PostHog用SQL优先级与Workflows自动化重构客户成功
Scaling customer success with PostHog
直接展示了如何用现有产品功能重构内部运营流程,给出了SQL优先级模型和Workflows自动化组合的具体落地细节,独立开发者可参考其低成本扩展思路。
Our customer success team works with every customer above $20k ARR, but this leaves thousands of customers without someone to talk to.
我们的客户成功团队服务于 ARR 超过 2 万美元的每位客户,但这留下了数千名无人对接的客户。
Helping all these customers was the responsibility of the onboarding team, but as we did it more, we realized our efforts were more like customer success scaled so that’s what our team name is now.
帮助所有这些客户原本是由入职团队负责的,但随着我们做得越来越多,我们意识到我们的工作更像是规模化后的客户成功,因此我们团队现在的名字就是如此。
The work of the
该
relies heavily on automations for customer records, account routing, prioritization, email sequences, follow-ups, alerts, sales handoffs, and more. We relied on an external customer success platform (CSP) to do this before.
的工作严重依赖自动化来处理客户记录、账户路由、优先级排序、邮件序列、后续跟进、警报和销售交接等事务。此前我们依靠一个外部的客户成功平台(CSP)来执行这些工作。
Now PostHog is our customer success platform, thanks to Customer Analytics and Workflows. Here’s why and how.
现在,凭借客户分析和工作流功能,PostHog 成为了我们的客户成功平台。以下是原因和方法。
Scaling customer success is about signals
扩展客户成功的关键在于信号
When you think of “onboarding”, you probably imagine a kickoff call after a closed deal, an implementation plan, and a series of training sessions. Ours looked different.
当你想到“入职”时,你可能想象的是成交后的一次启动会议、一份实施计划以及一系列培训课程。但我们的做法有所不同。
PostHog is self-serve, so a customer signs up, starts sending data, and explores products without talking to us. By the time we reach them, they may have already implemented several products and be approaching their first real bill. Very often, they haven’t asked for help. They may not even know our team exists.
PostHog 是自服务模式的,因此客户注册后会自行开始发送数据并探索产品,无需与我们交谈。当我们联系他们时,他们可能已经部署了多个产品,并且即将收到第一笔正式账单。很多时候,他们并没有寻求帮助,甚至可能不知道我们团队的存在。
Signals from product usage, billing, and customer activity help us identify moments when customers might be struggling, missing out on value, or a churn risk. We then optimize a customer’s setup and billing, explore their data and identify gaps, find better ways to answer business questions, guide them to discover other useful aspects of PostHog, and make sure they will stay with us longer.
来自产品使用、计费和客户活动的信号帮助我们识别客户可能遇到困难、错失价值或存在流失风险的时机。随后,我们会优化客户的设置和计费方案,探索他们的数据并找出差距,寻找更好地回答业务问题的方法,引导他们发现 PostHog 的其他有用功能,并确保他们会更长久地留用我们的产品。
Additionally, we're the first customer-facing team at PostHog to operate at scale: four people support around 700 customers through one shared program. We don’t have individual books of business; instead, we rely on a pretty intricate automation system to make sure nobody falls through the cracks and that we don’t step on each others’ toes while still being responsive and available.
此外,我们是 PostHog 首个以规模化方式运营的客户-facing 团队:四个人通过一个共享项目支持约 700 名客户。我们没有单独的业务管理范围;相反,我们依赖一套相当复杂的自动化系统,以确保没有人被遗漏,同时我们在保持响应速度和可用性的前提下避免互相干扰。
Our processes need to keep up with our customer “queue” which never stands still (forecasted spend changes with usage, accounts become eligible or drop out, some get handed over to Sales team, etc.). We need to figure out who needs our attention, understand customer context, and coordinate our work so we can help as many customers as possible, without turning our “help” into an impersonal email campaign or spending half our day managing a CSP.
我们的流程需要跟上永远在变化的客户“队列”(预测支出随使用情况而变化,账户变得符合条件或退出,部分转交给销售团队等)。我们需要确定谁需要我们的关注,理解客户背景,并协调我们的工作,以便尽可能多地帮助客户,而不会将我们的“帮助”变成非个性化的电子邮件活动,也不会花费半天时间来管理 CSP。
That's why a successful migration to PostHog was so vital for our team.
