Gingiris commited on
Commit
de3f89c
·
verified ·
1 Parent(s): 2ee8701

Update gr-user-interview to 0.3.0 with evidence-led workflows

Browse files
LICENSE ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ MIT License
2
+
3
+ Copyright (c) 2026 Iris / Gingiris
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
README.md ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ - zh
6
+ tags:
7
+ - agent-skill
8
+ - gingiris
9
+ - workflow
10
+ pretty_name: gr-user-interview — Evidence-led workflow
11
+ ---
12
+ # gr-user-interview
13
+
14
+ Version 0.3.0 · Updated 2026-11-08.
15
+
16
+ 用户访谈与需求验收框架,覆盖原始证据、方案、优先级及回访交接。按行为和复测验证需求,不以访谈场数证明 PMF。覆盖访谈招募、问题设计、 访谈执行、数据合成、决策驱动。从 gingiris-user-interview 提炼。
17
+
18
+ ## Install / 安装
19
+
20
+ ```bash
21
+ npx skills add Gingiris-1031/gingiris-skills --skill gr-user-interview
22
+ ```
23
+
24
+ Read [SKILL.md](SKILL.md) and its linked references for the complete workflow.
25
+ 完整方法与模板请阅读 SKILL.md 及其引用文件。
26
+
27
+ This update separates observations, hypotheses and outcomes. It contains reusable methods and constructed examples, not private workshop recordings or student records. No guaranteed business outcome.
28
+ 本次更新区分观察、假设与实际结果;仅包含通用方法与教学示例,不含私人逐字稿、学员记录,不保证商业效果。
29
+
30
+ Source: [maintained GitHub package](https://github.com/Gingiris-1031/gingiris-skills/tree/main/skills/gr-user-interview).
SKILL.md ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ name: gr-user-interview
3
+ description: '用户访谈与需求验收框架,覆盖原始证据、方案、优先级及回访交接。按行为和复测验证需求,不以访谈场数证明 PMF。覆盖访谈招募、问题设计、 访谈执行、数据合成、决策驱动。从 gingiris-user-interview
4
+ 提炼。
5
+
6
+ '
7
+ metadata:
8
+ author: Iris / Gingiris
9
+ version: 0.3.0
10
+ source: https://github.com/Gingiris-1031/gingiris-skills/tree/main/skills/gr-user-interview
11
+ tags:
12
+ - user-interview
13
+ - user-research
14
+ - pmf
15
+ - product-market-fit
16
+ - jtbd
17
+ - jobs-to-be-done
18
+ - customer-discovery
19
+ - qualitative-research
20
+ - startup
21
+ - claude-code
22
+ - ai-agent-skill
23
+ - agent-skill
24
+ - latest
25
+ ---
26
+
27
+ ## ⚠️ 2C 产品的渠道调整
28
+
29
+ 本 skill 默认 dev / B2B 渠道。**2C 消费品 / 教育 / 应用**:获客主战场是 **垂类社区 + 短视频 + 垂直 KOL**,按地区公开数据选第一平台(如印尼/泰国短视频已反超 Facebook)。KOL 优先 nano / micro 垂类——粉丝越多互动率越低,micro > mega 性价比更高。完整 2C 渠道数据库 + 公开来源见 → `gingiris-seo-geo/references/2c-adaptation.md`。
