PostHog Secures $75 Million Series C to Build AI Product Manager

PostHog获7500万美元C轮融资，拟打造AI产品经理 该开发者分析平台向自主AI产品战略的转型获得了资金支持；此前，其创始人通过坚持极具主见的公开透明化运作和刻意两极分化的设计，成功打入工程师市场。 本次讨论梳理了PostHog的发展历程：从经历长达六个月持续产品失败的艰难时期以及2020年3月混乱的种子轮融资，一路成长为如今拥有16或17款产品、估值达14亿美元的独角兽企业。CEO James Hawkins概述了早期的困境如何催生出一个对开发者友好的开源分析平台并在Hacker News上成功发布。他重点介绍了公司近期在7500万美元融资支持下，向AI驱动的自主产品经理方向做出的战略转型。此外，内容还详述了PostHog非同寻常且极具争议的营销策略，包括旧金山奇特的广告牌、高度透明的“公开构建”（building-in-public）理念，以及复古操作系统风格的网站。 PostHog 完成 7500 万美元 C 轮融资，估值达 14 亿美元，正式成为 Y Combinator 新晋独角兽，目前拥有约 160 名员工并服务近 30 万名客户。 (0:05) 在经历了长达 6 个月、平均每 5 到 6 周推翻一次想法的“转型地狱”后，创始人将反复重新配置数据分析工具的痛点转化为开源产品分析方案，在 Demo Day 前一个月成功破局。 (2:05) 2020 年 3 月全球疫情暴发期间，团队历经与 160 家投资机构的会面才艰难凑齐种子轮，其关键转变在于放弃迎合投资人、全力押注开源开发者社区。 (12:40) 公司正利用新融资全力转向 AI 驱动的产品自主化，开发能自动分析会话录屏及报错数据、并在工程师下班后自主提 PR 修复代码问题的 AI Agent。 (16:50) 通过坚持“公开构建”（Building in Public）、打造高互动度的复古操作系统风格网站以及反传统的幽默营销，PostHog 成功赢得了开发者群体的深度信任。 (23:01)

PostHog CEO James Hawkins attributes the company's rise to a $1.4 billion valuation to enduring rapid, failed iterations, adopting radically transparent and eccentric marketing, and using a broad product analytics stack as a foundation for autonomous AI development. Escaping Pivot Hell Through Developer-First Analytics (2:05) PostHog's core product emerged directly from the friction of cycling through failed startup ideas every five to six weeks. The founders initially built a sales territory management platform, securing 15 interested sales executives, only to find that 14 never clicked the sign-up link. Hawkins realized they had failed to follow user research principles like those in The Mom Test and concluded that sales leaders provided poor signal-to-noise feedback. This experience taught them to build for engineers and problem solvers whose actions matched their words. After another pivot to a technical debt survey tool stalled, the founders focused on their own recurring frustration: repeatedly setting up existing analytics tools that were designed for product managers, blocked by ad blockers, and unable to be self-hosted. Four weeks before their Y Combinator Demo Day in 2020, they built an open-source analytics alternative. Reframing the project from generic self-hosted infrastructure to 'open-source product analytics' transformed their trajectory, resulting in the most upvoted developer tool post on Hacker News that year and shifting customer acquisition from outbound pitching to organic inbound pull. Funding Disparities and the Autonomous AI Pivot (12:40) PostHog's fundraising history highlights a sharp contrast between early survival and later expansion. Closing their seed round during the March 2020 onset of COVID-19 required pitching 160 investment firms and piecing together $5,000 angel checks after venture funds withdrew. Hawkins admitted their initial pitches failed because they gave uninspiring, consensus answers. They secured traction only after developing strong conviction around prioritizing an open-source developer community over conventional enterprise sales teams. Five years later, PostHog raised a $75 million Series C round at a $1.4 billion valuation led by Y Combinator growth partner Shailendra. Hawkins used this capital to fund a strategic transition toward autonomous software engineering instead of horizontal SaaS expansion. Believing artificial general intelligence is technically achievable because the brain is a physical system, PostHog is developing a desktop application modeled after a human product manager. The tool analyzes session replays, analytics, error tracking, and LLM traces to automatically draft and ship pull requests resolving user problems without manual intervention. Distinctive Brand Marketing and Polarizing Design (23:01) Because PostHog operates without a traditional sales army, the company relies on extreme transparency and unconventional marketing to build developer trust. From inception, the founders published company operations, pricing structures, and internal challenges publicly. To humanize the brand, the company showcased team bios—including personal details and pets—and shared candid blogs about their move to San Francisco, creating relatable touchpoints for a skeptical developer audience. PostHog embraced eccentric marketing tactics, including a series of San Francisco billboards styled like 1950s Americana advertisements for tomato sauce featuring Hawkins in retro attire. The campaign deliberately disregarded direct conversion tracking to maximize brand awareness. Similarly, PostHog redesigned its website into an interactive, retro computer operating system featuring an unconventional job board and extensive product documentation. While the redesign initially caused a 10% dip in conversion rates, Hawkins noted that subsequent adjustments resulted in higher net traffic and conversion, validating their willingness to polarize non-technical visitors to delight developers. Scaling Breadth and Building Operational Leverage (0:05) PostHog expanded horizontally across the customer data stack, maintaining 16 to 17 tools in production or development rather than moving upmarket into high-touch enterprise sales. The business expanded from 70 employees to roughly 160 over the course of a year, aiming for 200 while serving approximately 300,000 customers across free and paid tiers, including several thousand paying accounts. Hawkins compares PostHog's trajectory to SpaceX, arguing that a company must build pragmatic, revenue-generating infrastructure before attempting more ambitious technological leaps. In his view, constructing an extensive, multi-product analytics suite earned PostHog the data infrastructure and organizational leverage necessary to pursue autonomous AI agents. Despite reaching unicorn status, Hawkins noted that he still feels the company has barely begun. PostHog illustrates how high-velocity experimentation and unapologetic, developer-focused branding can turn repetitive failure into a multi-product software business. The company's future growth now depends on whether its consolidated customer data platform can effectively power an autonomous AI agent capable of autonomously writing and shipping production code.

