该不该请公关公司?把 Lulu 的思路落成中小企业执行表
Hiring a PR Agency: A Founder-Led SME Playbook
中小企业可以外包媒体沟通、发布协调等执行工作,但应把目标客户、公司故事和关键承诺掌握在内部。聘请公关公司的前提,是知道自己缺哪一种能力,并且有人能持续提供事实、接受采访、作出决定。
这对应 Lulu Cheng Meservey 在 Flack 的原文《Should you hire a PR agency?》,发表于 2025 年 3 月 14 日,附视频文字稿。她创办了 Rostra,曾负责 Substack 与动视暴雪的传播工作。本文结合她的公开宣言、访谈节目要点与 AMEC 评估框架,给出面向初创企业的执行方案。
原文:https://www.getflack.com/p/should-you-hire-a-pr-agency
Lulu 的判断,怎样落到一家小公司?
原文的边界很清楚:故事与战略留在公司,执行忙不过来时找外援;挑选实际干活的人,验证判断力、口碑和合作节奏;用业务目标检验价值。她建议不确定时先签较短合作,例如三或六个月。
她在 2024 年的 Go Direct 宣言中强调,创始人应建立直达重要受众的叙事和渠道。直接沟通仍可接受帮助,关键是公司不依赖中间人才能讲清自己。
宣言:https://www.getflack.com/p/go-direct-the-manifesto
《How To Build A Cult》对应 The Knowledge Project 的第 246 期访谈,节目官方公开要点强调人的可信度、清晰表达与信念。这里值得中小企业借用的,是让客户认识承担责任的人,并持续看到可信证据。
官方节目页:https://fs.blog/knowledge-project-podcast/lulu-cheng-meservey/
以下工作表、预算与时间安排是本文的应用建议,尚未作为一项企业案例实施。 为便于落地,我们用一家虚构的 18 人团队作贯穿示例:它向制造企业销售售后工单软件,希望接触真正负责售后流程的人。
第一张表:创始人先交付一页叙事简报
请创始人与一位接触客户最多的同事共同填写。每一项最多两三句,能拿出事实的地方同时附上出处。先写客户听得懂的版本,再考虑媒体角度。
| 要写清的事情 | 虚构工单软件公司的示范 |
|---|---|
| 谁遇到什么问题 | 制造企业售后主管找不到工单交接责任人,客户反复催问 |
| 为什么值得现在解决 | 公司正在扩张服务网点,现有表格交接开始失效 |
| 我们作出的选择 | 先把责任与进度做透明,而非一开始承诺全自动处理 |
| 可核对的证据 | 准备经授权的客户案例、产品演示和统计方法;没有数据就先补采集 |
| 哪些承诺不成立 | 不承诺接入当天消除全部投诉,也不捏造节省比例 |
| 希望读者下一步做什么 | 预约一次针对自身工单流程的演示 |
这张表由公司签字确认。外援可以追问、整理、挑战,但“客户是谁、证据是什么、愿意承诺什么”需要内部负责人作决定。创始人不擅长写作时,可以录一段口述,让同事整理,再逐项审核事实与措辞。
把一句空泛的“重新定义售后智能化”,改成一个具体问题:“客户第二次来电时,为什么还没人知道谁负责?”接下来用真实演示回答。这个例子是本文创作的表达练习,不是 Lulu 提供的企业案例。

第二张表:到底买哪一段执行?