30
+
31
+ ---
32
+
33
+ # gr-user-interview — 用户访谈
34
+
35
+ ## 什么时候用
36
+
37
+ - "我要找 PMF,怎么做访谈"
38
+ - "用户反馈很杂,不知道抓哪个"
39
+ - "我采访了 5 个人,都没 insight"
40
+ - "怎么避免用户说假话"
41
+ - "不知道该访谈谁 / 约不到人"
42
+ - "激活率、付费率该怎么定义和对标"
43
+
44
+ ---
45
+
46
+ ## 从访谈到上线验收
47
+
48
+ 访谈之后需要与产研协作时,读取 [需求交接与回访模板](references/research-to-acceptance.md):分开原始证据和拟议方案,按基础能力、onboarding、商业化归类,给出优先级理由、负责人、可观察的验收结果和回访对象。bug 与 feature request 使用不同视图,不能把收集到的意见直接当成交付承诺。
49
+
50
+ ## 核心原则(5 条)
51
+
52
+ 1. **问 behavior,不问 opinion**
53
+ ❌ "你会用这个功能吗?" → 会得到礼貌性"yes"
54
+ ✅ "上次遇到 X 问题是什么时候?你当时做了什么?" → 得到真实行为
55
+
56
+ 2. **20 分钟访谈 >> 5 个 feature 问题**
57
+ 深度一个场景,比问 5 个浅问题更有价值
58
+
59
+ 3. **录音 / 整稿 > 当场记笔记**
60
+ 当场记会错过微表情和语气
61
+
62
+ 4. **小批量后暂停,复盘**
63
+ 按决策与样本差异安排;重复意见只是线索,继续找反证
64
+
65
+ 5. **用词镜像**
66
+ 用户用什么词,你也用什么词 —— 不要替换成产品语言
67
+
68
+ ---
69
+
70
+ ## 访谈结构模板(20-30 分钟)
71
+
72
+ | 阶段 | 时长 | 问什么 |
73
+ |---|---|---|
74
+ | Warm-up | 2 min | "你今天做了什么?"(开放式,松弛下来) |
75
+ | Discover | 5 min | "上次做 X 是什么时候?能讲讲吗?" |
76
+ | Deep-dive | 10 min | "当时你用了什么工具?怎么用的?哪里卡住?" |
77
+ | Alternative | 5 min | "有没有试过其他方法?为什么没继续?" |
78
+ | Wrap-up | 3 min | "如果有一个魔法能解决这个问题,它会怎么工作?" |
79
+
80
+ ---
81
+
82
+ ## Token Top10 访谈法(招牌打法)
83
+
84
+ **后台用量 Top10 拉 list → 自愿屏幕演示(录制需单独同意)→ 脱敏记录在获准工具中生成 feature/bug 草稿,审阅后再写入 Linear → 次日回访。节奏按研究问题与复盘能力决定。**
85
+
86
+ 1. 按 token 消耗 / 用量排 Top10,每周刷新——重度用户的反馈密度最高
87
+ 2. 经参与者同意后共享屏幕或录屏,允许拒绝:看真实操作 > 听口述
88
+ 3. 仅在获准向选定 AI 服务发送脱敏记录后提取 feature/bug 草稿;审阅后按已有授权写入 issue 工具
89
+ 4. 次日回访"你提的 X 已修/已排期"——反馈闭环就是留存杠杆
90
+ 5. 仅向获准的产研成员分享已同意录制且脱敏的片段
91
+
92
+ **邀约节奏**:
93
+ - 只向同意接收研究邀请的人发送邀请;未回复或拒绝后停止,不按固定次数追访。
94
+ - 使用用户同意的联系渠道;不反查社交账号追访,不因未回复换渠道联系。
95
+ - 话术用宽泛理由("早期团队想做访谈"),不暴露真实筛选原因——避免用户表演
96
+
97
+ **组合技**:
98
+ - KOL 访谈二合一:合作 YouTuber 顺便做用户访谈 + 可用性测试,一次合作两份价值(20-30 分钟视频访谈 ROI 最高)
99
+ - Wisprflow 打法:launch 前人工 onboard 前 500 用户逐个访谈,用势能带 launch
100
+
101
+ ---
102
+
103
+ ## 激活与付费口径
104
+
105
+ | 项 | 定义要求 |
106
+ |---|---|
107
+ | 激活 | 对应用户获得首次价值的可观察行为,不只注册或登录 |
108
+ | 注册到付费 | 同 cohort、同窗口、去重人数与真实付款;无固定合格线 |
109
+ | 流失 | 按正常使用周期定义,区分未激活与曾使用后停止 |
110
+ | 归因 | 时间线帮助解释,仍需渠道与行为证据,不能单凭先后判因果 |
111
+
112
+ ---
113
+
114
+ ## 深度参考
115
+
116
+ 📂 **https://github.com/Gingiris-1031/gingiris-skills/tree/main/skills/gingiris-user-interview**
117
+
118