PostHog联合创始人兼首席执行官James Hawkins阐述了公司如何从六个月内频繁失败的“转型地狱”中突围，最终凭借开源开发者分析工具切入市场，以14亿美元估值完成7500万美元C轮融资。他主张初创公司必须在营销与产品上做到“非同凡响”，通过高度透明、人格化品牌建立信任，并将筹集资金用于押注AI驱动的自主产品管理体系。 转型阵痛与开源分析工具的确立 (2:05) 在进入Y Combinator（YC）之前，PostHog创始团队经历了漫长的试错阶段，在六个月的时间里每隔五到六周便更换一个想法。其首个销售领域管理工具在联系了15位表达强烈兴趣的早期初创企业销售主管后，最终无一人真正完成注册。Hawkins指出，这一挫败让他们意识到销售主管往往出于礼貌给出虚高评价，而工程师和技术支持人员等天然的问题解决者在言行一致性上更为可靠。随后他们转向针对代码拉取请求（PR）的技术债务调查工具，并保持每周安排10场客户会议、发布3个关键功能的执行节奏；尽管投入巨大努力，甚至发现销售一款每月50美元的SaaS产品与过去销售十万美元的企业合同同样艰难，但客户仍不愿付费，从而证明该产品仅属于非刚需范畴。 转折点发生在YC Demo Day前一个月。由于此前频繁经历产品推倒重来，团队不得不多次从零搭建分析系统，进而切身体会到现有分析软件主要是为产品经理设计、缺乏开发者视角的痛点。Hawkins的技术联合创始人希望直接编写原生SQL查询底层数据，并借助自托管规避广告拦截器的干扰。意识到大量企业正在维护粗糙的自建分析系统后，团队在批次结束前的最后四周内开发出一款开源产品分析工具并在Hacker News发布。该帖子迅速成为2020年该平台上获赞最多的开发者工具，公司也由此从被动推销客户转变为应对大量主动涌入的功能需求与用户反馈。 融资反差与AI自主产品经理转型 (12:40) PostHog的融资历程呈现出鲜明反差。2020年3月正值新冠疫情暴发，原先表达意向的知名风投机构迅速撤资，创始人不得不拜访160家投资机构，靠筹集5000美元的单笔天使支票拼凑出种子轮。Hawkins反思称早期融资受阻源于过度迎合投资人，缺乏坚定主张；当团队明确表态将全力押注开源开发者社区、拒绝过早雇佣庞大销售团队后，才逐步赢得支持。相比之下，公司随后的A至E轮融资总计仅接触了20至30家机构；最近一轮7500万美元的C轮融资甚至仅与YC增长合伙人沟通便迅速敲定，主要为长期战略储备资金与心理缓冲。 此轮融资直接支撑了PostHog在战略方向上的重大跃迁。Hawkins认为人脑作为物理实体意味着通用人工智能（AGI）在技术层面必然可行，因此PostHog决定放弃向客户关系管理（CRM）或客户支持等相近领域横向扩张，全面转向产品自主化。公司正在研发一款桌面应用程序，能够直接处理会话回放、产品分析、错误追踪与大语言模型监控日志，并在工程师夜间休息时自动生成并提交修复Bug的PR代码。其最终构想是构建一个AI驱动的自主产品经理平台，直接承担传统产品经理的需求定义与问题排查职能。 激进透明度与人格化建立信任 (23:01) 针对开发者群体对商业软件普遍抱有的怀疑态度，PostHog从创立之初就确立了“公开构建（Building in Public）”的原则。Hawkins认为，即便开源工具本身免费，开发者仍需投入宝贵时间完成部署并将核心数据托付给软件平台。为了打破隔阂，Hawkins在首次发布前便系统梳理了Hacker News社区常见的质疑点，并在官网上主动披露公司的商业化路径、定价逻辑以及内部经营困境，以此将自身对业务的深刻理解转化为最有效的营销抓手。 此外，团队将品牌人格化视为差异化竞争的核心要素。PostHog早期一篇记录团队夜爬旧金山双峰山（Twin Peaks）的搬迁博客迅速引发广泛共鸣，向社区直观展现了代码背后的真实人物。如今，公司在官网上公示了全部约160名员工的个人主页，允许用户在查看分析功能时点击了解负责该模块的具体工程师及其家庭宠物等生活细节。Hawkins解释称，向用户证明团队并非冷冰冰的企业机器，能够让开发者在互动中感受到真实温度，进而显著增强社区黏性与品牌信赖感。 颠覆性广告牌与两极化复古官网 (26:58) 在缺乏庞大销售团队的架构下，PostHog采取了反常规的营销策略。团队在旧金山街头投放了一系列风格怪异的广告牌，例如让Hawkins身着花边洋装扮演20世纪50年代美国电视节目主持人，以番茄酱广告的形式推广会话重放功能。公司内部设立了专门的Slack评估渠道，以近乎挑剔的社区文化淘汰生硬造作的商业笑话。Hawkins指出，高速公路上的广告牌根本无法直接带来SDK安装转化，其核心目标是在注意力竞争中击败手机与电台，通过足够荒诞与极度细致的设计向开发者传递公司的品味与技术敏锐度。 作为自服务客户的主要成交阵地，PostHog将其官网改造成模拟复古操作系统的交互界面，彻底摒弃了传统SaaS严格设计的销售漏斗。该网站同时承载了16至17款处于不同阶段的产品文档、公司公开手册、超10万订阅者的技术周刊，以及提供“配备何种笔记本电脑”与“研发人员占比”等特殊筛选条件的开发者招聘板。尽管这一改版在上线初期导致转化率下滑约10%，并在网络讨论中引发“惊为天人”与“糟糕透顶”的严重两极分化，但随着流程微调，网站的访问流量与最终转化率均显著超越历史水平。 PostHog的演进印证了在高度同质化的SaaS行业中，依靠高速试错确立真实需求、借助开源透明度与差异化风格建立品牌壁垒的有效性。然而，公司当前估值与业务扩展高度依赖向多产品平台横向整合的推进，其旨在以AI全面接管代码修复与产品决策的自主代理模型，仍面临复杂工程场景下实际可靠性与市场接受度的严峻检验。

[0:05] James Hawkins: haven't got started yet [0:05] James Hawkins: homework being marked by the teacher [3:59] James Hawkins: whatever the coffee equivalent of wasted is [7:18] James Hawkins: full gas [10:08] James Hawkins: sitting cross-legged on the top of a mountain waiting for inspiration [12:40] James Hawkins: sucked at pitching [12:40] James Hawkins: get some revenue [12:40] James Hawkins: build some features [12:40] James Hawkins: all in [16:50] James Hawkins: A human product manager has product market fit—companies buy those [21:06] James Hawkins: unlocked rocket launchers [24:57] James Hawkins: The humanization I think is really important to stand out basically [24:57] James Hawkins: someone's at home [26:58] the sweet taste of understanding [26:58] James Hawkins: weirdness slightly funny scale [32:28] this is literally the best thing I have ever seen [32:28] this website is atrocious [37:21] James Hawkins: I think we sold one [37:21] James Hawkins: Tim alignment [38:01] James Hawkins: have to be remarkable

[0:05] I'm here today with James Hawkins, CEO [0:06] and founder of Post Hog from the YC [0:08] Winter20 batch. James is here hot off [0:10] the news of raising a 75 million series [0:13] Z round of funding at a $ 1.4 billion [0:16] valuation, making Postto the latest YC [0:18] unicorn. Today we're going to talk about [0:20] how Postto got started, how they pulled [0:22] off what I think is one of the best YC [0:24] pivots of the last decade, and how they [0:26] use attention and humor to compete with [0:29] not just other tech companies, but with [0:31] everything out there for attention. [0:33] Welcome, James. Thank you for having me. [0:35] James, could you tell us what is Post [0:38] Talk? Explain it to us a little bit. [0:39] >> Sure. So with YC, I always feel like [0:41] this is my homework being marked by the [0:43] teacher. Um, we help users to debug [0:47] their product to ship features faster [0:50] through kind of things like feature [0:51] flags and so on. And we help keep all [0:52] their kind of customer and product data [0:53] in one stack. [0:55] >> And tell us a little bit about um the [0:57] state of the company, how how big it is. [0:59] You guys are pretty big these days. G [1:01] give our give our viewers a snapshot of [1:03] what's going on at Post Talk. [1:04] >> Sure. Yeah. So, we're about 160 people. [1:07] Um, we're around 300,000 customers [1:09] across like free a lot of free customers [1:11] and paid. um we have several thousand [1:13] paid. We have I think 16 or 17 products [1:16] kind of in production or um in [1:18] development. They're all at different [1:20] kind of stages. And we've been I guess [1:21] like at the moment the main thing we're [1:22] doing and the reason kind of why we did [1:23] a lot of raising this year. Um we're [1:27] pushing very hard on getting more [1:29] products still out of the door. So [1:30] instead of kind of going up market, [1:31] we're going just across more of kind of [1:33] the customer data that people are [1:35] generating. Um, and we're automating a [1:37] ton of things through AI, which has been [1:39] a long and arduous pro process. Um, but [1:42] we're starting to get to the point where [1:43] I'm starting to get proud of what we've [1:44] achieved so far, but I still kind of [1:46] feel like ah, we haven't got started [1:48] yet. There's a lot more to go. Um, which [1:50] I don't think is a feeling it will ever [1:51] go probably, but um, yeah, it's like [1:53] proving a lot of fun and we're just [1:54] starting to scale up pretty aggressively [1:56] now. Like we're I think we start at 70 [1:58] people this year. Um, and so we're [2:01] supposed to be about 200ish by the end [2:02] of the year or something. So it's time [2:03] to get properly and to kind of scale up [2:05] maybe at this point. [2:05] >> So you've got this team, you've got all [2:07] these customers, you have all these [2:09] products that you're developing and [2:10] improving on, but what what was the [2:12] initial product that Post HTog started [2:13] with? What was that the first thing that [2:15] you guys brought to market as Post Hog? [2:17] >> Sure. So yeah, the first thing that [2:18] worked for us, we had a bunch of pivots [2:19] before um was self-hosted product [2:22] analytics because we pivoted so many [2:24] times and every time we had to set up [2:25] product analytics over and over again [2:26] and we're just getting frustrated having [2:28] to implement it. I think we found that [2:30] there's an awful lot of really strong [2:32] competition. Um, but we felt all the [2:34] products were built for kind of product [2:35] managers to force engineers to go [2:37] implement. But as like reasonably [2:38] technical, especially my co-founders, a [2:40] very technical co-founder was just [2:41] getting annoyed at like I can't I want [2:43] to write SQL. I want to see the data [2:44] underneath. I would like to keep this in [2:46] my infrastructure so I don't lose data [2:47] due to ad blockers. So we thought, okay, [2:49] we think there's room for a self-hosted [2:51] alternative. And we also we spent some [2:53] time talking to potential customers. Um, [2:56] and we found that quite a lot of people [2:58] had self-built their analytics stack and [2:59] we're maintaining this kind of junky [3:01] system which we felt if we can [3:04] productize something like that because [3:05] new infrastructure and we could see this [3:06] working quite well. Um, so yeah, we [3:08] started with just open source product [3:09] analytics on hacker news. We spent like [3:11] the last four weeks of the batch uh [3:14] desperately trying to get that out of [3:15] the door and then we hit like we [3:17] actually we didn't really pivot we [3:18] iterated I would say afterwards and now [3:20] we're all cloud-based with multi-roduct [3:21] but that was the thing that landed um [3:24] for us finally I think of postg um well [3:26] there's this meme um among founders of [3:28] pivot hell and it's a thing that YC [3:30] founders talk about a lot where you're [3:32] going from idea to idea to idea and [3:34] you're trying to find something that's [3:35] going to work and you guys not only came [3:37] out of pivot hell and have built this [3:40] awesome [3:41] But what you built was literally forged [3:43] in the fires of pivot hell, [3:45] >> right? Like while doing these pivots, [3:47] you had this pain that only came because [3:49] you were doing the pivots and trying to [3:50] do them at a high level, setting up the [3:52] analytics over and over again. Tell us a [3:54] little bit about how you guys actually [3:56] got to YC and and started this journey [3:59] of of pivoting and finding startup ideas [4:02] that might work. my co-founder uh quit. [4:05] We used to work together. Um and when I [4:06] heard news that he was leaving, um I had [4:10] also been thinking about leaving and [4:11] that kind of triggered me to go, okay, I [4:12] want to do my own thing. Um so I made [4:14] sure to grab him and instead of getting [4:15] him to go off and work in um another [4:17] startup, whatever, I thought, hey, let's [4:18] just do our own thing. [4:19] >> And when he when he was leaving, was he [4:21] leaving to join another company? Was he [4:22] leaving to start something? [4:23] >> Yeah, I think he was actually trying to [4:24] get a job at Facebook at the time. He [4:26] was doing an incremental kind of career [4:28] move. Um and I think I showed up being [4:29] like I think I have this idea that we'll [4:31] probably have product market fit for in [4:32] a couple of weeks time. [4:33] >> Of course. Of course they are. [4:34] >> Of course. Yeah. It's only a hop skip [4:36] and a jump away and then we can just you [4:37] know um I had like a I'd created a list [4:40] of problems I'd encountered [4:41] professionally thinking okay I should [4:42] solve a problem I've had myself and we [4:44] started going through the list but we [4:45] thought we wanted to bootstrap [4:46] initially. So the two of us were just [4:48] like working out of coffee shops. One of [4:51] the main problems we found was every [4:52] time we went to see a customer, we'd [4:53] have to buy another round of coffee. And [4:55] so we'd often end up getting like [4:57] whatever the coffee equivalent of wasted [4:58] is basically. Um so eventually we we did [5:02] this for about 6 months and we worked [5:03] through an average of one idea every [5:06] like five probably average about five or [5:07] six weeks. [5:08] >> Do you remember what any of those ideas [5:09] were? [5:09] >> Yeah. Um so the initial thing we tried [5:11] to build was a sales territory [5:13] management product. So, I used to run a [5:15] sales I was VP of sales and I felt that [5:17] like in sales, you're literally wasting [5:19] over 90% of your time because deals [5:22] don't usually win. They don't usually [5:24] close. Like they usually just float [5:25] around and gradually get killed off. And [5:27] we thought if we get better at not [5:29] focusing on deals that are closing, but [5:30] on the ones that aren't closing, getting [5:31] a bit more ruthless at using statistics [5:33] to pull them out of people's pipeline [5:34] and change the territory for the sales [5:36] team is what I thought might work. Um, [5:37] and it was just kind of really [5:38] complicated um to build and I think [5:41] because I'm much more technical than a [5:43] traditional sales leader would be, I [5:44] just think we built something kind of [5:46] complicated and I don't think our go to [5:48] market was right or anything. Um, yeah, [5:49] we had this idea, we built it. I think [5:52] the thing that was important or the [5:53] harsh lesson I learned was my co-founder [5:55] was building it. I was just trying to [5:56] get interesting prospects to buy it from [6:00] us. Um, so I had about I got through I [6:02] got about 15 sort of series B, C, D um, [6:05] sales leaders queued up saying they [6:07] wanted to use this. Well, they said they [6:09] wanted to use it. We got the first [6:11] version done in a couple of weeks. Send [6:12] them the link to create an account. Of [6:14] the 15 um 14 of them didn't even click [6:17] the link. One click the link and then [6:18] didn't create an account. And we're like [6:20] um it was like first good we should have [6:22] read the mom test was the we hadn't done [6:24] any of the user interviews correctly. [6:25] Um, but I also it made me feel that [6:27] building for sales people or sales [6:29] leaders in particular, I feel the signal [6:31] to noise ratio is pretty poor. Um, [6:34] because they're really quite willing to [6:36] hop on quick calls and are very very [6:37] positive and kind of very friendly [6:38] people. A little bit later it kind of [6:40] dawned on us that we think we'd be [6:43] better because we're not very good at [6:44] product. We're terrible at product [6:45] initially because we're kind of bad. We [6:47] need to deal with people that are what [6:49] they say is close to what they'll do. [6:50] And we felt that like natural problem [6:52] solver people like engineers or customer [6:54] support people probably would be a [6:56] better audience for us to work with [6:57] because out of all the many variables [6:58] you can get wrong. Like it might be like [7:00] your product's incredible but you're [7:01] terrible at explaining what it does. [7:02] You'll still fail. Um and we thought [7:04] that okay one of the variables we can [7:06] remove from this equation is the person [7:09] being a little bit um easier to read [7:11] basically. And so that was a big part of [7:13] why we wanted to build something dev [7:15] tooly later on was that sort of [7:17] realization. [7:18] >> Um but yeah that was the first thing [7:19] that we worked on. And so when you guys [7:20] did Y Combinator, the product that you [7:22] had was another pivot was having [7:24] developers take surveys about technical [7:27] debt when they would submit a poll [7:29] request. [7:29] >> And so you'd already done a few products [7:31] by that point. [7:32] >> You came into YC with another one. Did [7:35] you guys have a like how did you kind of [7:37] keep your spirits up or how did you [7:39] track your trajectory? Did you feel you [7:41] were getting stronger or better or [7:43] closer to the target as you were going [7:44] through these pivots? I would like to [7:47] think that was true. I'm not sure it [7:48] was. We didn't reflect on that. We [7:49] should have asked like are we actually [7:49] getting better each time around? Um I [7:52] think we should have run a retrospective [7:53] basically. I don't think we actually [7:54] were getting that much better each time. [7:56] There are definitely some things that [7:56] were consistently very important to us [7:58] though. Like for example, we went full [8:00] gas on each idea. Like we didn't sort of [8:02] like call it in. Like we would go in [8:04] person to every single customer that was [8:06] remotely interested, we would go in [8:07] person to go meet. Like we started in uh [8:09] England and came to San Francisco to do [8:11] YC. like we'd catch like a bunch of [8:12] trains and buses to get to a customer in [8:14] some random place and like we would put [8:16] the effort in to go really push it [8:18] because then kind of what happens when [8:19] we started charging for a lot of these [8:20] products. You would find that oh I'm not [8:23] failing this cuz I don't have a [8:24] relationship or like I'm kind of [8:25] friendly with this customer at this [8:27] point and they still don't want to buy [8:28] from us. This is a really good sign this [8:30] product isn't solving the problem. This [8:32] really demonstrates to me it's a nice to [8:33] have. Like in my previous job, um, we [8:35] used to sell comp like enterprise [8:37] software for millions of dollars a year. [8:38] And I can remember multiple times [8:40] thinking I'm having to work kind of as [8:42] hard to sell a like $50 a month SAS [8:44] subscription as I was to sell like a [8:46] $100,000 a year plus deal in my last [8:48] job. [8:49] >> I know that I've tried really hard. Um, [8:51] and so I can rule out one of the other [8:52] variables I guess is like, are you [8:53] actually trying as hard as you could be? [8:55] And we were. I think that helped us feel [8:57] confident enough to go this just isn't [8:58] working. So yeah, we just we worked very [9:00] hard. And it look I I remember it [9:02] feeling weird like on LinkedIn for [9:03] example I talk about like hey this is [9:05] what we're up to and it would be like [9:06] one week it was like some dev tool next [9:08] week it's a CRM thing. Um but we just [9:11] didn't care like we just threw ourselves [9:12] into each problem. [9:14] >> There's an aspect of shamelessness that [9:15] you have to have to just kind of keep up [9:17] the momentum and keep trying and not get [9:19] bogged down. I think it's easy for it to [9:20] feel kind of lame almost. Like yeah, you [9:22] have friends who are like in investment [9:24] banking or they have these like proper [9:25] careers and we're doing something that [9:26] just seems so kooky, but like that's the [9:28] bit I think that is a specific reason [9:30] why most people don't run companies [9:32] because that first bit is like all is so [9:34] existential and all over the place. I [9:36] think we just have faith that if we just [9:37] keep making the input we're looking at [9:38] is like are we making very regular real [9:40] progress? So on the ideas we'd often at [9:42] the start of the week be like okay by [9:44] the end of this week what is the list of [9:45] things we will feel good if we get them [9:46] all done and it would be something like [9:48] okay we want to do like 10 custom [9:49] meetings build these three big features [9:52] um and we just held ourselves to account [9:54] on a like very short cadence all the [9:56] time as people I think we got much [9:58] stronger at things like product