先给采购需求命名。原文谈到的媒体 outreach 是主动与记者沟通,不等于付费买版面;商务合作、广告与编辑部独立报道应分开预算和验收。
| 当前卡点 | 可购买的明确任务 | 公司仍负责的部分 |
|---|---|---|
| 创始人不会把产品讲清楚 | 采访式梳理与表达训练 | 定位、事实和最终表述 |
| 发布期间忙不过来 | 采访排期、资料整理、媒体问询协调 | 发言人出席和及时审批 |
| 不知道目标客户看什么 | 受众与渠道研究 | 确认目标客户和实际业务价值 |
| 缺少可信材料 | 客户案例制作、产品演示整理 | 获得授权、核验数据、开放产品 |
| 有持续的行业沟通工作 | 指定团队的月度执行服务 | 叙事方向、预算和业务反馈 |
对示例公司来说,第一笔钱更适合购买一个边界清楚的项目:完成一份真实案例、安排一次产品讲解、测试两种客户可接触到的分发渠道。是否升级为长期服务,等工作量与价值显现后再决定。
拥有一篇文章不等于拥有分发能力。公司应同时保留自己的原始说明页、演示入口和客户反馈记录;外部报道可以扩大触达,但不应成为唯一能找到产品解释的地方。这里是 Go Direct 思路在小团队渠道安排上的应用。
公关公司怎么筛?用同一份材料看实际表现
可以邀请少量候选团队参加一次有边界的沟通,必要的深度诊断付费完成。把同一页叙事简报交给对方,要求回答三个实际问题:谁应该读到这个故事、为什么现在会在意、现有证据哪里最弱。
接着看交付,而非只看客户 Logo。要求日常负责人参加沟通,提交一份任务安排和成本拆分;再找做过相近项目的客户了解配合过程。记录沟通中的答复速度与任务理解;相关媒体经验应能说明具体经历,而不是只展示通讯录。最后确认哪个人负责对接、遇到争议谁作判断,以及关键人员调整后怎么办。
下面是一份可以填写的面谈记录。它不是经过统计校准的评分模型:
| 观察项 | 请对方现场展示 | 留下的判断依据 |
|---|---|---|
| 受众理解 | 解释客户角色及其信息需求 | 是否理解购买场景,而非泛列“大众媒体” |
| 内容判断 | 指出简报中一处不可证实的说法 | 是否愿意纠正我们,而不只是夸奖 |
| 执行清晰度 | 列出首个阶段的负责人、依赖和交付物 | 是否能指出公司必须投入的工作 |
| 合作可靠性 | 说明一个项目受阻后的处理经过 | 能否给出经得起核对的客户反馈 |
如果团队只讨论“我们认识很多人”,请把话题拉回这次任务:它能帮助公司接触哪些相关人群,靠什么内容,下一步由谁执行。媒体名单应当是方法的一部分,无法代替方法本身。
该付多少钱?先分开历史行情与自己的预算
Lulu 在 2025 年给出的美元经验区间是:常规科技公关约 每月 2万—3.5 万美元,专门化或高知名度团队约每月 6万—10 万美元。它们不是今天中国中小企业的报价标准。
对于当前采购,向候选方索取同一任务边界下的报价,分开服务费、制作、付费渠道和额外支出。不要只比较总价;同时比较实际投入人员、交付范围与公司需要补的工作。本文没有开展 2026 年国内市场报价调查。
下面只演示预算如何拆分。假设公司愿意为一次三个月的传播实验预留 3 万元现金上限:
| 假设用途 | 现金上限 |
|---|---|
| 边界明确的外部执行服务 | 18,000 元 |
| 案例与演示材料制作 | 6,000 元 |
| 经批准的渠道或活动费用 | 3,000 元 |
| 预留 | 3,000 元 |
| 合计 | 30,000 元 |
这是预算练习,不是报价、推荐成交价或结果承诺,也不包含内部人员时间。如果真实报价不能覆盖任务,就减少交付范围、延后实验或先由内部完成,不要把它包装成全年全套公关。
合作文件中应明确交付物、审批责任、费用上限、素材交接和结束安排。付费内容与争取独立报道分别记录;不要把“承诺付费位置”与“保证记者独立采写”混成一项商品。具体合作条款以双方书面约定为准。

第三张表:每周看什么,才知道有没有帮助?