+ - `references/templates.md` — 不同阶段(Discovery / Validation / Retention)的问题模板
119
+ - 访谈证据与抽样边界
120
+
121
+ ---
122
+
123
+ ## 合成流程
124
+
125
+ 1. 每次访谈后 24h 内整理 3 张卡片:
126
+ - **Pain**(用户的真实痛)
127
+ - **Workaround**(他们目前怎么凑合)
128
+ - **Quote**(可引用的原话)
129
+ 2. 每 5 个访谈做一次复盘:
130
+ - 独立出现的 Pain → 待验证需求假设,保留反证与抽样限制
131
+ - 具体 Workaround → 替代方案线索
132
+ 3. 每 20 个访谈做一次决策:
133
+ - 继续这个假设 / 换方向 / 需要更多数据
134
+
135
+ ---
136
+
137
+ ## 级联推荐
138
+
139
+ - 访谈发现新 PMF 方向 → `gr-blog-post` 写一篇(验证市场反应)
140
+ - 访谈发现对手名字被提 ≥ 3 次 → `gr-competitor` 深度分析
141
+ - PMF 明确 → `gr-b2b-growth` 或 `gr-oss-marketing` 启动增长
142
+ - 需要招募访谈对象 → 从 `gr-seo-patrol` 的 GSC 流量挑读者,或直接上「Token Top10 访谈法」拉后台用量榜
143
+ - 在跑红人合作 → 用「KOL 访谈二合一」顺路拿访谈(红人打法见 `gingiris-kol-outreach`)
144
+ - 访谈后要定激活/付费目标 → 先定义真实价值行为、cohort 和观察窗口
145
+
146
+ ---
147
+
148
+ ## 反模式
149
+
150
+ - ❌ 把“愿意付费”的回答当成付款证据;可以了解理由,再验证明确范围与价格的实际购买
151
+ - ❌ 发问卷替代访谈(信息密度差 10x)
152
+ - ❌ 只访现有用户不访流失用户(流失才是真金)
153
+ - ❌ 一次访谈超过 45 分钟(疲劳 → 敷衍)
154
+ - ❌ 访谈中推销自己产品(污染答案)
155
+ - 只向同意接收研究邀请的人发送邀请;未回复或拒绝后停止,不按固定次数追访。
156
+ - ❌ 邀约话术暴露筛选原因("因为你用量高所以找你" → 用户开始表演)
157
+ - ❌ 激活定义写成"注册/登录过"(必须是行为阈值)
158
+ - ❌ 访谈完不回访(次日回访"已修/已排期"是最便宜的留存动作)
159
+
160
+
161
+ ## 纠偏与冷启动实验
162
+
163
+ 按场景读取 [访谈纠偏](references/interview-recovery.md) 或 [冷启动目标与成本](references/cold-start-decision-card.md)。需求研究、可用性测试与销售答疑分开;用户自己完成任务,主持人的帮助单独记录。
164
+
165
+ <!-- gingiris-community-cta:start -->
166
+ ## 加入出海社群与购买资料包 · Community & Resources
167
+
168
+ - **加入出海社群:** 微信添加 **staywildgrow**,留言「**加入社群**」即可。
169
+ - **购买资料包:** 微信添加 **staywildgrow**,咨询购买 Gingiris 出海资料包,价格 **599 元人民币**。
170
+
171
+ - **Join the global growth community:** Add **staywildgrow** on WeChat and send **“加入社群”** (join the community).
172
+ - **Get the resource pack:** Add **staywildgrow** on WeChat to purchase the Gingiris global growth resource pack for **CNY 599**.
173
+
174
+ 社群与资料包为可选入口,不影响本开源 Skill 的安装和使用。
175
+
176
+ The community and paid resource pack are optional; neither is required to install or use this open-source skill.
177
+ <!-- gingiris-community-cta:end -->
references/cold-start-decision-card.md ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Cold-start decision card / 冷启动目标与成本
2
+
3
+ Use when turning research into a small acquisition or validation experiment. Set one primary decision, not a list of channels to fill.