much [9:59] later on once post dog had sort of taken [10:02] off but I think we were just like very [10:03] we just yeah put our heart into things [10:05] we're very fast that was kind of all we [10:07] needed I think realistically you guys [10:08] were working through these pivots maybe [10:10] getting better, maybe not. Trying to be [10:12] wholehearted, trying to be consistent, [10:13] trying to be really committed to the [10:14] bit, maybe as a comedian would say, when [10:17] you started with the open source product [10:19] analytics, what became the final pivot, [10:21] what signals did you get early on or [10:23] signs that maybe this is it, maybe we're [10:25] really on to something was a month [10:27] before YC demo day. Tell us a little bit [10:29] about how that unfolded. [10:31] >> Yeah, so I think um we had a slightly [10:32] higher level of excitement about this [10:34] idea ourselves. I think we had had a [10:37] couple of ideas where in our heart of [10:38] hearts we weren't that interested in [10:40] them. Like we were kind of doing them [10:41] because we like felt like we had to get [10:42] something to work. Um whereas I think [10:44] this one and we had an office hours here [10:46] and I can remember and we're like well [10:48] we're kind of thinking about building [10:49] this thing that will do two at the time. [10:51] We thought there are two things are [10:52] important. We thought we want to capture [10:53] all user data completely automatically. [10:55] Like we don't want anyone to have to set [10:56] up event tracking manually. We didn't [10:58] realize someone had already built this. [11:00] Um and that didn't turn out to be [11:01] particularly important. But then the [11:02] second thing we're like we want to be [11:03] able to self-host this thing so that it [11:05] can say your infra and we came out the [11:06] office hours I was going ah that's not [11:08] the right framing it's open source [11:09] product analytics then it's not just [11:11] about it being your infrastructure it's [11:13] also just like a whole vibe like I think [11:14] immediately we were like we feel very [11:16] confident I could see that really [11:17] landing on hacker news that we both my [11:19] both read it and it just felt like the [11:21] kind of thing you would see on the front [11:22] page so the level of excitement was a [11:24] little bit higher for us I do think it's [11:26] a bad idea though to like sometimes I [11:27] see other founders they're like I'm [11:28] waiting for an idea that I'm excited [11:30] about but I think The reality is for us [11:32] it's we had to just build a bunch of [11:34] things to find the thing we're excited [11:35] about. What we didn't do was like sit [11:37] cross-legged on the top of a mountain [11:39] waiting for inspiration. [11:40] >> Um so by trying a few things we started [11:42] going this idea just doesn't feel good. [11:44] But we know that because we have tried [11:45] it. [11:46] >> You had to find something in the middle [11:47] of what people want and what you're [11:48] excited about. [11:49] >> Yeah. Exactly. Um and so will I think [11:51] doing stuff gives you much stronger I [11:52] think it's like a Brian Armstrong quote [11:53] but like doing stuff gives you much [11:54] stronger signal more information. And [11:56] then I think when we started booking [11:58] interviews, it was a little bit easier [12:00] to get people to talk to us. Um, but the [12:02] big deal was on Hacker News when it was [12:04] just on the front page and had like it [12:05] wasn't the best ever. Like at the time [12:07] it was pretty good compared to what [12:08] happened before. [12:08] >> It was a good launch. We had a good [12:09] hacker news launch. [12:10] >> But if you look at it at like sort of [12:11] like 20 23 24 vintage hacking news [12:14] launches, it was very mediocre, but it [12:15] was enough to get an initial audience. I [12:18] I think um I because I go back to [12:20] people's old demo day slides a lot of [12:21] the times when helping batch founders [12:23] figure out what they should say. So I I [12:24] pull up the post hog slide every now and [12:26] then and I believe according to the [12:28] slide and I think it's true it was the [12:29] most uh upvoted dev tool hacker news [12:33] post of the year. So pretty good. I [12:35] remember being in the office hour you [12:36] probably shared the idea with a bunch of [12:37] people but I remember the first time I [12:39] heard it and it just made a lot of [12:40] sense. Uh there were a bunch of product [12:42] analytics uh companies out there but no [12:45] one had really gone all the way to make [12:47] this truly developer focused by going [12:49] open source. Um and so it just sounded [12:51] like something someone should try. Uh, [12:53] and uh, I wasn't that surprised that it [12:55] hit when it the way it did and was [12:57] really excited for you guys at the time. [12:58] So, okay. So, you had this um, you were [13:01] starting to get some users. You had to [13:02] go out and raise money for this thing at [13:04] demo day and you'd only been working on [13:05] it for a few weeks at that point. What [13:07] was that like? Like, how do you get your [13:08] conviction up to go, you know, pitch [13:10] investors about a product that's really [13:12] only a month old at this point? It was a [13:13] very mixed. Um, we at first we felt [13:17] really good. We're like, okay, it feels [13:18] like we can have this done in like two [13:18] days or three days. Slam dunk. [13:20] >> Yeah. Yeah, I think we're coming in [13:21] going like, "Yeah, we've had a really [13:22] solid launch." I think we were still [13:23] buzzing a little bit after like oh it's [13:25] such a a feeling of kind of relief that [13:27] suddenly we've got something where it [13:28] just instantly the day we launched it [13:30] went from being a push basis to a pull [13:32] basis getting people interested like [13:33] suddenly it's just like we're trying to [13:34] service all these random problems or [13:36] feature requests and stuff we can see [13:37] coming in and so we went out to raise I [13:39] think very confident like we thought [13:40] we'd kind of be pushing quite big [13:42] valuation at the time and I think we [13:45] felt that we will have this done in just [13:47] a few days um but our the timing was [13:49] horrible because it was 2020 in March [13:51] And the same week basically co went from [13:54] being this somewhat niche thing to [13:56] changing the world and everyone just [13:58] suddenly pulled out. So we kind of went [14:00] from feeling like oh we've got like all [14:01] the fanciest VC firms in the world are [14:03] interested in this thing to everyone's [14:05] pulled out and we're now doing like [14:07] we're trying to get angel checks for [14:09] 5,000 bucks a pop to try and get all the [14:11] way up to like raising a few million [14:12] dollars. This is going to take a long [14:14] time. Um so it was very peaky and we [14:17] also had a lot of stress over like we're [14:18] in the US still. My wife was at home and [14:21] pregnant and I was worried about getting [14:23] stuck in America and my co-founder left [14:26] and because he was also worried about [14:27] getting stranded, overstaying his visa [14:28] and stuff. So there was just all this [14:29] chaos kind of happening. We just kept [14:31] plugging away basically. Um I think it [14:33] was very easy to it would have been very [14:35] easy I think to get disheartened and it [14:37] was painful but we just didn't I guess [14:39] we just kept going um was the main [14:40] thing. And so we were doing angel checks [14:42] um happily taking 5K off of people um [14:46] until eventually we got um a proper term [14:50] sheet um and that led to another one and [14:51] then we um got the round done very [14:53] quickly and then the market completely [14:54] picked up like even just a few months [14:56] later end up doing like a series A like [14:58] a month after the seed round had closed [15:00] but I think the statistics are something [15:01] like [15:02] >> I think we met 160 different firms in [15:05] the seed round. I think for like if I [15:08] added up the total number of pe number [15:09] of firms I've met for the series A, B, [15:11] C, D, and E, [15:13] probably like 20 uh in total like 30 in [15:16] total. It's it was by far the hardest [15:18] round we had to raise. Uh but also we [15:20] sucked it. It's not just them market [15:22] like we sucked at pitching. We're trying [15:24] to people please I think too much. Um so [15:26] we would be like well they were like [15:27] what are you going to do with the money [15:28] and it's like well you know we'll like [15:29] kind of get some revenue. We'll build [15:30] some features which is very vanilla [15:32] feeling. Um, and we had a lot of people [15:34] rejecting us because they felt we were [15:35] too early. But the reality was it's [15:37] like, well, we're dealing with a lot of [15:38] these investors have never invested at [15:39] that time. They never invested in open [15:40] source at all. Like the vast majority [15:42] hadn't. Um, because it was quite it was [15:44] much