Lulu 对媒体声量占比、报道中关键信息出现率这类周报指标评价很低。本文保留传播活动记录来诊断执行环节,并要求继续检查客户反应与业务推进。
AMEC 的 Integrated Evaluation Framework 将工作产出、受众反应、行为结果与组织影响区分开,并要求事先确定目标和基线。它支持用不同层次证据评价传播,而非只数发布了多少条内容。
框架:https://amecorg.com/amecframework/
将这一框架用于示例公司,可以形成下面的周报。字段与例子由本文设计:
| 层次 | 公司记录什么 | 这项记录不能证明什么 |
|---|---|---|
| 交付 | 案例上线、目标渠道接触与采访完成情况 | 内容发出不等于客户看懂 |
| 理解 | 演示前客户最常问的问题、能否正确复述用途 | 少量访谈不代表所有潜在客户 |
| 行为 | 来自适配公司的演示预约与后续讨论 | 点击不等于有效商机 |
| 业务贡献 | 哪些材料被销售实际使用、哪些机会继续推进 | 同期成交不能全部归因于公关 |
“合格演示预约”要先定义,比如对方确有售后流程需求、参与者有相关职责、愿意讨论部署条件。销售人员每次记录首次得知产品的路径,以及后来帮助其理解的材料;保留“不知道来源”这一选项。
官网链接参数、预约表和客户口述可以提供线索,但跨设备、转发和线下交流会造成缺失。因此这些记录反映贡献线索,不是严格证明传播带来增量成交的因果实验。
如果需要计算成本,可以用“包含内部投入的实验总成本 ÷ 合格机会数”,并称为每个机会的分摊实验成本。它不是已经证明的获客成本或公关投资回报率;机会数为零时,直接报告尚无合格机会。
同时保留一项合作观察:团队是否主动暴露问题,并把新信息变成下一步行动。它应配具体实例,不能用“感觉专业”替代事实,也不能让一张漂亮的数字周报掩盖长期没有学习。
90 天怎样安排,什么时候结束合作?
这是本文建议的检查周期,不是 Lulu 保证见效的期限:
- 第 1—2 周:补基础。 填完叙事简报,确认案例可用性、演示入口和内部负责人;基础缺失就暂停采购。
- 第 3—6 周:做一次完整发布。 产出材料、分发、接住问询,检查预约是否来自适配客户。
- 第 7—10 周:根据反馈修正。 调整含糊的说法、无效的渠道或没有证据的承诺,不因排期到了就继续堆文章。
- 第 11—12 周:决定继续、缩小或结束。 用已承诺的交付、目标客户反馈、业务推进线索与总成本共同判断。
长期传播的价值不一定在一个周期内全部显现,但一个周期应当带来可检查的工作与学习。如果说不清做了什么、反馈是什么、下一步为什么这样改,增加预算也不能解决这个管理问题。

常见问题
创始人不擅长写作,还能坚持 Go Direct 吗?
可以先保留自己的判断和第一手材料,再让同事或外援整理表达。公司需要能核实每个关键承诺,并让客户知道谁为它负责。
没有足够预算请月费团队怎么办?
缩小任务。先做一份可信案例、一次受众研究或一轮演示反馈,按明确交付询价;不要为模糊的“品牌声量”透支预算。
曝光量和报道数量完全不看吗?
它们可以说明分发与交付发生了什么,但不能独立代表理解、信任或业务结果。把它们放回完整证据链中,而不是直接当作成交证明。
相关阅读:如何把工具服务与持续顾问工作组合成交付,可参看本站 Plausible 咨询服务分析:https://blog.mushroom.cv/blog/plausible-analytics-consulting-white-label/
© 2026 Author: Mycelium Protocol. 本文采用 CC BY 4.0 授权——欢迎转载和引用,须注明作者姓名及原文链接,不得去除署名后以原创发布。
A startup can delegate media outreach and launch coordination while retaining responsibility for its audience, story and promises. Before buying agency support, identify the missing capability and appoint someone internally who can supply evidence, make decisions and meet interview commitments.
The primary source is Lulu Cheng Meservey’s Flack essay dated March 14, 2025. She created Rostra after communications roles at Substack and Activision Blizzard. This article uses its public text, her manifesto, official podcast notes and AMEC’s evaluation framework. The worksheets and fictional example below are our proposed application, not a tested company case.
Original: https://www.getflack.com/p/should-you-hire-a-pr-agency
What should stay inside the company?
