How to Commercialize Open Source (Ali Ghodsi, Co-Founder & CEO of Databricks)
发布时间 2021-03-18 21:24:05 来源
Ultimately, you will be a multi-product company. We don't go around today and thinking, Google is an amazing company because they have this search algorithm called PageRank. That's why they are great. No, nobody thinks of that that way, right? If you're going to be successful, you're going to transcend this one project that you're starting now.
最终，你们将成为一家多产品公司。我们今天不会走到处想，Google是一个了不起的公司，因为他们有这个被称为PageRank的搜索算法。这就是为什么他们很棒。不会有人这样想，对吧？如果你想取得成功，你需要超越你现在开始的这一个项目。
Welcome back to B2B SEO. Over the next few months, some of the episodes of the podcast will feature the very best discussions from FC Build, our recent enterprise conference that included interviews with 50 CEOs and executives. On this episode, I chat with Databricks co-founder and CEO Ali Goatsi for a second time.
欢迎回到B2B SEO。在接下来的几个月里，我们的一些播客会采用FC Build（我们最近的企业会议）的最佳讨论，其中包括50位首席执行官和高管的采访。在这一集中，我将再次与Databricks联合创始人兼CEO Ali Goatsi进行对话。
Last time around, Ali focused on the early days of Databricks and talked about the 0-10 journey. If you haven't listened to that episode, it's one of my favorites and I would highly recommend it. In this episode, Ali talks about how to scale and build a $28 billion software powerhouse. We talk about many of the challenges of scaling, including refreshing leadership teams, expanding the product footprint, and personally growing as CEO. Ali also talks about his future plans and why it's still day one at Databricks.
上一次，Ali关注了Databricks的早期发展，并谈到了其0-10的历程。如果您还没有听过那一集，那是我最喜欢的一集，我强烈推荐您去听听。在这一集中，Ali谈到了如何扩张并建立一家价值280亿美元的软件巨头。我们探讨了许多扩张的挑战，包括更新领导团队、扩大产品范围以及CEO的个人成长。Ali还谈到了他未来的计划，以及为什么在Databricks，仍然是第一天。
I want to welcome Ali Goatsi, co-founder and CEO of Databricks, where I'm a seed investor to the FC Build Conference. Ali is a PhD from the KTH Royal Institute of Technology from Sweden, who joined UC Berkeley in 2009 as a visiting scholar and worked closely with Scott Shrenker, Jan Stoika, Matej Zaharia, at the Amplab. He then co-founded Databricks with Jan and Matej in 2013, leading engineering and product management, and became the CEO in 2016.
我很高兴地欢迎Ali Goatsi，Databricks的联合创始人兼CEO，他是我在FC Build会议上的种子投资人。Ali毕业于瑞典皇家理工学院，获得博士学位，于2009年加入加州大学伯克利分校作为访问学者，并与Scott Shrenker、Jan Stoika和Matej Zaharia在Amplab密切合作。然后，他于2013年与Jan和Matej共同创立了Databricks，负责领导工程和产品管理，并于2016年成为公司的CEO。
While Ali probably can't comment on valuation, I routinely get offers to buy my shares at north of 10 billion. And it's widely rumored that the company is going to go public sometime soon. Ali, thank you for joining us today. Thank you for having me.
虽然Ali可能无法对估值进行评论，但我经常收到超过100亿美元的购买股份的要约。而且有广泛的传言称，该公司很快将上市。非常感谢你今天能加入我们。谢谢你的邀请。
Ali, perhaps you can start by sharing the founding story of Spark and Databricks. Yeah, happy to. We were researchers at UC Berkeley, and it's a public university has not that much funding, but the interesting thing is that Silicon Valley forward tech companies, the Facebook, Google's of the world, they're interested in the talent at these universities. So they spent quite a bit of amount of money at UC Berkeley, you know, in research and most of it just to get closer to the smart talent.
阿里，也许你可以先分享一下Spark和Databricks的创立故事。好的，很高兴分享。我们是加州大学伯克利分校的研究员，这所公立大学没有太多的资金，但有趣的是，硅谷的科技公司，如Facebook、谷歌等，对这些大学的人才非常感兴趣。所以他们在加州大学伯克利分校投入了相当一笔资金，主要是为了接近聪明的人才。
That enabled us to basically get a glimpse into what they were doing under NDA, of course. And we got to see what they were doing, and they were doing things very, very differently. Okay, so they were using massive, massive amounts of data, and they were using that data in a very strategic way to disrupt the industry that they were sort of competing with.
那使我们基本上能够窥探到他们在NDA下的所作所为，当然要保密。我们看到了他们的所作所为，他们所做的事情非常非常不同。他们使用了大量的数据，并以非常战略的方式利用这些数据来破坏他们所竞争的行业。
You know, whether it was Facebook and the media industry, or Airbnb and the hotel industry, or Uber and the cab medallions, they were actually enabling their organization to use data across the organization, you know, hundreds of people, thousands of people using this data, and then they were coming up with various ways to optimize using machine learning AI, and they were getting ahead of the competition that way.
你知道的，无论是Facebook和媒体产业，还是Airbnb和酒店行业，或者是Uber和出租车牌照行业，它们都能让自己的组织跨越整个组织使用数据——数百名员工、数千名员工使用这些数据。然后，他们通过使用机器学习和人工智能找到各种优化方式，从而超越竞争对手。
So, you know, we wanted to do the same thing, and we were Berkeley hippies. So we said, we'll open source what these guys are doing. They're not potentially, you know, they're more proprietary stuff. We'll open source it, we'll give it to the world for free, and you know, all the enterprises on the planet can then do the same thing, instead of just these huge Silicon Valley for tech companies, and that is that, and then we'll move on, and we'll publish research papers, and we'll be happier as to our lives. But it turned out a little differently for research.
因此，你知道，我们也想做同样的事情，而我们是伯克利的嬉皮士。所以我们说，我们要开源这些人正在做的事情。他们有更多的专有技术，我们将开源它，免费提供给全世界，然后全球所有企业都可以做同样的事情，而不仅仅是那些硅谷的科技公司，然后我们就可以继续前进，发表研究论文，享受更快乐的生活。但对于研究而言，事情的进展有些不同。
Yeah, I mean, what basically happened is that Ben Horowitz showed up in 2013, and kind of said, you guys are idiots, you know, if you really want to, you're onto something, you've got something good going here, but if you want to change the world, you can't expect other people to do it for you. You got to take that. You're not doing it yourself. Yeah, you got to do it yourself. If you guys don't do it, nobody else is going to do it. So this is yours to lose. So that kind of got us thinking that, you know, maybe we should, because, you know, confidentially, I haven't shared this before, but we went and knocked on a bunch of companies' doors, companies like Cloud Air and so on and we said, we have this awesome technology.
嗯，实际上发生的事情是，本·霍洛威兹在2013年出现了，说你们真的很傻，知道吗？如果你们想要改变这个世界，你们不能期望其他人替你们完成。你们必须自己去做，如果不是你们自己做，就没有任何人会做。这是你们的机会，你们要好好把握。这让我们开始思考，也许我们应该去找一些公司，比如Cloud Air之类的公司，我们告诉他们，我们有这个了不起的技术。
Why don't you take it, you know? And they said, well, but you created it, who's going to be governing it? I'm saying, look, we're academics. We don't want to make any money. You take it, you run with it. And you know, nah, we don't want that. This is academic, you know, Mambo jumbo. We don't really think that there is anything here. It's not enterprise-ready. Nah, we're not interested.
为什么不接受呢？他们问道，但你创造了这个东西，谁来管理它？我说，我们是学者，我们不想赚钱。你们接手，你们运营。但他们说，不，我们不想，这只是学术性的胡言乱语。我们不认为这里有什么实际价值，也不满足企业使用的条件。我们不感兴趣。
So, you know, we were given the runaround for, you know, probably four years across Silicon Valley. We would go on 101, we would, we were poor, so we would rent a car. We'd go down on 101, we'd visit Facebook and all these places. I'd please take this technology and they didn't want. What did it at the time? Yeah.