more unusual then. And I started [15:46] realizing that I don't even believe that [15:48] we'd want to spend money across all [15:49] these areas. Like we actually just want [15:50] to go full all in on the open source [15:52] project for a while. Um, again, I think [15:54] that's very bad advice for something [15:55] that doesn't have competition that's [15:57] making lots of revenue. So be very [15:58] careful with something like that. But we [15:59] started getting much more opinionated [16:01] and that's what started that eventually [16:02] landing. Hey, like you're out of your [16:03] mind if you think we're going to go big [16:04] on like hiring a big sales team or [16:06] trying to go into the enterprise or [16:07] something like it's just about building [16:09] this like nice big inbound kind of [16:10] community. We think is the if we can't [16:12] do that nothing else will follow and we [16:14] need to make really establish that [16:15] properly. Um so yeah, we got much punch [16:16] here. [16:17] >> It's funny how that is. It's almost like [16:18] investors are pretty agnostic on what [16:21] the plan is. So long as there's a plan [16:23] >> and that you you're really passionate [16:25] about it and you've thought about it and [16:26] you're ready to go execute on it. [16:28] >> I think there's a Mark Benoff quote [16:29] which is something like um it's better [16:30] to be different than right. I think that [16:32] it's almost the equivalent here where I [16:34] think it's if you're trying to raise [16:35] money, it's better to have a plan that [16:37] than it is to necessarily be like [16:39] somewhat correct, like be either very [16:41] clearly right or wrong because you'll [16:42] learn something from that. So, that [16:44] round was really hard. You talked to 160 [16:47] investors. It took months to get the [16:49] money. Yeah. Okay. [16:50] >> Fast forward to this this latest round [16:52] that you guys just raised. Obviously, [16:54] it's 5 years later. Things are very [16:56] different. The company's in a very [16:57] different place. You have revenue and [16:58] customers. What's it like to raise a $75 [17:02] million round? You know, how is that [17:04] relatable to folks that maybe are out [17:05] there trying to raise the first money [17:07] for their company? What's that [17:08] experience like? [17:09] >> Yeah, perversely. It's easier. Uh like [17:11] we had um we spoke to one investor for [17:13] this round. We raised from Pete 15. Um [17:15] this one called Chelendra that is the [17:17] main partner, but also Anav um is [17:19] involved as well. The round before he [17:21] tried to take part in, we only had just [17:23] gotten to know him though, so we kind of [17:24] felt like we want to spend a bit longer [17:25] with you before we raised with you. [17:26] Basically, we did spend a little bit [17:28] more time with them because we liked him [17:29] as a person. They offered to do the next [17:31] round preemptively and at first we [17:34] thought it was a bad idea um because we [17:37] just felt we don't know how we would [17:39] spend this. We haven't spent like the [17:40] round before the round before almost. [17:42] But then at the same time what was [17:43] starting to happen was we were [17:46] getting increasingly confident on AI. Um [17:49] what actually happened was I went on [17:50] vacation and read um about some of the [17:54] motives for why some of the co-founders [17:56] built open eye in the first place and I [17:58] thought the reasoning was really [17:58] interesting. It was humans have a brain [18:01] um it's a physical object. We can [18:02] therefore build one basically and [18:04] there's no technical reason we can't [18:05] build ADI essentially. Then I just this [18:07] is nerdy reading but I started thinking [18:09] like what are the specific differences [18:10] between your brain and LMS today and I [18:12] kind of concluded there's a lot of [18:14] things that could be built but I don't [18:15] understand why they don't exist yet. So [18:17] I therefore think that AI will I don't [18:20] think it will flatten out like I think [18:21] the rate of improvement may vary but I [18:23] think over a long time frame it will be [18:25] incredibly important to the world. So I [18:26] got much more um bullish on its [18:28] importance and then thinking at work I'm [18:31] like okay I can definitely see a path [18:32] like I just spent some time reflecting [18:34] with my co I'm like okay how could this [18:36] shake out basically like should we [18:37] change our strategy if we both believe [18:39] this is this important so at the time we [18:41] were thinking about building like a CRM [18:43] a support platform all kinds of other [18:45] data customer data oriented products [18:47] >> and everything was kind of working but [18:48] we felt actually we can do so much more [18:49] in the realm of product at first like we [18:52] can build kind of product autonomy out [18:54] um so we want to go deeper within our [18:56] products that help you understand what [18:58] to build. So we wanted to work on like [19:00] at the moment we're working on like a [19:01] desktop app for example that ships pull [19:02] requests based on your customer data. So [19:04] it'll look across all your session [19:05] recordings or your analytics or your [19:07] error tracking your LM traces go hey [19:09] these are the issues we can see in your [19:10] customer base we literally fix them [19:13] whilst you're asleep. Um here are the [19:14] pull requests. So instead of like please [19:16] build me this feature it will be the [19:18] flow is pull based. So it's like hey [19:20] we've built these features based on [19:22] what's happening. um you can just [19:23] review, close, edit, merge them [19:25] depending what you want to speed up [19:26] development further. Um and we just were [19:29] like, hey, let's just go we want to be [19:30] able to go all in on this idea. Um and [19:32] so we raised to have just like the [19:34] comfort and the confidence. So a lot of [19:35] it psychological like I don't think [19:36] we'll manage to suspend it. Um but we [19:39] want that's kind of the point. It's like [19:40] this means that we can just fully [19:42] embrace um our product strategy and take [19:44] it bigger and just like increase the [19:45] size of swing that we're taking. Well, [19:47] and it sounds like what you're saying is [19:48] uh you're able to fully explore the [19:50] implications. Yeah. Of the original idea [19:52] of post hog, which is understand what's [19:54] happening in your product and what your [19:55] users are up to. And there's all this [19:56] stuff downstream of that like once you [19:58] understand it, then you can do these [20:00] things and take these actions. Yeah. Cuz [20:01] we started thinking like, okay, it's a [20:03] bit like if you're trying if your [20:04] customer is like a painting that you're [20:05] trying to understand. If you have one [20:07] data type, um it's a bit like seeing a [20:09] painting but only be able to see the [20:10] color of blue. You're not going to get a [20:11] nuance understanding of your customer. [20:13] Like looks like they're not using this [20:14] feature. we should email them to cross- [20:16] tell them the new feature we've just [20:17] built. But at the same time, your other [20:18] data might be telling you something [20:19] conflicting like they're really pissed [20:20] off in support right now and this is a [20:22] terrible time to automatically send them [20:23] an email or something. So we kind of [20:24] felt like and that's what a human would [20:26] do it today is you kind of look across [20:28] your tools and we thought oh we can [20:30] basically just build something that does [20:31] this. Um and so we set off trying to [20:33] build basically a product manager that's [20:34] a physical thing we can build it. [20:35] >> Yeah. We don't want to build something [20:36] that has like that we can't conceive of. [20:38] We want to build something that exists [20:40] but just like a better version. And we [20:41] were like well there's a human product [20:42] manager has product market fit like [20:44] companies buy those that is sort of [20:45] where the idea came from. Um so instead [20:48] of just building because normally our [20:49] strategy had been what are the products [20:50] have product market fit have customer [20:51] data we will build all of them it's more [20:53] convenient but then we well also like [20:54] you could consider the team members um [20:56] and thinking about sales has product [20:58] market fit in companies so does support [20:59] so does engineering um so can we build [21:01] those things now because it's now [21:02] possible uh I think it's going to be [21:05] correct that that works but we'll see. [21:06] >> All right. All right. You mentioned to [21:08] me before we started talking today that [21:10] you're h this is like unexpectedly maybe [21:13] you're having more fun than at any other [21:15] point in the company so far. [21:17] >> Um and that is a little counterintuitive [21:18] to people. Uh a lot of founders say oh [21:20] the early days are the most fun and