Meservey separates narrative ownership from execution support, stresses the actual working team and credible references, and evaluates value against business goals. For uncertain engagements, she mentions three- or six-month initial contracts.
Her 2024 manifesto argues for a direct relationship with important audiences without depending on intermediaries. Receiving help remains compatible with that ownership.
Manifesto: https://www.getflack.com/p/go-direct-the-manifesto
The Knowledge Project episode 246 is the relevant cult-building interview. Its public notes emphasize a recognizable human, clear language and trust. We reviewed those notes rather than accessing the member transcript or claiming to have listened to the full episode.
Episode: https://fs.blog/knowledge-project-podcast/lulu-cheng-meservey/
Worksheet one: a founder-owned narrative brief
Imagine an 18-person software company serving manufacturing after-sales teams. This fictional business wants relevant prospects to book a workflow demonstration. Its one-page brief should answer:
| Field | Example to develop and substantiate |
|---|---|
| Customer problem | Service managers cannot identify the owner of a handoff |
| Why now | New service locations expose spreadsheet limitations |
| Product choice | Make ownership and progress visible before promising automation |
| Evidence | Authorized customer story, working demo and documented measures |
| Boundaries | No invented savings figure or universal launch-day guarantee |
| Next action | Book a demonstration of the prospect’s own workflow |
The founder and a customer-facing colleague own these answers. A helper can interview, edit and challenge them. Recording a spoken explanation is a practical starting point for a founder who finds writing difficult.
Our writing exercise replaces a vague category claim with a concrete customer question: why does nobody know who owns the issue when the customer calls again? The product demonstration must then answer it with evidence. This is our example, not a case supplied by Meservey.

Worksheet two: buy a defined task
Separate editorial outreach, paid placements and commercial partnerships. A supplier arranging paid space is not guaranteeing independent reporting.
| Constraint | A possible external assignment | Internal responsibility |
|---|---|---|
| Unclear explanation | Interview-led editing and presentation practice | Positioning, evidence and final promises |
| Launch overload | Scheduling, materials and inquiry coordination | Spokesperson availability and approvals |
| Unknown audience channels | Focused audience research | Confirming the customer and business relevance |
| Missing proof | Customer-case and demo production | Permissions, data and product access |
| Recurring workload | Named-team execution retainer | Direction, budget and business feedback |
For the fictional company, start with one useful case, one demonstration and a small distribution experiment. Keep a company-owned explanation page, demo destination and feedback record, so a media article is an additional route rather than the only product explanation.
When considering suppliers, give each the same brief. Ask who should care, why now and which claim lacks evidence. Have the working lead participate, inspect a task plan and cost breakdown, and speak to relevant former clients. Observe response quality and ask for relevant, verifiable media experience rather than a contact database. Agree how responsibility changes if a key person leaves. These are our proposed procurement steps, not a calibrated agency ranking.
What should a small company pay?
Meservey’s 2025 examples put typical tech retainers at $20,000–35,000 monthly and specialized/high-profile support at $60,000–100,000. These are historical context, not current Chinese SME quotes.
Obtain comparable proposals for a defined scope. Distinguish service fees, production, paid channels and expenses. We did not survey the 2026 local market.
For illustration only, a three-month RMB 30,000 cash cap might reserve RMB 18,000 for defined execution, RMB 6,000 for materials, RMB 3,000 for approved channels/events and RMB 3,000 contingency. It excludes internal time and is neither a market price nor an outcome promise. Reduce the assignment or postpone it if real quotes cannot cover it.
Document deliverables, approvals, spending limits, handover and exit arrangements. Do not treat a small project budget as a promise of a full ongoing PR department.

Worksheet three: connect activity to business evidence
Meservey criticizes share-of-voice and message-pull-through reports as weak business signals. We retain delivery records for diagnosis, while requiring evidence further along the customer journey.
AMEC separates outputs, audience response, outcomes and organizational impact, starting with objectives and baselines. Its framework helps avoid treating publication counts as a complete result.
