gr-user-interview

⚠️ Channel adjustments for 2C products

This skill defaults to dev / B2B channels.2C consumer goods / education / apps: the main battleground for customer acquisition is vertical communities + short videos + vertical KOLs, choose the primary platform based on regional public data (e.g., short videos in Indonesia/Thailand have already overtaken Facebook). For KOLs, prioritize nano / micro verticals—the more followers, the lower the engagement rate; micro > mega offers better value for money. For the complete 2C channel database + public sources see → gingiris-seo-geo/references/2c-adaptation.md。

gr-user-interview — user interviews

When to use

"I want to find PMF, how do I do interviews"
"User feedback is messy, I don't know which to focus on"
"I interviewed 5 people and got no insight"
"How do I avoid users lying"
"I don't know who to interview / can't get people to meet"
"How should activation rate and payment rate be defined and benchmarked"

From interviews to launch acceptance

When collaboration with product and engineering is needed after interviews, read Requirements handoff and follow-up template: separate raw evidence from proposed solutions, categorize by foundational capabilities, onboarding, and monetization, and provide prioritization rationale, owners, observable acceptance results, and follow-up targets. Use different views for bugs and feature requests; do not treat collected opinions directly as delivery commitments.

Core principles (5)

Ask about behavior, not opinion ❌ "Would you use this feature?" → you'll get a polite "yes" ✅ "When was the last time you encountered problem X? What did you do then?" → you'll get real behavior
A 20-minute interview >> 5 feature questions Going deep on one scenario is more valuable than asking 5 shallow questions
Recording / full transcript > taking notes on the spot Taking notes on the spot will miss micro-expressions and tone
Pause after 5 people and debrief Don't wait until you've interviewed 20 people in a row to summarize; after 5 people you've likely already seen the pattern
Mirror their wording Use whatever words the user uses—don't replace them with product language

Interview structure template (20-30 minutes)

Stage	Duration	What to ask
Warm-up	2 min	"What did you do today?" (open-ended, help them relax)
Discover	5 min	"When was the last time you did X? Can you tell me about it?"
Deep-dive	10 min	"What tools did you use then? How did you use them? Where did you get stuck?"
Alternative	5 min	"Have you tried other methods? Why didn't you continue?"
Wrap-up	3 min	"If there were a magic solution to this problem, how would it work?"

Token Top10 interview method (signature play)

Pull the Top10 list by backend usage → shared-screen interview (screen recording required) → feed the transcript to AI to extract features/bugs into Linear → follow up the next day. ≥2 sessions per day.

Rank the Top10 by token consumption / usage, refresh weekly—power users have the highest feedback density
Must share screen + record screen: watching real operations > listening to verbal accounts
Have AI automatically extract feature requests / bugs from the transcript and put them directly into the issue tool
Next-day follow-up: "The X you mentioned has been fixed/scheduled" — the feedback loop is the retention lever
After the interview, clip 5-10 minutes of key segments for product and engineering

Outreach cadence：

Four touches on days 1/3/7/14(Follow up once each on Day 1 → 3 → 7 → 14)
Dual-channel: email template + reverse-lookup email to LinkedIn DM; if no reply to email, follow up on LinkedIn for another round
Use a broad reason in your messaging ("an early-stage team wants to do interviews"), don't reveal the real screening reason — avoid users performing for you

Combo moves：

KOL interview two-in-one: while partnering with a YouTuber, also do user interviews + usability testing — one collaboration, two kinds of value (20-30 minute video interviews have the highest ROI)
Wisprflow playbook: manually onboard and interview the first 500 users one by one before launch, using momentum to drive the launch

Activation benchmarks quick reference

Item	Benchmark
Activation definition	Must be abehavioral threshold: complete ≥2 tasks / ≥20 rounds of conversation for 3 consecutive days / use for more than 5 minutes — not "has logged in"
Registration → payment qualification line	>5%（Pro-C 5-8%、SMB 10-15%）
Churn definition	No return visit within 24h after two conversations = churn → automatically trigger a win-back email
Payment attribution	Cross-reference four timestamps: registration × activation × payment × most recent login

In-depth reference

📂 https://github.com/Gingiris-1031/gingiris-skills/tree/main/skills/gingiris-user-interview

references/templates.md — question templates for different stages (Discovery / Validation / Retention)
HeyGen 937 interview case study breakdown

Synthesis process

Within 24h after each interview, organize 3 cards:
- Pain(the user's real pain)
- Workaround(how they currently make do)
- Quote(quotable verbatim)
Do a retrospective every 5 interviews:
- Pain that appears repeatedly ≥ 3 times → candidate PMF direction
- Workaround that appears repeatedly ≥ 2 times → competitor signal
Make a decision every 20 interviews:
- Continue this hypothesis / change direction / need more data

Cascading recommendations

Interview reveals a new PMF direction → gr-blog-post write a piece (to validate market reaction)
Interview reveals a competitor's name mentioned ≥ 3 times → gr-competitor deep analysis
PMF is clear → gr-b2b-growth or gr-oss-marketing start growth
Need to recruit interviewees → pick readers from gr-seo-patrol 's GSC traffic, or directly use the "Token Top10 interview method" to pull the backend usage leaderboard
Running influencer partnerships → use the "KOL interview two-in-one" to pick up interviews along the way (see gingiris-kol-outreach）
After interviews, need to set activation/payment goals → first lock down the behavioral thresholds against the "Activation benchmarks quick reference" above

Anti-patterns

❌ Asking "would you be willing to pay" → 100% fake answers
❌ Sending a survey instead of interviews (10x worse information density)
❌ Only interviewing existing users and not churned users (churn is the real gold)
❌ An interview longer than 45 minutes (fatigue → perfunctory answers)
❌ Pitching your own product during the interview (contaminates the answers)
❌ Giving up after sending the invitation only once (four touches on days 1/3/7/14 is the baseline)
❌ Invitation script reveals the screening reason ("we're reaching out because your usage is high" → the user starts performing)
❌ Activation defined as "registered/logged in" (it must be a behavioral threshold)
❌ No follow-up after the interview (a next-day follow-up saying "fixed/scheduled" is the cheapest retention action)