4
+
5
+ - Segment, task, stage and eligibility; why this cohort fits.
6
+ - One initial channel, example leads with source/fit (not scraped bulk outreach), message drafts matched to the task, and one useful content example.
7
+ - Outcome: following, reply, meeting, activated user, retained use or actual payment. Define each separately, with window and deduplication key.
8
+ - Cash, tools and human hours; budget is an upper bound, not an obligation to spend. Zero cash does not mean zero cost.
9
+ - Baseline and forecast assumptions; unknown conversion/CAC/LTV stay unknown. Distinguish cash CAC from fully loaded cost and never divide by projected payers as if they were actual customers.
10
+ - Owner, time/budget cap, next review, continue/change/stop conditions.
11
+
12
+ If the original request is “paying customers” but the sensible next step is “interviews”, explain original target → proposed target → evidence/reason → question still unanswered → when to revisit payment. Do not silently replace paid acquisition with follower growth. Research drafts do not send messages without authorization.
13
+
14
+ Constructed example: third-party posts already mention an open-source tool, but interested users cannot find its author. Test an author entry and a relevant task invitation. Count replies, completed tasks and payments separately; stars, clones, followers and asking about price do not establish paid demand.
15
+
16
+ For content variants, keep project, format and landing page comparable, name the variable changed, retain IDs and measurement windows. Report counts as well as rates. Natural recommendation traffic is not a randomized experiment; tiny samples can be inconclusive. Only expand when downstream quality and the business cost limit support it.
references/interview-recovery.md ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Interview recovery and handoff / 访谈纠偏与交接
2
+
3
+ Choose the objective before the session. Discovery reconstructs recent real behavior; usability testing observes a defined task; sales/support answers adoption questions. They can share a meeting, but label the transition and do not score a persuasive demo as unassisted usability success.
4
+
5
+ | Trigger | Recovery | Record |
6
+ |---|---|---|
7
+ | Host explains features at length | Ask “What were you doing the last time this happened?” | User task, current workaround, actual impact |
8
+ | User asks product questions throughout | Acknowledge and note the question; offer to answer after the research segment, unless it blocks participation | Unresolved concern, agreed follow-up |
9
+ | Host operates the demo | Invite the user to attempt the agreed task; help only when needed, noting the intervention | Action, hesitation, error, assistance and result |
10
+ | No real task is defined | Ask about the user's product/workflow and select a representative task together | Why this task represents their needs |
11
+
12
+ Example rescue: “这个问题我记下了,稍后单独回答。先回到你最近一次处理这件事:当时具体做了什么?” Do not ignore material safety, cost or consent questions to retain control of the interview.
13
+
14
+ Before a usability test, explain prototype limitations and use a suitable sandbox where possible. Screen sharing and recording are optional and consensual. A research invitation does not authorize production integration, financial transactions or sharing private data. During the task, avoid narrating advantages; ask what the user expects and observe what they do.
15
+
16
+ Separate: cannot understand (expression/interface); understands but does not need it (value/persona); needs it but cannot adopt (capability, access, input or integration). A demo can test understanding and navigation; claims of time saved or business outcomes need real-use evidence.
17
+
18
+ ## Evidence and follow-up
19
+
20
+ Keep user quotation/timecode, task, observed action, host prompt, blocker, current workaround and proposed solution in separate fields. Handoff clips include start/end time, consent scope, redactions, issue owner, observable acceptance condition and revisit date. A clip index is often more usable than an unexplained full recording.
21
+
22
+ Re-test the changed task with an appropriate participant, ideally including the person who encountered it. Tell users “fixed” or “scheduled” only after checking the actual state. Record a proposed trial, an agreed trial and a completed action separately. Willingness-to-pay statements and requests for more credits are clues, not payment.
23
+
24
+ ## Sampling and evidence strength
25
+
26
+ Choose participants for the decision: paying, heavy-use, new, churned and non-adopting users answer different questions. Start with a manageable batch and reassess gaps and counterexamples; no fixed interview count or conversion percentage certifies PMF. Do not manufacture feature requests to meet a quota. A participant who does not need a change can supply valuable counterevidence.