they [21:22] kind of look back nostalgically. Why do [21:25] you think you're having more fun now [21:26] than ever before? [21:27] >> The the work feels much more leveraged [21:28] now which is quite entertaining. I think [21:30] before you're so much of our focus was [21:32] trying to get attention from people [21:34] because you have no attention by [21:36] default. It's the world is very noisy [21:38] and so I'm just like sending emails and [21:40] LinkedIn messages into the ether [21:41] basically and it just feels like I'm [21:42] doing an awful lot of stuff that has no [21:44] impact. Obviously when you do that like [21:45] a tiny little spark it then leads to a [21:47] big fire. Whereas now um it feels like [21:50] we're playing computer game. We've just [21:51] unlocked like rocket launchers or [21:52] something. So it's like, oh, we can like [21:54] go absolutely nuclear on this particular [21:56] idea and like really fully build it out [21:59] in a way that would be very hard to [22:01] we're like lucky enough to be able to [22:02] build this because it involves having to [22:04] have like 15 or 16 products and also a [22:07] bunch of AI stuff and we need to be able [22:08] to like have the time to build this [22:10] because like there's been remarkable [22:12] work at like intercom where they built [22:13] pin um and what we're trying to do is [22:15] analogist to that I think but it's not [22:17] making posto work with kind of what a [22:19] couple of products it's trying to make [22:20] it work across like 16 of them. Um, I [22:22] feel like we're going much deeper than [22:24] we've ever been. But it's enabled us to [22:26] make a bet like that. Like I think [22:27] similarly if we're starting from [22:29] scratch, trying to build what we're now [22:30] trying to build would be impossible. I [22:33] think like SpaceX or something um is a [22:35] much more grandiose idea. Like on day [22:37] one it's not building um like the [22:40] transportation from Mars. It's like day [22:42] one's like we need to build actually a [22:43] relatively unambitious satellite [22:45] launching business. I think we've been [22:47] we've sort of done quite a lot of the [22:48] homework. That means we're able to now [22:50] >> do these take the bigger test basically. [22:52] >> I wonder how many products Post Hog [22:53] would need to have to be as grandiose as [22:56] ambitious as SpaceX. Yeah, it's an [22:58] interesting comparison. You know, you [23:01] mentioned like having fun getting this [23:03] work out there. Like um I know building [23:04] in public is something that you're keen [23:06] on. What does that look like as a CEO is [23:09] getting this attention, getting that [23:10] spark and and putting it out there and [23:12] trying to start a fire. How do you think [23:14] about that and doing it in the open [23:15] where people can see what you're up to [23:16] and hear about it and follow along? [23:18] >> Trust was the original reason for it. [23:20] Like we just felt that even if you're [23:21] offering like most of our users don't [23:22] pay anything like you the open source [23:24] project and no one paid but it still [23:25] takes up there is a cost like it's trust [23:27] in terms of what are they doing my data [23:29] um and also just time with setting this [23:30] tool up. I think especially software is [23:32] getting more like as it's getting [23:33] quicker to build. There's more software [23:34] appearing. It's getting just more and [23:36] more competitive. I think it's got it's [23:37] become a step change more competitive [23:39] with AI. And so we felt that okay the [23:41] fundamental thing we have to achieve in [23:42] marketing is we need to highlight how [23:44] we're different because it's in a busy [23:45] industry. Uh and then we need to build [23:47] trust with users and like the foundation [23:49] of trust we felt was transparency. So um [23:51] instead of just having like a one pager [23:53] landing page we're going to like really [23:55] fully explain everything we can about [23:56] who we are what we're trying to do. I [23:58] think that gave us the trust we needed. [24:00] Um, and we thought about kind of I look [24:01] through other hack and use launches and [24:03] just sort of summarize like okay these [24:04] are all the pieces of critique I can see [24:06] and so we're going to make sure we have [24:07] transparent answers to these types the [24:09] questions would be things like they [24:10] don't have a clear business model I [24:11] don't think this is sustainable for [24:13] example there's like this sort of [24:14] inherent suspicion of ideas and we [24:17] thought okay let's just like think [24:18] through [24:18] >> these types of questions that we know [24:20] that our audience developers would have [24:22] and we'll just address them on the [24:23] website. So we like you know we didn't [24:24] have a pay product. We said hey this is [24:25] how we think the pay product is going to [24:26] work in future. [24:27] >> And that was from like day zero. You you [24:29] put some of those things in practice. [24:30] >> Yeah. Like before we launched Hacker [24:31] News. I think the other thing we [24:32] realized from a content and like a [24:33] marketing perspective was I think [24:35] experts writing about things is more [24:36] compelling often than non-experts. If [24:38] you look at like the front page of [24:39] hacker news there'll be like incredibly [24:41] technically savvy people talking about [24:43] very complicated topics quite often. [24:45] >> And I thought man I'm like don't I'm not [24:47] clever enough to get there but I need to [24:49] figure out a way to get us visible in [24:50] hacker news is like the probably the [24:51] most popular place for our audience. And [24:53] I thought well I am the thing I'm [24:54] actually an expert in is our business [24:56] and so I can write about what we're [24:57] learning and doing and so I feel like a [24:59] really natural topic. So we like we [25:01] wrote a blog post also doing the Y bad [25:03] batch like shortly after the launch of [25:05] what it's like moving San Francisco. [25:06] >> I remember that post. Yeah. [25:07] >> Yeah. Um and again that was also viral [25:09] but I think I build a lot of trust [25:10] because it humanized us underneath. So I [25:12] was like put photos in of like we [25:13] climbed up like Twin Peaks for example [25:15] in the night at night. I think people [25:16] realize like oh there are human beings [25:18] you know in the same way that someone [25:18] might be really aggressive online but as [25:20] soon as you meet in person they're much [25:21] more chill. It's the same sort of [25:22] feeling and we're trying to make that [25:23] clear like with all our products today. [25:25] You can see like the literal engineers [25:27] that we have every single team member on [25:29] the team pages. You can see all the so [25:30] you can look at like product analytics. [25:31] You can see the handful of engineers [25:32] building it. You can click their bio and [25:34] read about like their pet cat for [25:35] example. So that you're like oh there [25:37] are people building this. The [25:38] humanization I think is really important [25:39] to stand out basically. Yeah. It it lets [25:40] people know someone's at home. You go [25:42] visit this house there'll be someone [25:43] there and you'll have a neat [25:44] interaction. I actually I I think about [25:46] your your moving to San Francisco post [25:49] quite often over the years since because [25:51] if you look at it, there's nothing that [25:53] remarkable in the post, [25:55] >> but you got it on Hacker News. People [25:57] read it, they talked about it, it was [25:58] discussed, and I think you're right. [26:00] It's because you just showed some of [26:02] your humanity behind uh this product [26:04] that people could use. And everyone's [26:05] looking for ways to relate to to folks [26:07] like themselves. I remember when I read [26:09] that I thought, "Wow, they can they can, [26:11] you know, climb up Twin Peaks and write [26:13] about it and get developers to want to [26:15] check it out. They've got, you know, [26:17] this is going to go somewhere." Um, [26:19] building in public is one of the ways [26:21] that you guys have really connected with [26:22] developers. I think another aspect of [26:24] that though is is humor and u bringing a [26:28] sense of like marrynt, amusement, and [26:30] maybe even a little chaos to how the [26:32] Post Hog message gets across. One of the [26:34] manifestations of that was present all [26:36] over San Francisco this last year with [26:38] the the crazy billboards uh that you [26:40] guys put up. In my opinion, like there's [26:42] billboards all over the city. When [26:44] people move to San Francisco, they're [26:45] shocked by all the tech ads everywhere [26:48] on all the billboards, but I don't think [26:50] I've seen anyone do quite what you guys [26:52] did with your marketing. And so, I [26:53] thought it'd be really fun to just take [26:55] a look at some of these together and and [26:57] get your take on what's going on here. [26:58] So I guess this