因此，你知道的，在硅谷我们被拖来拖去了大约四年。我们开车上101，因为我们很穷，所以要租车。我们会去拜访Facebook和其他公司，请求采纳我们的技术，但他们都不屑一顾。现在想想，那时我们做的事情是什么？是的。
Well, you know, things for Cloud Data have turned out a little differently compared to DataPrix. So, you know, it's amazing how big a deal Hadoop was at the time and, you know, and what's happened since. But shifting here from the founding story of DataPrix to company, perhaps, you know, maybe we talk a little bit about your story, which is so unique as well. You know, academia growing up in Sweden to UC Berkeley, to being a VP of engineering, and now eventually to CEO.
嗯，你知道，云数据的情况与DataPrix相比有些不同。所以，你知道，当时Hadoop有多么重要，以及之后发生了什么，真的很惊人。但是从DataPrix的创始故事转移过来，也许我们可以谈谈你的故事，它也是如此独特。你从瑞典的学术界成长到UC Berkeley，再到成为一名工程副总裁，现在最终成为了CEO。
Yeah. What are some of the highlights?
是的。有哪些亮点呢？
Yeah. So, look, I actually think, the way I think about these companies, it's, they go in stages, right? In the first stage of DataPrix, when I was the head of product, and I'll get to it in a second, you're really just focused on making sure that you have product market fit. Yep. And that's more important than anything else.
嗯，你看，我认为这些公司的发展是分阶段的。在DataPrix的第一个阶段，也就是我担任产品负责人时，我们在努力确保我们的产品符合市场需求，这比其他事情都更重要。
So, in some sense, the first part of my career prepared me for that, right? What you do as a researcher in computer systems, is you spend a lot of time trying to come up with technology that really works and people adopt and use. In some sense, it's sort of at school, where you learn how to get product market fit. Yeah. You know, these systems are schools.
因此，在某种程度上，我职业生涯的第一部分为此做好了准备，对吧？作为计算机系统研究员，你会花很多时间尝试开发真正有效并被人们采用和使用的技术。从某种程度上来说，这就像在学校里学习如何获得产品与市场的匹配度一样。是的，你知道，这些系统就像学校一样。
So, that's what I did the first part of my life. You know, I grew up in Sweden. We actually built a bunch of open source technology back there. But then I got the chance to just come visit UC Berkeley one year. And the rest is history. Yeah, I got permission from my girlfriend at that time, and I said, look, one year, and it's okay, one year, you have to be back. So, I spent one year here, and it was so amazing. We started messes, spark, you know, attack on all these projects that were having a lot of impact.
所以，这就是我人生的前半部分。您知道，我在瑞典长大。我们当时实际上建造了很多开源技术。但后来我有机会去加利福尼亚大学伯克利分校访问一年。剩下的故事就都成了历史。我的当时女友同意我可以去一年，然后就要回来。于是我在这里待了一年，这是如此令人惊奇。我们开始开发Messes、Spark，以及攻击所有那些具有很大影响的项目。
So, you know, I said, hey, can I stay another year? And then, can I stay another year? Can I stay another year? And I think after. That was 11 years ago. Yeah, exactly. And after three years, I think she dumped me. But, yeah. So, then in 2013, we got this opportunity.
所以，你知道的，我说：“嗨，我能再呆一年吗？”然后，“我能再呆一年吗？”“我能再呆一年吗？”之后，我想是 11 年前。对，确切地说是 11 年前。三年后，我想她把我甩了。但是，在 2013 年，我们获得了这个机会。
But when we started Databricks, the first phase was the phase of really understanding the enterprise customer. And understanding that, you know, what we had built at UC Berkeley was not really a fit for enterprises. So, it actually needs to be quite a bit adapted. So, that's what we did the first three years.
但当我们开始创建Databricks时，第一个阶段是真正了解企业客户。了解到我们在加州大学伯克利校区所建立的产品并不完全符合企业的需求，因此需要进行相当的修改和适应。因此，在过去的三年中，我们一直在进行这方面的工作。
Then, when I became CEO 2016, that's when you actually, once you have the product market fit, you know, you know, this product is something the market wants and needs. Now, you have to really spin up the go-to market side. And this is oftentimes where they bring in a professional CEO from outside, which has also its downsides. That's when I transitioned. And, you know, I was lucky that I had seen two startups earlier and I had done an MBA earlier in my career. So, I had a little bit of preparation, but that's when we focused on the go-to market side.
2016年我成为首席执行官时，产品市场适配之后，我们知道这款产品正是市场所需要的。此时，我们必须实现市场推广。这通常需要公司从外部引入一位专业的CEO，但这也有其不足之处。那时我进行了转型，很幸运之前经历过两个创业公司并且在职业生涯初期拥有了MBA学位，因此有所准备。这是我们开始重点关注市场推广的时候。
And today, I would say we're past that go-to market phase two. Now, Databricks is sort of in the scale, multi-product, optimizing, you know, the whole sort of organization. When you have thousands of employees, operational efficiency, and operational rigor is more important than anything else. So, that's the phase we're in now.
今天我会说我们已经完成了前往市场的第二阶段。现在，Databricks 处于规模、多产品、优化等整个组织的状态。当你拥有数千名员工时，运营效率和严格的操作比其他任何事情都更重要。因此，这是我们现在所处的阶段。
That's a great articulation. I think these three phases of finding product market fit, figuring out a scalable, repeatable go-to market, and then at some point, you're, you know, you're scaling the organization around that go-to market.
这是一种很好的表达方式。我认为找到产品市场适配的三个阶段是：确定可扩展、可重复的市场推广方法，然后在某个时间点，你会围绕着这种市场推广方法扩大整个组织规模。
You know, one of the questions that founders asked me again and again, I probably get this question once a day and there's no right answer, but I'm curious to get your perspective is, what is product market fit? How do you know you have it? Yeah.
你知道，创始人们一遍又一遍地问我一个问题，我可能每天都会被问到这个问题，没有确切的答案，但我很想知道你的看法，那就是，什么是产品市场契合度？你如何知道是否具备这种契合度？
Look, I think product market fit. There are lots of books you can read on how to get product market fit. And so on. I think most of them are bullshit. You know, this is my mask, you, you, you live in it. The hard way, the early days were easy.
我认为产品市场适配已经实现了。有很多书可以教你如何实现产品市场适配等等。我认为大部分都是胡扯。你知道，这是我的假想，你，你，你都生活在它之中。艰难的方式，早期的日子很容易。
I will just say this, and I don't have a, you know, people won't like this answer, but it's, there's science and then there's art. How do scale go-to market machine? There's a science behind it. And the people have done it. If you read the right books and you hire the right people, you can do it. Operating efficiency is the same thing. Product market fit is not science. It's an art.
我只想说，这个答案可能不会让人满意，但是有科学和艺术之分。如何扩大营销机器规模呢？这背后有一定的科学原理，而且已经被有相关经验的人总结成书。只要雇佣合适的人才，你也可以做到。然而，市场适配度并不是简单的科学原理，更多的是靠创意和艺术。
So, if I say, how do I create symphony nine? Like, tell me what I, what do I need to do? Or I want to do a painting that's as beautiful as Mona Lisa. What do I need to do? Do I need to like, you know, practice my, you know, muscles in the arm? Or like, do I need to like listen to certain music to get in a certain mood? And, you know, it's, it's hard because it's humans on the other side that have to really love what you're creating. So, I would just say fast iteration cycle and trying it out and testing it is, is the key thing.
如果我想问，我该如何创作第九交响曲？比如说，告诉我该做些什么？或者我想画一幅像蒙娜丽莎一样美的画，我需要做什么？要练习手臂的肌肉吗？或者听特定的音乐来进入特定的情绪？但是，这很难，因为观众是人类，他们必须真正喜欢你所创作的东西。因此，我建议快速迭代循环和试验是关键。
We had an advantage. We had an unfair advantage at Databricks, which was when you're at UC Berkeley in a lab, and you're working on many of these projects, and you're funded to, you know, the research, research. Right? So, you have many, many years to try it out. And there were probably hundreds of projects that were created in those years that I was at UC Berkeley. Most of them didn't have product market fit. The ones that had are the ones that, of course, we spun out and did make sense. So, in some sense, it's, you know, it's, we get to cheat a little bit at these research labs. But I don't think there's like a formula you need to follow.
