional value" and "healing"
A 6-month gap between scans can flip your findings

Use re-runs to validate strategy decisions against current sentiment, not against outdated understanding.

What Scout can't reveal

What people do offline. Posts show opinions and stated behavior, not always actual behavior.
Private discussions. No DMs, no closed groups, no logged-in-only content.
Long-tail niches. A group too small to generate public posts won't surface even if it exists.
Causal relationships. Scout finds correlations in what people say. It doesn't prove that X causes Y.

For "why does X happen" questions, follow up Scout with interviews.

A realistic cost/time picture

Setup	Time	Scout output
5 rounds, 1 platform	Half day to 1 day	Surface-level profile, basic verbatim
10 rounds, 2–3 platforms (default)	1–2 days	Solid profile, 5–10 highlights, 3 personas
15 rounds, multi-platform	2–3 days	Deep profile, 10–20 highlights, segmentation detail

The default (10 rounds, AI-chosen platforms) is right for most studies. Bump up for niche markets or B2B audiences. Drop down for time-sensitive quick checks.

Common mistakes when running Scout

Too narrow. "Scan users who bought my product" — these users aren't publicly identifiable on social media.

Too broad. "Scan consumers" — AI will pull from everywhere, output will be too generic to be useful.

Trusting percentages. Scout reports "60% of users say X" based on observed posts — that's a directional read, not a population statistic.

Skipping the verbatim. The structured profile is a summary; the verbatim highlights are the texture. Read both.

Never re-running. A 12-month-old Scout output may describe a market that's already moved.

Related: Plan Mode, Interview, Discussion, Persona Library