first one is about [27:00] session replay ostensibly, but it's a [27:03] also like an ad for tomato sauce and [27:06] it's promoting the sweet taste of [27:08] understanding. Where did you guys come [27:09] up with this and what made you think to [27:12] compare to Tomato sauce to session [27:13] replay for example? [27:14] >> There are a few things I think kind of [27:16] by design we wanted it to it had to be [27:19] different. So kind of our grand theory [27:21] of billboards was how interesting your [27:23] brand is. If your brand's normally here, [27:25] everyone's a bit more interesting on [27:26] billboards. Like if you just look at [27:27] companies seem to come out of their [27:28] shell a bit more on billboards than they [27:30] do on their websites. [27:31] >> Why do you think that is? [27:32] >> I think they're trying to stand out [27:33] basically, but I think it's a bit like, [27:36] again, this is going to sound harsh, but [27:37] I think it's a bit like watching like a [27:39] first grader play football where they're [27:40] going to dabble. Um, but like they're [27:42] starting from a base of being a first [27:43] grader where it's like, well, we're [27:44] already pretty weird online, so our [27:46] billboards need to be bizarre. like they [27:48] need to be like totally out there. So, [27:50] we should like lean harder into this, [27:51] which we think will work because the [27:53] whole point is like it has to stand out [27:55] in the environment and we need people to [27:57] we want people to talk about and see it [27:59] and we're not trying to get conversion [28:00] like it's you're unlikely to decide to [28:02] install like our SDK whilst you're [28:04] driving your car down the freeway. [28:05] Instead of trying to please everyone, [28:06] we're not going to care about [28:08] conversion. We are just going to try and [28:09] raise awareness. And then the other [28:10] thing I think that had struck us was [28:12] from actually doing a bunch of social [28:14] stuff, we're like, "Oh, I can like spend [28:16] the time to write a useful post based on [28:17] something we're learning and that will [28:18] work reasonably well to get some like if [28:20] it's helpful, people will share it." But [28:23] somewhat depressingly if we write like [28:25] if I just write something I think is [28:26] funny, it can sometimes go a thousand [28:28] times further in terms of reach. So if [28:30] I'm just trying to raise awareness, it [28:32] has to be funny. Um it and that's [28:34] actually the primary thing we want to [28:35] get across and and so yeah we just [28:38] really wanted to like crank as far as we [28:39] could over on the like weirdness [28:41] slightly funny scale. Um it is a real [28:44] skill. I think we've been trying very [28:45] hard to I think a lot of corporates or [28:47] whatever do like corporate tryh hard and [28:48] it's very hard to not come because [28:51] you're obviously trying hard. [28:52] >> Your stuff is almost a parody of [28:54] corporate try hard. You've transcended [28:56] corporate tryh hard. It's a very um like [28:58] on hacker news for example people talk [28:59] about like on hacker news it's like [29:00] troll by fire basically where there's [29:02] unbelievable level of criticism our [29:04] marketing channel in slack is like that [29:05] too where people like your joke is just [29:07] not funny um it has to be better this [29:09] you look like you're trying so it's with [29:11] a very um heavy feedback culture in [29:14] marketing. [29:14] >> Let's let's talk about that. So before [29:16] um a post hog uh tweet or billboard gets [29:20] out into the wild [29:21] >> what is the incubation? Yeah, it's a [29:23] very small number of people will be [29:24] involved like probably like three or [29:25] four and it is a very high trust direct [29:30] place and so we're really trying to get [29:31] something that we think is genuinely it [29:34] might be sarcastic but something that's [29:36] genuinely funny to get onto a billboard [29:38] um or genu like just something that's [29:40] that would make our laugh or that we [29:41] would want we are kind of doing them for [29:42] ourselves like is it something I think [29:44] would be quite funny to see is sort of [29:46] the bar it feels like you're competing [29:48] against B2B software because like [29:50] there's all these like AI billboards [29:51] that are quite similar to each [29:53] Um, but I think the reality is if you [29:54] want people to pay attention, it's like [29:55] no, you're competing with like what [29:56] they're listening to on the radio or [29:57] what they're doing on their phone or [29:58] whatever else. Um, and so the bar is so [30:01] many times higher. Um, so it's like it [30:03] needs to be like in the at least in that [30:05] realm. Um, it's not just about being [30:07] like doing a slightly better [30:09] incrementally better job than other [30:10] software companies and which is like [30:12] impossing as Mr. Beast or something, but [30:14] like the mentality is like oh I wish we [30:17] need to be playing that game at least a [30:18] bit more. Um, we felt that, um, we would [30:21] make fun of adverts in general and so we [30:24] thought like, okay, we'll do like a very [30:25] like Americana style [30:26] >> That's right. [30:27] >> ad, but we'll make it like it's like a [30:28] 1950s thing, but for like AI software, [30:31] whatever. [30:31] >> Tastes better than spaghetti. [30:33] >> Yeah. [30:33] >> And you're done up here like the host of [30:35] a um like a 1950s food show. [30:38] >> Yeah. [30:39] >> Uh, talking about your B2B B2B SAS [30:42] product. Okay. And so why why 1950s [30:45] Americana? Like what where was the [30:46] appeal there? Yeah, I think if you sort [30:47] of say like what's the most [30:48] stereotypical kind of billboard ad you'd [30:50] imagine. I think you sort of would [30:51] imagine you would go back to like what's [30:53] the origin of billboards? That's almost [30:54] when you look at them from a positive [30:55] perspective. I think of like oh like [30:56] back in the day ads were kind of cool. [30:58] They're like funnier, better thought out [31:00] whereas now they've just got a bit death [31:01] to humans. [31:02] >> Yeah. Yeah. Or it's just like big words [31:04] on a billboard with like no design [31:05] element to it at all. And we thought [31:06] like actually let's try and make [31:07] something that's kind of a little bit [31:08] like warm feeling um and pretty stupid [31:11] like I was wearing like a frilly frock. [31:13] >> And what about the decision to put [31:14] yourself in all the ads? So, I was like, [31:15] I don't they don't even realize this is [31:17] me or they think they've got some like [31:19] random middle-aged guy to turn up these [31:21] ads. Um, either case is probably fine. [31:23] Um, yeah, we did a lot with hedgehog. We [31:25] did do some with hedgehogs as well, but [31:27] we felt again it's just not quite as [31:29] >> edgy. It's hard to make it not feel like [31:31] too professionally [31:32] >> done almost where it's like kind of [31:34] kooky to have like [31:35] >> a very average person on the ads or [31:37] whatever. Um, instead of like something [31:39] that we could clearly put tons of time [31:40] into. [31:41] >> Too hard on yourself, James. Uh, I mean, [31:42] I agree. Like, I think these are great. [31:44] My wife gives me so much grief. I get so [31:46] excited every time I see a post dog ad [31:48] around the city and my wife is like the [31:50] spaghetti guys again. But it lets people [31:52] know that someone's home. Yeah. [31:53] >> That there's actual people behind this [31:55] and that they really care a lot. You [31:57] know, even things down the details in [31:58] the typography like it sends a message. [32:01] We are going to make a product that's [32:02] also this well thought out and it's also [32:04] trying to get you what you need to the [32:06] same level. [32:07] >> Yeah. I think that's just like a level [32:08] of um signaling that like you know what [32:10] you're doing to pull off an ad like [32:11] this. Um because it I know it does show [32:14] a level of savviness like the website is [32:16] similar in thought process of like we're [32:18] trying to demonstrate someone who isn't [32:19] yet using our product that oh no they [32:21] kind of get technology or they like kind [32:24] of just get me as a person a little bit [32:26] better than like a faceless corporation. [32:28] Let's talk about your website. I think [32:30] you've got the best website of like any [32:32] B2B company maybe in the YC portfolio [32:35] and I get in trouble for saying that but [32:36] it's incredible. Your website's awesome. [32:38] What was the genesis of going in that [32:40] direction? cuz it wasn't always like [32:42] that. You mentioned a little bit ago [32:43] that early on you guys had a lot of back [32:45] content, a lot of lore on Post Hog early [32:49] on, but the website itself was still [32:50] pretty typical um static website with [32:53] some information for a SAS company. But [32:55] now it's this whole experience [32:58] simulating, you know, using a computer [33:00] in post