我们拥有了一个优势。在Databricks，我们有一个不公平的优势，那就是当你在UC Berkeley实验室工作，你在许多项目上工作，并且你获得了资金，进行研究，研究。对吧？所以，你有很多年的时间去尝试。在我在UC Berkeley时，可能有数百个项目被创建。其中大多数没有产品市场适配性。那些拥有市场适配性的项目，当然是我们推出并合理化的。因此，在某种意义上，我们在这些研究实验室中可以有点作弊。但我认为并没有必须遵循的公式。
And I think your spot on, I think the, the trick to product market fit is you sort of sense it and you feel it when you have it. When one of my CEOs, Moad Aaron, has used the expression, which is, you know, which is the very high bar, is when the average salesperson can sell the product to the average customer. But that's a very high bar. But I think when you start to feel like demand is coming and you're not having to push to get every customer to develop the product, there's a certain point when that inflection point happens and you kind of know you have it.
我认为你说得很对，我认为确保产品市场适用性的关键是感受到它，当你拥有它时，你会感到它。我的一位CEO Moad Aaron曾经用过这个说法，即当普通销售人员能够向普通客户销售产品时，这就是非常高的门槛。但这是非常高的门槛。但我认为，当你开始感觉到需求正在增长，而你不必要推动每一个客户来开发产品时，当这种变化点发生时，你就知道你已经拥有了产品市场的适用性。
Yeah. I want to talk a little bit about the GoToMarket. You talked about how the number one challenge for technical founders is figuring out the GoToMarket. Yeah. And you talked a lot about how it's more of a science. Yeah. Can you talk a little bit about your journey in figuring out the GoToMarket for Databricks?
是的，我想谈一下GoToMarket。你曾经谈到技术创始人最大的挑战是找到适宜的市场。是的，你也谈到它更多是一门科学。你能谈一下你在找到Databricks的GoToMarket上的经历吗？
意思：这段话讲述了一个关于创始人在找到合适市场方面的挑战。提到了GoToMarket更像是一门科学，并询问对方能否谈一下Databricks寻找适宜市场的过程。
Yeah, let me take a step back and just start by saying, usually the movie goes as follows. And it's actually kind of a sad movie. It's technical startup founder. Yeah. We're technically deep. Build this awesome thing. It has product market fit, able to get some revenue out of it, but not enough, the board, you know, and see, and then they bring in, then they say, look, we need to make a change. We need to change the CEO.
是的，让我先退一步，简单地说一下，通常电影的情节如下。这其实是一部悲伤的电影。它述说了一个技术创业者。是的，我们是技术型的。他建造了一样很棒的东西，拥有了产品市场适配性，并能从中获得一些收入，但收入还不够。之后董事会开始考虑，然后他们说，看，我们需要做出一些改变。我们需要换掉CEO。
And then they think, okay, what's broken? What are we trying to fix? Well, certainly these guys have the tech covered, right? Go to Marked, but, you know, it's the revenue is not there. So let's parachute in a professional CEO. And then the board then naturally says, well, who would be really good at growing revenue? Well, someone, some CEO, professional CEO, who's prior in their life either to done sales or marketing, because that's what we want to grow. Yep. So that's the person that they hire. So that's usually what happens. And in my opinion, that's kind of, it's why that grows the revenue in the next two, three years, but it'll actually usually be the death of the company.
然后他们会思考，好的，有什么问题需要解决？我们想要修复什么？显然，这些人已经掌握了技术，对吗？去找Marked，但你知道，收入不足。所以让我们空降一个专业的CEO。然后董事会自然会问，谁会擅长增加收入呢？嗯，一个CEO，一个在以前生活中或是销售或是市场营销方面表现很好的专业CEO，因为这就是我们想要增长的。是的，那就是他们要雇佣的人。这通常就是发生的事情。我认为，这种做法通常可以在接下来的两三年内增加公司的收入，但实际上往往会导致公司的死亡。
Because those people don't actually understand products, and they don't understand markets, how products work. So the company kind of loses its sense of direction. Unless you can keep the co-founders deeply engaged and working really well with the new CEO, and they can continue. Which is very hard. Yeah, it's hard, right? It's hard. That's a marriage, you know, that, you know, really needs to work out well from both sides. And, you know, they really need to trust each other and be mature, and that's hard.
因为那些人实际上不理解产品，也不理解市场，以及产品的工作原理。因此，公司会有点失去方向感。除非你能让联合创始人深度参与并与新CEO良好合作，并且他们能够继续合作。这非常困难。对，它很难，是吧？这就像一次婚姻，需要双方都能良好地合作，彼此信任，成熟，这很难。
In the rare cases where the tech guy, like me, gets a chance to actually do the go-to-market side, I think there's a few things you need to think about. It's not just, hey, there's a go-to-market thing, you just deploy it. You just need to get the pros, you do go-to-market, and then you're done. It's actually not that simple, because it depends on who, with market you're selling to, how they will buy, and who that persona is, and how do you get that budget from them? What's required to do that? That's different for different products.
在像我这样的技术人员在罕见的情况下有机会真正做市场推广的情况下，我认为你需要考虑几件事情。这不仅仅是“嘿，有个市场推广的事情，你只需要部署它。”你需要考虑你要卖给哪个市场，他们如何购买，以及你是面对哪个人格（目标人口），以及你如何从他们那里获得预算。每个产品所需的情况都是不同的，所以这并不简单。
Actually, one of the mistakes we made at Databricks early days was that we thought, you know, we'll let them swipe a credit card, and then come do AI on our platform. And actually, that didn't happen. Turn on enterprises don't like doing that. Exactly. And even if they do, it doesn't matter, because here's the thing. Databricks builds AI, you know, for enterprises. And this AI is extremely strategic for them, okay? So strategic that they're happy to pay $5 million of dollars, okay?
实际上，我们在 Databricks 处于早期阶段时犯的一个错误是认为我们可以让用户刷一张信用卡，然后在我们的平台上进行 AI。但实际上，这并没有发生。大企业不喜欢这样做。即使他们这样做了，也没有意义。因为这是这样的一种情况：Databricks 为企业构建 AI，这对他们来说是非常战略性的，好到他们愿意支付 500 万美元。
How do you get in any company on the planet, a million dollars approved to spend on anything? That's a big decision. Is that some engineer that's sitting there, you know, coding who swipes a credit card and just puts a million dollars on it? So enterprises, in our case, where the AI speeds are higher, the average selling price is higher, and you need to have budget for this decision, and you need to get high up in the organization.
你怎样才能在全球任何一家公司中得到一百万美元的开支批准？这是个大决定。难道是有个工程师，在那里编码，然后刷了个信用卡，就把一百万美元花掉了吗？在我们的情况下，AI的速度更快，平均售价更高，因此需要有预算，并且需要得到高层管理层的支持才能做出这个决定。
That's a particular go-to-market, you know, the enterprise sales motion that's needed for that, that's the bus all you need to build. And you don't get to just pick that. So like, when you receive, you don't just get to say, oh, you know, I want my SP to be $5 million, and I don't want to have any sales people, and so on and so forth. You have to, for the product you have, and the market that you want to address with it, you have to find the right go-to-market strategy for that.
这是一个特定的市场定位策略，你知道，为了实现它需要一个企业推销的动作，那就是你需要建立的总线。你不能自由选择这件事情。因此，当你接收到产品时，你不能仅仅说，哦，我想让我的销售额达到500万美元，而且我不想有任何销售人员等等。你必须为你所拥有的产品和你想要触及的市场找到正确的市场定位策略。
For Databricks, it ended up being enterprise sales. But if you take something like Altrix or Tableau, or many other products, it's different. It's as many users as possible. Each license cost maybe $1,000, $2,000. And once you have enough of those people in the organization, you can maybe consolidate it. Same thing with Slack. That's a different go-to-market motion.