这就是为什么成功迁移到 PostHog 对我们的团队至关重要。
Bringing customer data and operations together
将客户数据与运营整合在一起
Previously, we managed all processes in an external CSP. Our collection of automations and integrations was ever-growing and somewhat worked, but it required creative workarounds and lengthy troubleshooting. As our needs grew, maintaining these became increasingly complicated.
此前,我们在外部 CSP 中管理所有流程。我们的自动化和集成组合不断增长且勉强可用,但需要创造性的变通方案和冗长的故障排除。随着需求的增长,维护这些方案变得越来越复杂。
We wanted the flexibility to make decisions based on billing, product usage, engagement, and previous conversations, without piecing together information from different tools and holding our breath to keep the system from breaking.
我们希望拥有基于账单、产品使用情况、参与度和过往对话做出决策的灵活性,而无需从不同工具中拼凑信息,并提心吊胆地担心系统崩溃。
When PostHog started building Customer Analytics, we saw an opportunity to simplify things. Instead of sending our own customer data to an external CSP, why not run our processes directly where that data already lived?
当 PostHog 开始构建客户分析(Customer Analytics)时,我们看到了简化的机会。与其将客户数据发送到外部 CSP,为什么不直接在数据所在的地方运行我们的流程呢?
The entire data layer was built by
整个数据层由
- a SQL model in PostHog's data warehouse that calculates account eligibility, a priority score based on billing forecasts, upcoming bills, recent engagement, and other signals. The advantage is that when we need to adjust our prioritization or eligibility conditions, we can change them in one centralized place rather than update multiple integrations one by one.
- PostHog 数据仓库中的一个 SQL 模型构建而成,该模型根据账单预测、即将到来的账单、近期参与度及其他信号计算账户资格和优先级评分。其优势在于,当我们需要调整优先级或资格条件时,可以在一个集中位置进行修改,而不是逐个更新多个集成。
Customer Analytics pulls from this and offers a single shared view, sorted by that priority score. It helps us identify which customers need attention first, while giving us access to their usage, spending, conversations, meetings, and progress through our program.
客户分析(Customer Analytics)从中提取数据并提供统一的共享视图,并按优先级评分排序。这帮助我们识别哪些客户最需要关注,同时让我们能够访问他们的使用情况、支出、对话、会议以及在我们项目中的进展。
Knowing which customers need help is only half the challenge - we also need to act on that information, and that's where Workflows come in handy.
知道哪些客户需要帮助只是挑战的一半——我们还需要据此采取行动,这正是工作流(Workflows)派上用场的地方。
Turning Workflows into the scaling engine
将工作流打造为扩展引擎
On the surface, Workflows might seem to be for marketing-related automations, but we quickly proved that they could do much more than send emails and could help us launch our operations at scale.
表面上看,工作流似乎仅用于营销相关的自动化,但我们很快证明它们不仅能发送邮件,还能帮助我们大规模启动运营。
We needed a system that could act on customer data, coordinate our team's work, automate, and adapt when circumstances changed. Instead of recreating every old CSP automation separately, we consolidated much of our logic into a few core workflows, and supporting ones focused on specific tasks.
我们需要一个能够基于客户数据采取行动、协调团队工作、实现自动化并在情况变化时适应的系统。我们没有逐一重建旧的 CSP 自动化,而是将大部分逻辑整合到少数核心工作流中,并辅以专注于特定任务的支持性工作流。
A lot of this happened through conversations with AI agents, e.g., PostHog AI, MCP, and the Slack app. Usually we described in plain language what we want to automate, review what the agent builds, test the logic, and enable it once we're happy with how it works.
许多操作是通过与 AI 代理(如 PostHog AI、MCP 和 Slack 应用)的对话完成的。通常我们用自然语言描述想要自动化的内容,审查代理构建的结果,测试逻辑,并在对其工作方式满意后启用它。
Here are few specific examples:
以下是几个具体示例:
1. An onboarding journey that adapts to the customer
1. 适应客户的入职旅程
Every day, an “Enter queue” workflow adds newly eligible accounts to our onboarding program. Customer Analytics triggers an event when an account property changes, allowing subsequent workflows to pick up where they left off.