hogland. Where did this come [33:02] from? [33:02] >> I think a few things. I think we kind of [33:05] since like 2021ish, the website's our [33:07] sales team. like there isn't a single [33:08] customer that won't go through the [33:09] website before they buy [33:11] >> the product and especially because of [33:12] nature our audience. We're very self- [33:14] served and so like well we should invest [33:15] in this like it's that we don't have to [33:16] hire a sales team it turns out in the [33:18] first few years or we we had but it was [33:20] tiny kind of like two people and so well [33:22] I think it we can therefore justify [33:24] spending like a ridiculous amount of [33:26] effort on the website. The cost isn't [33:27] that high but like we can put it's kind [33:29] of the energy really because I think [33:30] this will make us stand out. So I think [33:31] a lot of people quickly 8020 something [33:33] you can extremely quickly get to a [33:34] polish website now like in a day you [33:36] have a polish website. So everyone just [33:38] does that and they're like cool I'm done [33:39] now this is 8020 they move on and I'm [33:40] like well like marketing to me is like [33:42] because we're in a busy industry it's [33:43] about standing out and so there's no [33:45] point doing 8020 like there's it's just [33:47] a missed shot. It's like actually all [33:49] the respect on the website has come from [33:50] like the last percent of effort where [33:52] it's like no we're going to go like so [33:53] insanely far past what is normal that it [33:56] is remarkable which means that people [33:58] will talk about it. Um because when we [33:59] talked to users early on it turned out [34:01] all of them are coming to us for people [34:03] were showing up because they'd heard [34:04] about us online or someone recommended [34:05] us. Um and then we said why did they [34:07] recommend you and it was all the tools [34:09] in one low pricing um develop a brand [34:12] and very technical support people [34:14] >> and we have literally just done those [34:15] things and but it's like okay we can't [34:17] just keep doing them as it is we should [34:18] be trying to [34:19] >> crank the dial up further so it's like [34:21] instead of building like three products [34:22] we should build like 300 products um how [34:24] could that work um or on the brand side [34:26] it's like okay post is kind of weird [34:28] it's like it needs to be weirder still [34:30] like it needs to be so weird that it [34:31] polarizes every day we'll get a whole [34:33] bunch of people talk about that website [34:34] and we we monitor all our brand mentions [34:36] online and it's going to be like either [34:38] this is literally the best thing I have [34:39] ever seen and we also get this website [34:42] is atrocious but we feel like we're [34:44] polarizing it largely in favor of our [34:46] audience more like I'm sure there are [34:47] some people who are like what um but [34:50] we're happy to [34:52] >> push so hard in the direction we think [34:53] is something cool for our audience that [34:55] the people that aren't really our focus [34:56] are going to we don't care if they [34:57] disproportionately hate it basically and [34:59] then partly the other thing that [35:00] phenomenon that was happening was it [35:01] didn't just come from like a pure [35:03] aesthetic perspective we genuinely were [35:05] trying to out how do we build a website [35:06] that fully documents and prices for like [35:09] 15 or 16 products. But we also have a [35:11] whole load of other things we built that [35:12] we think are useful. So for example, we [35:13] have a cool developer jobs board. It's [35:15] like a jobs board where the filters [35:16] instead of being just like location or [35:18] pay range or something are like what [35:20] kind of laptop do I get? Is the majority [35:22] of the company developers? The stuff [35:24] that an engineer actually would be [35:25] interested in think about. Yeah. Yeah. [35:27] And we reveal this as like kind of a fun [35:28] project, but we're like well this is [35:29] useful to our audience. We have that in [35:31] there. We have like a handbook that's [35:32] quite popular for people to read and [35:34] learn from. like we you can learn how [35:35] everything works like how much we pay [35:36] people like when we let people go like [35:38] everything is there and so that's like [35:40] another thing we have a newsletter with [35:42] like more than 100,000 subscribers [35:43] that's also in there and so it just felt [35:46] like the form of a website of just like [35:48] a thing lots of pages you scroll through [35:51] it felt like we're doing something more [35:52] multi much more multi-dimensional than [35:53] that that was yeah it's the two things [35:54] it wasn't purely aesthetic it was also [35:56] from experience but we actually thought [35:57] it would be better reflect what we're [35:58] doing post is quite a complicated [36:01] company like again it's tempting to be [36:02] like okay we should just have like a [36:03] single landing page we're like one [36:04] button you click and nothing no other [36:05] information. It's not what our audience [36:07] wants. [36:08] >> You're kind of like spitting in the eye [36:10] of the idea of a funnel. Um it's almost [36:12] like instead of a funnel that you're [36:14] trying to push people through, you're [36:15] just giving them a great piece of cheese [36:17] >> to dig into and enjoy and savor. Well, [36:20] we were wondering what would happen to [36:21] conversion rate and we were like we kind [36:23] of knew this is going to tank the [36:24] conversion rate is going to get we like [36:26] this get much more traffic and the [36:27] conversion rate is going worse. And I [36:28] was like, well, like if we ongoingly had [36:29] way more traffic and a worse conversion [36:31] rate, I if I could like double our [36:32] traffic and half our conversion rate, I [36:34] would take that. And so we launched it [36:35] kind of knowing that's probably what [36:36] would happen. But we did also think like [36:38] we think this is interesting enough that [36:39] people will actually switch on for once [36:41] online rather than just like sort of [36:42] like trying to rush to get to the [36:44] button. And like we also thought like I [36:45] mean if a developer can't work out how [36:46] to click get started when there's [36:48] already like three of them on the page [36:49] like you know that marginal person who [36:52] maybe would have would not converted [36:53] like I don't know. It just felt like [36:55] you're playing like a weird game trying [36:56] to appeal to that user. And so we think [36:59] like of the marginal user, this will [37:01] impress enough that they will convert [37:02] despite the fact the conversion [37:03] experience is very unusual. [37:05] >> And so yeah, the conversion rate did uh [37:07] it didn't actually tank. It was like [37:09] 10%ish off the top of my head. We're [37:10] going to do a post on it. It was worse. [37:12] Um we did like one experiment to change [37:14] the flow slightly. Now the traffic's [37:16] much higher. The conversion rate is [37:17] significantly higher too. Um so it's [37:18] performing better um from a conversion [37:20] perspective even though it's like wildly [37:21] complicated. My favorite part of the [37:23] website, you have this uh unhinged merch [37:26] shop that has some some deep cuts in it. [37:28] And my favorite of the deep cuts is um [37:31] visitors can buy a signed photo of James [37:33] and Tim, the founders of Post Hog, and [37:35] it's sold out. Someone actually came and [37:37] bought [37:37] >> I think we sold one. So, in fairness, I [37:40] I love the yacht rock aesthetic of the [37:42] photo, by the way. [37:42] >> Yeah, we've had quite a few voting um [37:44] off sites. This is supposed to represent [37:46] Tim alignment, but I I was thinking [37:48] about it on the way here and I was [37:49] actually think I think the reason that [37:50] feels funny is you go like this is [37:52] represent Tim and I being aligned and [37:53] close together, but this is what aligned [37:55] would be [37:57] >> we're close but something else this. [37:59] Yeah. [38:01] >> All right. Well, um thank you so much [38:03] for spending time with us today and [38:04] talking us through all this. Uh I know [38:06] we have so many folks out there that [38:08] follow um you the videos that we put out [38:10] there looking for ways to break through. [38:12] Yeah. And I think your point about like [38:14] you have to be remarkable uh is obvious [38:17] but overlooked so much today. And so [38:19] thank you for sharing some of the ways [38:20] that you guys have figured out how to [38:22] try to stand out and be remarkable. And [38:24] congratulations again on this latest [38:25] round that you guys just raised. We're [38:26] so happy for you guys excited here at [38:28] YC. [38:28] >> Yeah, thank you. Yeah, it was u [38:29] monumentally important to us to go [38:31] through the program. Um so um yeah, turn [38:32] if you're listening like go apply to YC. [38:35] >> All right, thanks for joining us today.