对于Databricks来说，他们最终采用了企业销售模式。但是，对于像Altrix或Tableau等许多其他产品来说，情况是不同的。它是尽可能多的用户。每个许可证的成本可能是1000美元或2000美元。一旦你的组织中拥有足够多的这些人，你可以可能进行整合。Slack也是同样不同的市场推广方式。
So you need a leader on the sales side that you need to hire that can understand, you know, this first, it's almost like product market fit, but it's for the go-to-market side. Yep. Who is strapping of the go-to-market side? So I think there are three phases on the go-to-market side.
所以你需要雇用一个销售领导者，他能理解这个，你知道，它几乎像是产品市场适应性，但是它是针对去市场的一面的。是的。谁是应对市场的专家？所以我认为在应对市场的一侧有三个阶段。
（意思是说你需要雇用一个能理解产品市场适应性并且能应对市场的专家，并且对应对市场的三个阶段有深入的了解。）
The first phase is to figure out, you know, the- What is the price point? Who's the buyer? What's the persona? What their decision-making process is? Yeah. So that's phase one. That's phase one. And the person who does that is very creative sales go-to-market leader.
第一阶段是弄清楚，你知道的，价格点是什么？买家是谁？人物是什么？他们的决策过程是什么？是的，这就是第一阶段。这就是第一阶段。而负责这项工作的人是非常有创意的销售市场领袖。
And it's basically the zero to 10 million story, you know. This is the person that will take you to the zero to 10. They're highly creative. They're actually really smart. And they're experimenting with you as a co-founder on how to do that. So that's phase number one.
基本上，这是一个0到1000万的故事。这个人将带领你从零开始到1000万。他们非常有创造力，而且非常聪明。他们与你作为共同创始人一起尝试实验如何达成这个目标，这就是第一阶段。
Phase number two, once you know that, okay, this thing sells. Now we just need to hire people. There's the growing, the growing. The growers, they're really good about how to put the, you know, when it sells ops, when it's this ratio of, you know, sales engineers and this and that. And they'll configure it for scale. Yep.
第二阶段，一旦你确认这件事情能够卖出去，那么现在我们只需要招人。有一个问题就是成长，这些种植者非常擅长如何配置销售运营、销售工程师等比例，以便于扩大规模。是的。
And then the third phase is the, you know, optimizers or the maximizers. This is the exacting the value. Yeah, at that point, it's, you know, how do you drive efficiency out of it? Because it's a very costly affair. Go-to-market can be extremely costly. You can, you know, you can burn down your whole bank account, literally, and many, many founders have.
第三个阶段是优化器或最大化者。这一阶段需要精确地提取出价值。在这个阶段，重点是如何从中提高效率，因为推广市场是非常昂贵的。你可能会烧光你的整个银行账户，很多创始人也曾这样做过。
So how do you actually optimize that? Is it going to be essentially one of the most important things you do to get the profitability? So that's a different leader. So when you're a technical co-founder, you need to find that pairing of a go-to-market leader that can take you to zero. Yeah, zero to 10. Now, it might be one, you know, you find the, you know, genius that can do the whole thing for you, but it's usually not the case.
那么你如何实际优化这个过程呢？它会成为获得盈利的最重要的事情之一吗？那就需要有一个不同的领导者。当你是技术创始人时，你需要找到一个能够带领你从零到十的市场引领者。也许你能够找到一个天才，能够为你完成所有的事情，但通常情况下并不是这样的。
Now, I think, I think you make a really good point. I mean, a lot of technical CEOs get paired up with the scale go-to-market leader way too early. And I think as you rightly pointed out, you want someone who actually is creative enough to try half a dozen different approaches, who can try a bottom-up, middle-out, I mean, there's all these different go-to-market approaches. And you don't actually know which one will work.
现在，我认为，你的观点非常好。我的意思是，很多技术型首席执行官过早地与市场推广领袖搭档。如你所说，你需要一个真正有创意的人尝试半打不同的方法，可以尝试从下至上，从中间出发，意味着有许多不同的市场推广方法。你并不知道哪一种方法会起作用。
And so I think having that flexibility in the first, you know, hundred customers or so is really critical. It's almost like figuring out product markets fit the same thing again, but mostly for the go-to-market motion. So get in there and you should be sound to the customers yourself as a technical profaner all the time with your go-to-market leader. Until you guys figured out the messaging, the persona, who you're targeting, how you get that budget, the pitch, all of that. Absolutely.
因此，我认为在最开始的100个客户中拥有这种灵活性真的非常关键。这几乎就像是再次确定产品市场适合度，但主要是针对市场推广。因此，你需要亲自与客户交流，并且一直与你们的市场推广领导人一起与客户沟通。直到你们确定了信息传递、人设、目标受众、预算、推销等内容。完全正确。
We'll be right back. Hi, I'm a part of the NERKRIN. I'm co-founder of Horizon I. Hope you don't mind if I interrupt this episode to tell you a little bit about my company. Horizon is a machine learning observability platform. With the adoption of AIML at an all-time high, it's more important than ever to understand how this technology is affecting your business. When models are deployed in production, we lose all sight of how they're actually performing. Even the engineers who built them couldn't tell you why they're buggy or not doing what they're supposed to do.
我们马上回来。嗨，我是 NERKRIN 的一员，也是 Horizon I 的联合创始人。希望你不介意我在这集节目中打扰一下，让你了解一下我们的公司。Horizon 是一个机器学习可观察性平台。随着 AI 和 ML 的普及程度达到历史最高点，了解这种技术对你的企业的影响比以往任何时候都更为重要。当模型部署到生产环境中时，我们就无法看到它们的实际表现。即使是建造它们的工程师也无法告诉你为什么它们有问题或者为什么它们不能按照预期运行。
Horizon is here to help by providing real-time analytics and observability. The Horizon platform helps your team determine when, why, and how your models are performing. We empower engineers to fix models with explainable analysis and catch upstream engineering issues. So if your team is fed up with the hours spent troubleshooting and debugging your models, you don't have to keep just hoping for the best. You can arise.
Horizon 提供实时分析和可观测性，以帮助你的团队确定模型的表现情况，包括何时、为什么以及如何表现。我们的平台提供可解释的分析，帮助工程师修复模型，并发现上游的工程问题。因此，如果你的团队对花费数小时来排除故障和调试模型感到厌倦，你不必只是盲目地希望最好的结果。你可以崛起。
Ali, I'm going to move around a little bit. There's been a bunch of audience questions around the business model, especially for open-source projects. In United Rixx, in fact, a rare example of a company that started off with Spark, one open-source project, you built a very successful business model while continuing to support Spark as a project. And now you have several other open-source projects that are part of the fold and the family. Can you talk a little bit about your business model and how that's evolved?
阿里，我要稍微走动一下。关于商业模式，特别是对于开源项目，有很多观众提出了问题。实际上，United Rixx是一个罕见的例子，该公司最初只有一个名为Spark的开源项目，您构建了一个非常成功的商业模式，同时继续支持Spark作为一个项目。现在，您还有其他几个开源项目属于United Rixx的范畴和家族。您能谈一谈您的商业模式以及其如何发展吗？
Yeah. So first of all, I think there's two different classes of open-source business models. One is what I call the Red Hat open-source model. Yep. Is I have some free software. I give that to you. And you need my support and services around it to be successful. So it's generally a services business model. Yep. That business model works if you're so lucky, like Red Hat, that there is really no major competition.
嗯，首先，我认为有两种不同类型的开源商业模式。其中一种是我所谓的红帽开源模式。就是我有一些免费的软件，我把它给你。你需要我提供的支持和服务才能成功。因此，这通常是一种服务型业务模式。如果像红帽一样非常幸运，没有真正的主要竞争对手，那么这种商业模式就能够运作良好。
As we saw, we could do many other technologies, oftentimes because it's free, you'll find three, four, five vendors that come in and say, hey, we'll offer that as well. Difficult to monetize your support and services. There will be people that can do support and services better and cheaper than you. And over time, this expertise that you have for the tech that you created becomes commoditized. And the value can usually get to decline over time.