每天,一个“进入队列”工作流会将新符合资格的客户账户添加到我们的入职计划中。当客户属性发生变化时,Customer Analytics 会触发一个事件,使后续工作流能够接续之前的进度继续执行。
Our main email sequence lives in a single workflow, but not every customer follows the same path. For eligible customers forecasting under $600/month, we send an automated first email offering help with setup and billing. For higher-paying customers, we take a more personal approach: we review their account and write tailored advice based on their setup and usage.
我们的主要邮件序列位于单个工作流中,但并非所有客户都遵循相同的路径。对于预计月消费低于 600 美元的合格客户,我们会发送一封自动化的首封邮件,提供设置和计费方面的帮助。对于高付费客户,我们采取更个性化的方法:审查他们的账户,并根据其设置和使用情况撰写定制建议。
Both groups then continue through the onboarding sequence, including a "last chance to get our help" message timed around their upcoming bill.
随后,这两组客户将继续完成入职序列,其中包括一条在即将到来的账单日前后发送的“最后获得我们帮助的机会”消息。
Workflows then handle the rest of the journey: flagging check-ins for the team, adding tasks to Customer Analytics, sending follow-ups where appropriate, and finally, completing the onboarding by setting a person property. Our main email sequence is currently on version 70, which is a great depiction of the number of edge cases we've worked through.
工作流随后处理旅程的其余部分:为团队标记签到、向 Customer Analytics 添加任务、在适当的时候发送跟进邮件,并最终通过设置人员属性来完成入职。我们的主要邮件序列目前处于第 70 版,这很好地体现了我们已经处理过的边缘案例数量。
It’s important to note that no workflow sends emails just because they're scheduled. Before every send, it checks the customer's current situation. Has someone recently contacted the customer? Has another team taken ownership? If the situation has changed, the message is skipped.
需要注意的是,没有任何工作流仅仅因为排期到了就发送邮件。在每次发送之前,它都会检查客户的当前状况。是否有人最近联系过该客户?是否有其他团队接管了该客户?如果状况已发生变化,则跳过该消息。
2. Billing optimization and proactive help
2. 计费优化与主动帮助
Sometimes the best thing we can do for customers is help them spend less, and we enjoy doing so.
有时,我们能给客户做的最好的事情就是帮助他们减少支出,而我们也很乐意这样做。
Via workflows, we reach out when customers encounter failed payments, offering assistance without requiring someone to monitor every account.
通过工作流,当客户遇到付款失败时,我们会主动联系并提供协助,而无需专人监控每个账户。
We also developed a system of proactive interventions, based on specific spend and usage signals, that lets us identify customers who could benefit from talking with us and keep their PostHog usage sustainable before a potential issue becomes a reason to leave.
我们还开发了一套基于特定支出和使用信号的主动干预系统,使我们能够识别出那些可能从与我们交谈中受益的客户,并在潜在问题成为离职原因之前,保持他们在 PostHog 的使用可持续性。
Combining customer data with Scouts and automation enables us to offer this kind of proactive help to far more people than we could reach out to manually.
将客户数据与 Scouts 和自动化相结合,使我们能够为比手动联系更多的人提供这种主动帮助。
3. Automating the work behind the scenes
3. 自动化幕后工作
Some of our favorite workflows don't send emails at all - they save time and reduce manual effort.
我们最喜欢的一些工作流根本不发送邮件——它们节省时间并减少人工操作。
For example, Workflows creates follow-up tasks after onboarding meetings, add Gong call summaries to customer notes using AI tasks, and generate tasks from customer research reports.
例如,工作流会在入职会议后创建跟进任务,使用 AI 任务将 Gong 通话摘要添加到客户笔记中,并从客户研究报告中生成任务。
Our “Stage Reactor” keeps track of changes in each account's stage, recording timestamps and managing specialist assignments.
我们的“Stage Reactor”会跟踪每个账户阶段的变更,记录时间戳并管理专家分配。
These might not sound like big game changers, but across hundreds of accounts, they add up to a considerable amount of coordination we don’t have to handle anymore. What’s even more important, they keep our processes consistent, regardless of who's working that day, and free up time for the work that actually requires our expertise.
这些听起来可能不是什么重大的变革,但在数百个账户的规模下,它们累积起来为我们节省了大量不再需要手动处理的协调工作。更重要的是,无论当天由谁负责,它们都能确保流程的一致性,并释放出时间让我们专注于真正需要专业知识的任务。
Shaping Customer Analytics along the way
在此过程中塑造客户分析能力
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力