27
+
28
+ Compensation pays for time, not favorable answers; disclose and record it. Do not treat an incentivized interview as invalid by default, or all free-user feedback as useless. Historical purchases and a clearly scoped paid offer are stronger pricing evidence than a hypothetical yes.
29
+
30
+ ## Facilitator exercise (constructed, not a customer result)
31
+
32
+ Bad: “Our integration is easy; let me show all its advantages.”
33
+ Better: “What are you trying to charge for? Please try configuring that in this test environment.”
34
+ Acceptance: the researcher identifies whether the person can complete the task, records assistance, and distinguishes an onboarding blocker from an unsupported product promise. No invented conversion result.
references/research-to-acceptance.md ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # From an interview to a verifiable product change
2
+
3
+ Use when a team has many feedback notes but cannot tell engineering what to change. Preserve the source context and distinguish what the user did, what they said, and the team's proposed solution.
4
+
5
+ ## Decision handoff
6
+
7
+ `Evidence / timestamp | User role and task | Observed blocker | Proposed solution | Benchmark | Core / onboarding / commercial | Priority and rationale | Owner | Acceptance steps | Follow-up user and date`
8
+
9
+ Acceptance needs a starting state, user action and observable outcome. “Improve onboarding” is not enough. For example: a new test account imports its sample document, completes the target task and can locate the saved result; record the actual result and failures. This is an illustrative acceptance pattern, not a past customer outcome.
10
+
11
+ Use a short shared document to align product and engineering. A competitor complaint is input; the team still needs to propose a solution and test it. Prioritize by blocked user value and dependency rather than how often a person repeats a request. Revisit the same user after a change where possible; shipping alone does not prove the problem was solved.
12
+
13
+ ## Feedback intake
14
+
15
+ Keep bug reports and feature requests in separate views of one shared log:
16
+
17
+ `Date | Reporter / source | Description | Reproduction or current workaround | Screenshot / recording where appropriate | Suggested priority | Owner | Status | Follow-up evidence`
18
+
19
+ Label suggested priority separately from the team's delivery commitment. Share only the account details and recordings needed for diagnosis, with suitable access controls. Public skill packages must not contain customer transcripts or private evidence URLs.
references/templates.md ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # 用户访谈模板 · 速查手册
2
+
3
+ > 用于真实行为、任务观察与回访验证;访谈次数不证明 PMF。
4
+ > 按 Discovery / Validation / Retention 三个阶段组织,直接抄结构、改产品名即用。
5
+
6
+ ## Token Top10 访谈法(日常运营循环)
7
+
8
+ 把访谈从"调研动作"变成"日常运营循环"的最短路径:
9
+
10
+ 1. 后台按 token 消耗 / 用量拉 **Top10 用户 list**(每周刷新)
11
+ 2. 约**共享屏幕访谈;录制需同意,可不录制**——看真实操作 > 听口述
12
+ 3. **脱敏 Transcript**:只向用户与参与者允许的服务发送,先生成 feature/bug 草稿,审阅后按已有授权写入 issue 工具
13
+ 4. **次日回访**:告知"你提的 X 已修/已排期"——反馈闭环本身就是留存杠杆
14
+ 5. 节奏按研究问题与团队复盘能力安排;仅向获准的产研成员分享已同意录制且脱敏的片段
15
+
16
+ 配套:新访谈员先 mock interview + 全程录屏复盘再上真实用户;KOL 访谈二合一(合作 YouTuber 顺便做访谈 + 可用性测试);访谈可产出需求证据;SEO公开引用与商业线索使用需另行获得明确许可,不默认改变研究数据用途。
17
+
18
+ ## 邀约模板与 1/3/7/14 触达节奏
19
+
20
+ - 只向同意接收研究邀请的人发送邀请;未回复或拒绝后停止,不按固定次数追访。
21
+ **访谈邀请**:使用用户同意的联系渠道;不反查其他账号追访,未回复或拒绝后停止跟进。
22
+
23
+ 邀约话术四要素:
24
+ 1. 说明目的(用宽泛理由如"团队早期想做用户访谈",**不暴露真实筛选原因**,避免用户带预设表演)
25
+ 2. 时长承诺(30 分钟)
26
+ 3. 录屏说明
27
+ 4. 价值交换(早期权限 / 路线图参与等)
28
+
29
+ **筛选优先级**:P0 付费用户、高频活跃用户 → P1 竞品用户、流失用户 → P2 注册未付费用户。
30
+
31
+ ---
32
+
33
+ ## 阶段一:Discovery(探索期问题模板)
34
+
35
+ **背景了解**
36
+ - 你的职业/角色?从哪个渠道知道我们的?用了多久?主要用哪个功能?