正如我们所看到的，我们可以使用许多其他技术，通常是因为它是免费的，你会发现有三、四、五个供应商会来说：“嘿，我们也会提供这个”。很难从你的支持和服务中获利。会有人可以比你更好、更便宜地提供支持和服务。随着时间的推移，你为所创造的技术所拥有的专业知识变得商品化。价值通常会随着时间推移而下降。
Exactly. The other business model that I think people don't think about enough, but it's like obvious and it's in front of their face. It's the 100 pound gorilla is the SaaS open source model. And just to be provocative, I'll say that one of the best companies in the planet is Amazon Rep Services. In terms of money, not business model.
没错。我认为人们不够重视的另一种商业模式是SaaS开源模式，它非常明显并摆在他们的面前。而且为了引起反响，我会说，在金钱方面，在星球上最好的公司之一是亚马逊服务代表公司。但这并不是指它的商业模式。
That's a business model in which you take open source software and say, hey, you can rent it from me. And you pay me rent for using the service. It's much stickier than the on-prem open source model where they could switch services provider, who does support for them. They could switch from, say, a cloud error to a hotmerks very easily and keep the software. That's the key thing. They could keep the software.
这是一种商业模型，即利用开源软件，并说：“嘿，你可以从我这里租用它。你使用这项服务需要向我支付租金。”这比传统的本地开源模型更加“黏性”，因为它们可以随时更换服务提供商，让支持他们的服务提供商提供帮助。他们可以轻松地从云端变换到Hotmeks，并且可以保留这个软件。这是关键的事情。他们可以保留这个软件。
In the cloud, when you rent the service from AWS or Databricks or anyone else, you can't keep the software if you want to switch vendors. So that business model is much, much better. Why? It's stickier. And it's harder to switch to other things. And it turns out over time, this subscription model grows the revenue much more sustainable. I totally make sense.
在云端，当你从AWS、Databricks或任何其他供应商那里租用服务时，如果想要更换供应商，你无法保留软件。因此，这种商业模式更好。为什么？因为它更具黏性，更难切换到其他东西。随着时间的推移，这种订阅模式可以更可靠地增加收入。我完全理解这个观点。
I think I can interrupt for a second, because the audience has been engaging in this and asking the next level or detail. The challenge with that open source model, with the SaaS model, is that you are now often competing with the cloud service providers themselves. Yeah. How do you answer the elephants?
我想我可以打扰一下，因为观众一直在参与并提出下一个层次或细节的问题。这个开源模型和SaaS模型的挑战是，你现在往往要与云服务提供商竞争。是的。你如何回答大象？
意思是说，该人认为自己可以打断别人来回答问题，因为观众一直在提问更深入的问题。他提到，使用开源模型和软件即服务模型的挑战在于你需要与云服务提供商竞争。在这种情况下，他后面提到“如何回答大象”是指如何面对困难、如何应对问题。
Yeah. Just because some people are confused, there are vendors that are doing open source and they started with this red hat model. OK? And the truth is, they're not very good at running a SaaS service in the cloud. So once their customer base told them that, look, we need to be in the cloud, they said, OK, we'll offer. I mean, I created the software. I'll run it in the cloud.
是的。只是因为有些人感到困惑，有些供应商正在进行开源并采用了红帽模型。好吗？实际上，他们并不擅长在云中运营SaaS服务。因此，一旦他们的客户基础告诉他们，我们需要在云端，他们就会说，好的，我们会提供。我是创建这个软件的人，我会在云中运行它。
The problem is, when they start running in the cloud, they realize, wow, it's really, really hard to run a SaaS service. Yep. That point, the cloud vendors are running super fast and taking their software and offering it up. And actually, they're offering a better service than the open source creators. So my advice is this, if you start a company on open source now, you need to be 100% in the cloud.
问题在于，当他们开始在云上运行时，他们发现，哇，运行SaaS服务真的非常非常困难。没错。在这一点上，云供应商正在以超快的速度运行，并将他们的软件提供出来。实际上，他们提供的服务比开源创建者更好。所以我的建议是，如果你现在用开源创立公司，你需要完全投入到云中。
And you need to be 100% SaaS. Don't touch on-prem. Get really good at managing SaaS software in the cloud. If you do that, then I think you can absolutely compete and beat the cloud vendors. Databricks is an example of that. Because it's actually really hard to operate and run software in the cloud. And who do you think is the best in the world to operate that software? The creators.
你需要100%的SaaS。不要碰On-Prem。在云端管理SaaS软件要变得非常优秀。如果你做到了这一点，我认为你绝对能够与云供应商竞争并胜出。Databricks就是一个例子。因为在云中运作和管理软件真的很难。而谁是世界上最擅长操作该软件的人呢？就是创造者。
Databricks has a competitive advantage when it comes to operating things like MLflow, things like Delta, things like Spark, the technologies we created, no one can run them as well at scale, reliably, and cheaply than us. The cloud vendors cannot do that. We literally are creating the technology for that.
Databricks在像MLflow、Delta、Spark这样的技术操作方面具有竞争优势，没有人能够像我们一样在规模、可靠性和成本方面运行它们。云供应商无法做到这一点。我们真正为此创建了技术。
Had we started on-prem though, that would not have been true. Because on-prem, we would have gotten really good at this support and services thing. And then five, six years later, we would have tried to just, you know, maybe we should just offer something in the cloud. That turns out it's very hard. So I think that's what's really going on.
如果我们一开始选择在本地部署的话，这个说法就不成立了。因为在本地部署，我们会非常擅长支持和服务方面的事情。然后五六年后，我们可能会尝试在云上提供一些东西，结果发现这非常难。所以我认为这就是真正的情况。
So I say you can compete with them. These guys have their hands full, running VMs, data centers, storage, and a lot of other things. They're not going to be world class at operating your open-source software that you invented if you focus on it. But you have to focus on it, and it's not easy.
所以我认为你可以与他们竞争。这些人已经忙于运行虚拟机、数据中心、存储和很多其他事情。如果你专注于你发明的开源软件，他们不可能在运营方面达到世界级水平。但是你必须专注于它，而这并不容易。
So I think clearly the message, I think, hopefully the audience is hearing from you, is don't have one foot each in two boats. You know, you've got to commit 100% to being on the cloud first. You've got to commit to being a SaaS service provider from day one. And it's often very different from the ethos of an open-source project community.
我认为你传递的信息非常明确，希望听众能够听到，就是不要两只脚都踏进两只船里。你需要100%的投入并决定成为云服务的供应商。你需要从一开始就致力于成为SaaS服务提供商。这与开源项目社区的理念常常非常不同。
Because those projects are really around improving the project, adding features, adding functionality, adding integrations. And now you have to build a whole new engineering culture around running a cloud service. Yep. Presumably, you end up building a lot of proprietary software in that process. You have to.
这些项目的主要目的是改善项目、增加功能、整合系统。现在，您需要建立一种全新的工程文化来运营云服务。而在这个过程中，您很可能需要大量开发专有软件。这是必要的。
Here's a dirty secret that nobody thinks about. Every, to my knowledge, every open-source project is an on-prem product. You go download it from GitHub. There's a version number. You download it. You can install it. That's shrink-wrap software. That's not a SaaS service.
这是一个被人忽略的不太光彩的秘密。据我所知，每个开源项目都是一个内部软件产品。你可以从GitHub下载它，还有一个版本号，你可以安装它。这就是收缩包软件，而不是SaaS服务。
SaaS service, like if you use Databricks today, you don't even get to see the version. It gets upgraded all the time under the scene and so on. So in some sense, it's two different technologies. One is a SaaS service. The other one is shrink-wrap gets released x times a year software.
SaaS服务就像你今天使用Databricks一样，你甚至不能看到版本号。它在幕后一直升级等等。因此，在某种程度上，它是两种不同的技术。一种是SaaS服务，另一种是每年发布x次的收缩包软件。
解释：SaaS是指对软件的一种使用方式，即以服务的形式提供软件，而非用户自行安装软件。在使用SaaS服务时，用户通常无法看到软件的具体版本号，因为软件会在幕后不断升级。与之相对，传统的软件则是以收缩包的形式发布，并在之后的某些时间里进行几次更新。这两种方式可以说是完全不同的技术形式。
So yeah, by definition, if you have an open-source project, and you also have a SaaS service, the SaaS service is actually a different thing. And a lot of the secret sauce can be there and how to manage and operate and run it. And that, it doesn't even make sense open-sourcing that. There is no open-sourced.