Framework: https://amecorg.com/amecframework/
Our suggested weekly record for the fictional company is:
| Layer | Evidence to collect | Limitation |
|---|---|---|
| Delivery | Materials live, outreach and interviews completed | Distribution does not prove understanding |
| Understanding | Questions before demos and correct product explanations | Small interviews are not a population survey |
| Behavior | Relevant-company demo bookings and follow-up | Clicks are not qualified opportunities |
| Contribution | Materials used in sales and opportunities progressing | Concurrent sales are not all attributable to PR |
Define a qualified meeting before launch: relevant need, an appropriate role and willingness to discuss deployment. Ask how the prospect first heard about the product and which material helped later. Allow unknown sources. Link parameters, forms and self-reports provide clues but miss forwarded, offline and cross-device exposure.
A calculation of total experiment cost, including internal time, divided by qualified opportunities is a cost allocation. It is not proven incremental customer-acquisition cost or PR ROI. If there are no qualified opportunities, report that directly.
Also record whether the team identifies problems and changes its next action. Attach examples, not a vague impression of professionalism.
How could a 90-day pilot run?
This is our proposed review cycle, not a promise of time-to-results. Weeks 1–2 establish the brief, evidence and owner. Weeks 3–6 complete a small launch and collect inquiries. Weeks 7–10 revise unclear claims or ineffective channels. Weeks 11–12 decide whether to continue, narrow or end the work using delivery, customer feedback, business contribution and cost.
Long-term value may take longer, but a cycle should produce inspectable work and learning. More spend cannot repair an engagement where nobody can explain the work, feedback or next decision.

Frequently asked questions
Must the founder write everything personally?
The founder can receive editing and production help while retaining judgment, evidence and responsibility for important promises.
What if a monthly agency is unaffordable?
Scope a useful, bounded task and obtain a quote for its deliverables. Do not buy an undefined promise of visibility.
Should impressions and publication counts be ignored?
They can describe distribution activity, but do not independently establish understanding, trust or business outcomes.
Related service-design discussion: https://blog.mushroom.cv/blog/plausible-analytics-consulting-white-label/
© 2026 Author: Mycelium Protocol. Licensed under CC BY 4.0 — free to share and adapt with attribution. You must credit the author and link to the original; removing attribution and republishing as original is not permitted.
关于本站 · 免责声明
🍄 Mushroom Research Blog 是非营利、免费公开的个人科技观察博客与公众号 XStack18,不接受商业合作、不代表任何企业或机构立场,也不谋求商业利益。我们以个人视角客观中立地记录和分析 AI、Web3 等领域的最新模型发布与技术动态——不止转述新闻标题或二手信息,而是给出有独立思考的深入分析,希望帮更多人获得有价值的一手科技认知。
⚠️ 文中介绍的开源代码与模型,仅供学习交流与技术借鉴。它们大多仍处于早期阶段,有待进一步研究和验证,请勿直接用于工作或生产环境;如需采用,请先自行充分测试,并核实其许可证与安全性。
Open-source code and models featured here are shared for learning and reference only. Most are early-stage and still need further study and verification — please don't use them directly in your work or in production. Test them thoroughly and check their licenses and security first.
- 本站文章均为作者基于公开信息的个人研究与观点整理,不代表文中提及的任何公司、产品、模型的官方立场,未与其构成商业关联或合作关系。
- 科技行业信息更新极快,我们尽力保证内容准确、及时,但不对完整性、实时性做绝对保证,具体请以相关企业/项目官方公告为准。
- 文中引用的第三方商标、产品名称、图片、数据等版权归原权利人所有,我们会尽量注明来源;如你认为存在版权疑问或侵权,请通过下方邮箱联系我们,收到通知后会尽快核实处理(更正、加注来源或删除)。
- 文章内容仅为技术科普与个人观点,不构成投资、法律或其他专业建议,据此进行任何决策的后果需自行判断和承担。
📮 侵权 / 勘误 / 合作咨询:[email protected]
💬 评论与讨论
使用 GitHub 账号登录后发表评论