37
+ - 同类工具还在用哪些?
38
+
39
+ **工作流探索**
40
+ - 这个产品帮你解决什么问题?之前你是怎么做的?现在流程有什么变化?
41
+
42
+ **痛点挖掘**
43
+ - 用的过程中最不爽的地方?遇到过什么 Bug?哪个功能最难用?
44
+ - "魔法棒"问题:如果给你一根魔法棒,你最想让产品变成什么样?
45
+
46
+ **关键技巧**:让用户共享屏幕演示真实操作,观察卡点——屏幕观察 > 用户口述。
47
+
48
+ ## 阶段二:Validation(验证期问题模板)
49
+
50
+ **竞品对比**
51
+ - 用过哪些竞品?各自什么场景下用?我们相比的优劣势?
52
+
53
+ **付费意愿**
54
+ - 之前为类似工具付过费吗?愿意为这个付多少?什么会促使你升级?
55
+
56
+ **Beta 五问法(结构化收尾)**
57
+ 1. 你觉得这是一个怎样的产品?(让用户自己贴标签)
58
+ 2. 最满意的功能是什么?
59
+ 3. 哪个环节最卡顿或困惑?
60
+ 4. 如果明天正式上线,你愿意付费吗?为什么?
61
+ 5. 最希望下个版本加入什么?
62
+
63
+ **Beta 任务设计**:任务模拟完整工作流而非孤立功能点;任务长度要触发"等待"场景(暴露性能问题);提供 prompt example;明确说"Beta 不稳定是正常的"。连线测试用 Think Aloud Protocol,利用加载等待期收集用户背景。
64
+
65
+ ## 阶段三:Retention(留存/流失期问题模板)
66
+
67
+ **流失用户访谈三问**
68
+ 1. 什么时候开始不用的?当时在做什么?
69
+ 2. 遇到了什么问题,还是找到了替代品?
70
+ 3. 如果我们解决了 X 问题,你会回来吗?
71
+
72
+ **流失归因四类**:性能类(卡顿/超时/崩溃)、工作流断层类(缺下游集成)、ICP 不匹配类、Messaging 不匹配类(触达话术 ≠ 产品实际价值)。
73
+
74
+ **流失可运营化**:把流失定义成行为(如两次对话后 24h 不回访)→ 自动触发挽回邮件。
75
+
76
+ ---
77
+
78
+ ## 访谈复盘模板(24 小时内完成)
79
+
80
+ - 用户画像:角色 / 渠道来源 / 使用时长 / 分级(S/A/B/C)
81
+ - 核心工作流:解决什么问题、之前怎么做
82
+ - Top 痛点(按提及强度排序)
83
+ - Feature request → 进 tracker(标注用户等级 + 付费状态)
84
+ - 付费信号:付过什么、愿付多少、升级触发点
85
+ - 下一步:回访承诺 / 大使计划邀请 / 转介绍请求
86
+
87
+ **每 5-10 场汇总一次**:共同需求/痛点 → 更新功能优先级 → 修正用户画像 → 调整产品方向。
88
+
89
+ ## 工具链速查
90
+
91
+ | 类别 | 工具 |
92
+ |:---|:---|
93
+ | 问卷 | Typeform |
94
+ | 录音转写 | Grain / Fathom |
95
+ | 用户数据库 | Airtable |
96
+ | 排期 | Calendly |
97
+ | 触达 | LinkedIn / Discord |