是的，按照定义，如果你有一个开源项目，而你也有一个SaaS服务，这些其实是两个不同的东西。许多秘密酱料可能存在于如何管理、运营和运行SaaS服务。因此，公开开源它甚至都没有意义。这是一个没有公开源代码的事情。
I don't think there is an open-source version of MongoDB or pick your favorite open-source project. I don't think there exists a SaaS version of it in open-source. Got it. And so those are two engineering teams that you have to build and two sort of engineering-shipping motions inside the company that you have to support over time.
我认为目前没有MongoDB或者其他开源项目的开源版本。我认为在开源中不存在它的SaaS版本。明白了。因此，你需要构建两个工程团队并支持公司内部的两个工程/交付活动。
Yeah, but put your weight behind the SaaS one because that one is revolutionary different. It's fundamentally at its core very different from how you do R&D on-prem. At its core, it's fundamentally different in sort of five different ways. So it's impossible to just figure that out later.
是的，但是请把你的重心放在SaaS上，因为那种方式是革命性的不同。从根本上来讲，它与在本地进行研发的方式完全不同。在其核心上，它从五个不同方面根本上不同。因此，仅凭后期的解决是不可能的。
Look, is there any other advice before we move on, change topics? Is there any other advice you have for founders who are building companies that are on open-sourced projects? Anything else, you know, in terms, because you've done now this with a couple of different open-sourced projects. You know, ultimately you can't be one trick pony. So ultimately you will be a multi-product company.
看，我们在继续之前，转换话题之前还有什么其他的建议吗？对于正在建设基于开放源代码项目的公司的创始人，你有什么其他建议吗？在这方面，您已经完成了几个不同的开源项目。最终，您不能只有一个技巧。因此，最终您将成为一个多产品公司。
We don't go around today and thinking, Google is an amazing company because they have this search algorithm called PageRank. That's why they are great. No, nobody thinks of that that way, right? If you're going to be successful, you're going to transcend this one project that you're starting now. Keep that in mind. And don't pire yourself too closely just to one tech. This is something we did at Databricks early days. When we started Databricks, we said, hey, our original invention was spark at that time. These days, it's a small portion of what we do. But back then, we were saying, should we name the company something with Spark?
我们今天不会一直想着谷歌是一家很棒的公司，因为他们有一个叫做PageRank的搜索算法。这就是他们伟大的原因。对吧？如果你想成功，你需要超越你现在正在开始的这个项目。记住这一点。不要过分依赖某个技术。这是我们在Databricks早期所做的事情。当我们开始Databricks时，我们说，嘿，我们最初的发明是Spark。但现在，它只是我们所做的工作的一小部分。但当时，我们在考虑，是否应该给公司取一个带有Spark名字的名字。
And as founders, we said no, because we will transcend Spark, we will have a whole portfolio of products eventually. And Spark will be just one of them. And now, in hindsight, it was really important because today, the most important open source project we have, I would say, is Delta. So that's a cautionary tale. There are companies out there, for instance, Docker to company and Docker to software is the same thing. It's same name. And you can stream them. It blocks them for a long time.
作为创始人，我们说不，因为我们将超越Spark，最终将拥有整个产品组合。 而Spark只是其中之一。 事实证明，这非常重要，因为今天，我认为最重要的开源项目是Delta。这是一个警示故事。 有些公司，例如Docker公司和Docker软件是相同的东西。它们的名称相同。你可以流式传输它们。它会阻塞它们很长时间。
Yeah, so this is just psychological advice to my fellow co-founders. Eventually, you have to divorce yourself from that open source project you created if your company is going to be successful. I think that's great advice. And it's one that you have to make that decision very early on. So thanks for sharing that.
嗯，这只是给我的联合创始人们的心理建议。如果你的公司要成功，你最终必须要与你创建的开源项目分道扬镳，而这是一个极好的建议。你必须在早期就做出这个决定。感谢你分享这个建议。
So one of the questions coming from the audience from Greg is, in the $0 to $10 million phase, as you're building out, you're figuring out you're going to market. Do you recommend that you have just one go to market leader, or should you have a separate leader for sales and a separate one for marketing?
来自Greg观众的一个问题是，在0到1000万美元的阶段，当你在构建自己的公司并确定你要去哪个市场时，您是否建议只有一个马上投放市场的领导者？还是应该分别设置销售和市场营销的领导者呢？
Most VCs will give you advice that you first hire the product marketing leader, because they come up with the marketing material, because the sales guy doesn't know what to say otherwise. And then you hire the sales guy. I actually disagree with him. And I think you should do it in reverse order. Titles don't really matter. But I think you need a partnering crime that essentially is going to be your GTM co-founder of yours. And that person might join several years later once you have product marketing with. And whoever that is, together with that person, you're going to figure it out with him or her. Maybe that person has a marketing title, but most likely they have a sales title.
大多数风险投资公司会建议您首先雇用产品营销领导者，因为他们能够提供营销材料，而销售人员否则不知道该说什么。然后您再雇用销售人员。我实际上不同意这种观点。我认为您应该采取相反的顺序。头衔并不是很重要。但我认为您需要一个合作同伴，本质上将成为您的GTM联合创始人。这个人可能在几年后加入，一旦您有了产品营销策略，与这个人一起确定。也许这个人拥有营销主管的头衔，但更可能是销售主管的头衔。
Because really what you're doing in that phase is you go into the customer, and you're basically giving them a value prop and putting a price tag on it. Yep. OK? And you're going to see if it sticks. And so you experiment with different stakeholders, with different types of companies. And you're doing this while you're changing the value prop, you're changing the product a little bit to see maybe it's a different product I should. So you need a thought part of that. I think that's the sales person.
在这个阶段，你要做的就是去面对客户，向他们提供价值主张并确定价格。是的，你要不断实验不同的利益相关者和不同类型的公司来看是否可行。同时，你需要改变价值主张并微调产品，以确定最适合的产品。因此，你需要有思考的能力。我认为这个阶段的人应该是销售人员。
Once you have that figured out that I think I know what it is. It's roughly this value prop, but probably the way I explain it is way too complicated. And it's typically this persona. And this is roughly what the product looks like. That's when I think you bring in a marketing leader.
一旦你确定了这一点，我认为我知道这是什么了。它大致是这样的价值主张，但我解释的方式可能太复杂了。通常是这种人设。产品大致看起来是这样的。这时我认为你需要引进一位营销领袖。
And I would say there's basically two different types of marketing leaders for that stage of the company. There is the messaging people, and then there's the managing people. The managing people can drive the man. And then there's the messaging people. They love the story, the narrative. They can tell you a nice story. My opinion is you hired that person, the storyteller, that then takes your super complicated, technically correct message and makes it really available to a wide audience so that it resonates with a big audience. And most importantly, that you can actually teach it to a sales guy in Texas who doesn't care about your company. He's going to work there one or two years to make some money and leave.
我会说，在这个阶段的公司中，基本上有两种不同类型的市场营销领导者。有传递信息的人，还有管理人员。管理人员可以驱动人。然后有传递信息的人。他们喜欢故事，叙述。他们可以给你讲一个好故事。我的意见是，你雇佣了那个讲故事的人，让他把你的超级复杂、技术正确的信息变得容易理解，以便广泛的观众能够共鸣。最重要的是，你实际上可以教给德克萨斯州的销售人员，他对你的公司不感兴趣。他可能在那里工作一两年赚钱然后离开。
You want that guy who has, he's not going to kill himself. He wants him to be able to quickly pick up that message and run with it. That's the second hire is that marketing leader. The third hire I think should be the finance leader. Because if you don't hire finance leader, the go-to-market leaders will, as I mentioned earlier, empty of bank account. And the wheels start to fall off. So you don't want to, as a CEO, spend all your time all day long, micro-managing where they're spending the money. And why are you paying 350 KOT to the sales guy?
你想要找那个人，他不会自杀。你希望他能够快速接收到这个信息并立即行动。这就是第二个招聘的营销领导者的目的。第三个招聘的应该是财务领导者。因为如果你没有雇用财务领导者，市场领导者会像我之前提到的那样，耗尽你的银行账户。然后，雪球就开始滚落。所以，作为CEO，你不想把所有的时间都花在微观管理他们的支出上。你又何苦要为销售员支付350 KOT呢？
And so on, you don't want to. That's you shouldn't be doing that. Have a professional CFO whose job is to make sure that you still have money in the bank account. I think that sequence makes a lot of sense. So it's hire a salesperson or sales leader that is creating, that has a business development bone in their body, that can help you figure out the go-to-market.
所以，你不想这样做。那么你就不应该那样做。聘请一个专业的财务总监来确保你的银行账户持续有资金。我认为这个顺序非常合理。所以，聘请一个销售人员或销售领导者，他们有开发业务的能力，可以帮助你找出市场营销的策略。
Then bring on board a marketing person to scale that approach. And I think you rightly pointed out that, you know, there's so many different skills in marketing. You have to pick one to hire for first, one spike. And the spike is around storytelling. And then they can build the mansion around them. And once these people start spending some money, some serious money, you need the finance person to come in and sort of corral them a little bit. Yeah, and here's why the marketing one, just to make that point, how many startups do you guys know where you go to their web page?
然后聘请一名市场营销人员来扩大这种方法的应用。我认为你说得对，市场营销中涉及很多不同的技能。首先，你必须挑选一种技能作为起点雇用人才。这个起点技能是关于讲故事的技能。然后他们可以在这个基础上构建一座大厦。当这些人开始花费一些严肃的资金时，你需要财务人员来进行一定的规划。是的，市场营销的原因是，你有多少了解到他们的网站的初创企业？
And you're like, I don't know exactly what they do. What do they do? And you talk to the founder, and they say, oh, it's very simple. We take Docker and we have a KBM layer. We modify the Linux kernel, in which we can actually do the virtual operations really, really fast. What do you exactly do? So that's why those companies are in dire need of product marketing storyteller. Completely agree. You know, we have these panels sometimes.
你会像这样想，我不太知道他们具体做什么。他们是干什么的呢？你和创始人谈话，他们说，“很简单，我们使用Docker并有一个KBM层，我们修改Linux内核，这样我们可以非常快速地进行虚拟操作。”你会非常想知道他们究竟做了什么。所以这些公司非常需要产品市场宣传员来讲故事。完全同意。有时我们会举行这样的小组讨论。
And we had, I remember the last panel, we had eight CEOs who all described our companies. We're all in one space, and security is an example in this case. And it sounded like all the eight companies did the same thing. And they actually did eight different things. But the storytelling is, because if you abstract it to two higher levels, it's meaningless. And it was too detailed, no one understands. Yeah, and finding that right abstraction, I think, is challenging.
我记得在最后一个小组讨论中，我们有8位CEO介绍各自的公司。我们都在一个地方，以安全为例。尽管这8家公司实际上做的是8种不同的事情，但听起来好像都在做同样的事情。然而，仅仅抽象到较高的层面，是没有任何意义的。而且，过于详细的描述会让人难以理解。因此，找到正确的抽象水平，实际上是具有挑战性的。
Yeah, absolutely. On the go to market side, you know, another thread that's been coming through in the questions from the audience has been around pricing. You know, you talked early on about how customers are willing to pay you millions of dollars, and you said you've got to just try a lot of things. You've got to throw things at the one and see what sticks. Any other observations and pricing beyond that?
没错，市场推广方面，观众提出了一个重要问题，也是一直以来被提及的——定价问题。你早先提到过客户愿意支付数百万美元，而你必须尝试很多不同的定价策略，试图找到可行的方法。此外，你还有其他的定价方面的观察吗？
Yeah, we should do a whole separate podcast just and pricing. So I think it's an extremely important topic. When people say product market fit, I see that as part of product is pricing. It's one of the four piece of market product, actually.
是的，我们应该专门开一个关于定价的播客节目。我认为这是一个非常重要的主题。当人们说产品市场适配时，我认为这是产品定价的一部分。实际上，这是市场产品的四个元素之一。
In general. So I think you need to take it super seriously. I have actually run a product pricing meeting since we started Databricks every week, still to the state, seven years in. And we still run it every week. Every week, it's called PGTM pricing, go to market meeting, and it's every week an hour and a half since the start. I'm not saying you have to do that. And otherwise, you're not going to get pricing. We're here for a different trick. So for every CEO, at least you run it for a couple of years and see what happens.
一般来说，我认为你需要非常认真地对待它。自从我们成立Databricks的那天起，我每周都会召开一次产品定价会议，直到现在为止，已经有七年了。我们仍然每周都举行这个名为PGTM定价、市场推广会议，自开始以来每次会议都持续一个半小时。我不是说你必须这样做，否则你将无法得到定价。我们在这里来做不同的事情。因此，对于每位CEO来说，至少要运行它几年并观察结果如何。
It's a very important topic, especially as your company is a little bit bigger. You don't need to be thousands of employees. If you're even if you're just 50 people or 100 people, it's actually a tricky thing to change. Because the sales guy will say, well, you just ruin my revenue. And the marketing guy needs to know how to actually market that price. The product guy needs to build it into the product. So there's so many stakeholders involved. Finance.
这是一个非常重要的话题，特别是因为你的公司稍微大了一点。你不需要成为成千上万的员工。即使你只有50人或100人，这实际上是一个棘手的改变。因为销售人员会说，你把我的收益搞砸了。市场营销人员需要知道如何将价格营销。产品人员需要将其构建到产品中。因此，有很多利益相关者参与，包括财务。
So you actually, the only one who can really pull that together and make sure that you get the right pricing structure initially in the early days is the CEO in my opinion. Sales of the account do it unilateral. Finance can do it unilaterally. Product guys don't have enough influence to just push that out. What I would say is you should have WTP discussions, withling less to pay discussions with your customers.
在我看来，实际上只有CEO能真正将其整合起来，确保您在早期阶段获得正确的定价结构。账户销售可以单方面完成，财务部门也可以单方面完成。产品人员没有足够的影响力来推动这一点。我想说的是，您应该与客户进行WTP（意愿支付金额）讨论，了解其愿意支付的最低价格。
You should ask them what they're willing to pay. You should price on a metric that's attached to the value that the customer is getting. So attached to whatever is the thing that really ultimately defines the value, ideally it's something that grows or has multiple drivers behind it, users, data, our spend, whatever it is, impressions, so that your revenue can grow faster than say the number of users are just using it.
你应该询问他们愿意付多少钱。你应该按照与客户获得价值相关的度量标准来定价。因此，应该与实际上最终定义价值的事情相关，理想情况下是某种增长或有多个驱动器支持的东西，例如用户、数据、我们的支出、印象，这样你的收入就可以比使用它的用户数量增长得更快。
If you can attach to such a growth driver, and you're going to have people that are going to come say, hey, we can't price it on that value metric, because if we do that, it prohibits growth. Don't listen to them, because if you pick the right pricing metric, if you're really pricing on value, of course, people will optimize that particular metric.
如果你能依附于这样的增长驱动因素，必然有人会来告诉你：“嘿，如果我们按照价值指标进行定价，妨碍了增长，我们不能这样做”。不要听他们的话，因为如果你选对了定价指标，并真正以价值为价格基准，人们自然会优化这个因素。
Your customers will try to lower that. And yes, it prevents that very thing. If you think users is the main thing you should price on, of course, if you price on that, you won't get as many users, because whoever user has to pay. But it's still the right metric. So I would say, you know, there are a lot of firms out there that can help you with doing the willingness to pay conversation. So I also think you should use them. I've used many of them over the years.
你的顾客会试图降低价格。是的，这会防止这种情况发生。如果你认为用户是你应该定价的主要因素，当然，如果你在这个方面定价，你不会得到太多用户，因为每个用户都必须付费。但这仍然是正确的指标。所以我会说，你知道，在市场上有很多公司可以帮助你进行支付意愿的交流。所以我也认为你应该使用它们。这些年来，我使用过很多这样的公司。
Sounds like you've put a lot of energy behind pricing. I think it's extremely important. I think, if you look at technology space in the last 40, 50 years, many of the way companies disappeared or appeared or won or products won, just came down to pricing. Like how office bundled its software versus Lotus 1, 2, 3, or word perfect, or how big companies buy the second player in the market that's not so good and offer it as a free skew as part of their enterprise agreement. It's all bundle pricing strategies.
听起来你对定价投入了很多精力。我认为这非常重要。我认为，如果你看一下过去40、50年的技术领域，很多公司消失、出现或赢得胜利、产品获胜，都归结于定价。比如说，微软是如何捆绑其软件与 Lotus 1、2、3 或 WordPerfect 竞争的，或者大公司如何购买市场上不太好的第二大玩家，并将其提供为免费的零售产品作为其企业协议的一部分。这都是捆绑定价策略。
So I would definitely spend a lot of energy and time on pricing. One last question on this go to market phase, and then we'll move on. Someone from the audience asked, the carto from the audience is asking the question that, you know, if you're a solo technical founder, you were relatively lucky. I think there were several of you. But if you're a solo technical founder and you're building something relatively hard, some deep tech just like you were, how do you balance between product development and go to market, especially the earliest stages?
所以我肯定会在定价上花费大量的精力和时间。关于这个市场进入阶段的最后一个问题，然后我们就可以继续进行了。观众中有人问道，如果你是一个单独的技术创始人，你可能比较幸运，我想有好几个人是这样。但如果你是一个单独的技术创始人，你正在开发一个相对较难的东西，就像你之前所做的那样，那么在产品开发和市场营销之间如何平衡，特别是在最早的阶段？
You raise a couple of million dollars in the seed. You don't have eight or 10 million dollars in the bank. It's like it's a hard balance every day. Yeah, it's difficult. I mean, for me, I think focus on product market fit, which is make sure that the thing you're building has value. I think it's just too hard to build the perfect company according to a playbook step by step. So if you try to, in seed, a, build the right product, for the right time, make sure that the willingness to pay is really high.
你在种子轮中筹集了几百万美元，但你银行里没有800万或1000万美元。这就像每天都要保持艰难的平衡。是的，这很困难。我认为应该专注于产品市场适应性，确保你正在构建的东西有价值。我认为按照手册一步一步地构建完美的公司太难了。因此，在种子轮阶段，你应该建立适合正确时间的正确产品，并确保愿意付费的意愿非常高。
And that, you know, it's just too many things to optimize at the same time. So I think stage number one, just make sure that this product actually has unique value that's at least 10x more than anything else that's out there for them. Are you a vitamin or are you a painkiller? You know, I forget to take my vitamins every day. That's how important they are. You know, but when I have a headache, I need my painkiller, okay? So are you a painkiller and are you a 10x painkiller? That's better than anything out there. Focus on figuring that out.
你知道，要同时优化太多事情，这是不可能的。因此，我认为第一阶段就是确保这种产品的独特价值至少比其他产品高出10倍。你是维生素还是止痛药？我每天都会忘记吃维生素，这就说明它们并不十分重要。但是当我头痛时，我需要止痛药。所以，你是止痛药，还是10倍于市面上其他产品的止痛药？专注于理清这个问题。
Once you have that painkiller, you can figure out, that's when you can start focusing on the go-to-market side. There's lots of examples. Companies like Facebook, Google, Snapchat, WhatsApp, who, you know, I don't think really had figured out their go-to-market strategy. They just had a vital. It's an amazing consumer demand. And investors invested them and said, they'll figure it out and they did.
一旦你有了止痛药，你就可以开始思考市场推广方案。很多公司都有这样的例子，比如Facebook、Google、Snapchat和WhatsApp等，我认为他们并没有真正想清楚自己的市场推广策略，他们所拥有的只是消费者的强烈需求。投资者投资了他们，并认为他们会想出解决方案，并最终他们也做到了。
You know, I want to shift gears back to sort of, you know, where GitHub exists today. It's a very different company from some of the early days that we've talked about and where a lot of the founders who are listening to this are.
你知道，我想把焦点转回GitHub今天所处的位置。它与我们谈论过的早期日子和很多正在听这个的创始人所经历的情况有很大不同。
You mentioned a few minutes ago that while you started with Spark, Delta has become the most important opensource project in your ecosystem. Can you talk about your journey from being a paink product to a platform for data scientists? And now today, you're a platform serving the needs of multiple user persona, including business analytics users.
你几分钟前提到，尽管你最初使用Spark，但Delta已成为你生态系统中最重要的开源项目。你能谈一下你从一个痛点产品到成为数据科学家平台的旅程吗？现在今天，你是一个为多个用户人物提供服务的平台，包括商业分析用户。
Yeah, absolutely. So it's similar to our story. We wanted to, when we worked Berkeley, product market fit stage, we just wanted to build something way better, 10x better at least. And actually, we came up with something that was roughly 100 times faster than a doop. But once you have that and you start the go-to-market muscle and you start listening to these enterprises, our vision was we wanted to enable these enterprises to do the kind of AI and data science that is super strategic and disruptive to their markets like Uber had or Airbnb had or Facebook had or Google had.
是的，完全是这样。这跟我们的故事很相似。当我们在伯克利工作时，处在产品市场拟合阶段，我们只是想要构建比doop快10倍以上的东西。实际上，我们创造了一款大约比doop快100倍的产品。但是一旦你有了这个产品，开始强化市场营销和倾听企业的需求，我们的愿景就是让这些企业可以进行超级战略性和颠覆性的AI和数据科学，就像Uber、Airbnb、Facebook和Google一样。
And we start working closer with them. It turns out that Spark itself is not enough. It's pretty complicated. And you need actually several pieces of the puzzle. My advice then is at that point, and it's very naturally to see many, many companies doing this, you start looking at what's the critical user journey or what's the jobs to be done? You can Google those two phrases. Yeah. What's the jobs to be done or this critical user journey that a user has to go through to be successful with the job they want to achieve?
我们开始与他们更加紧密地合作。事实证明，Spark本身并不足够。这相当复杂。实际上，您需要几个拼图的碎片。我的建议是，在这一点上，很自然地可以看到许多公司在这样做，您开始看看什么是关键用户旅程或什么是要完成的工作？您可以谷歌这两个短语。是的。什么是要完成的工作或用户必须完成的关键用户旅程，以实现他们想要实现的工作？
Most companies are just a small part of that, right? Is there a way you can consolidate that? You can expand that so you can own and to end more of that. If you can, that kind of vertical integration, you can give benefits that no one else can. Because at the scenes between those product jumps that everyone has to do, there's a lot of friction, a lot of costs, different buyers.
大多数公司只是整个产业链的一小部分，对吧？您有没有办法整合它们？您可以扩大规模拥有更多的产业链环节。如果您能实现这种垂直整合，您可以获得其他人无法获得的优势。因为在产品之间转换的环节中，每个人都必须进行很多摩擦、成本和不同的采购方。
So if you could do multiple of those, you can have a much superior experience as a user goes through that critical user journey to do the job that they need to get done. Whatever it is, I need to come up with a predictive model to reduce fraud on my site. What do I need to do step by step? So Databricks today, now what it has is, Spark is a way where you get all your data into a data lake. Yep. That's just the foundation of accessing all the data that's virtualized in this data mesh everywhere.
如果您能完成多个任务，那么在用户进行必要的任务时，将会获得更加优越的体验。不论是什么任务，我都需要设计一个预测模型来减少我的网站上的欺诈。逐步完成需要做哪些事情呢？今天的Databricks，现在就是使用Spark来将所有数据放入数据湖中的方法。是的，这只是访问虚拟化的数据网格中所有数据的基础。
Once you have that data, Delta is the technology that actually puts quality, reliability, and performance on it, so that data is usable by anyone in the organization. Now you have governance on your data, it's secure. Then we have on top of that something called MLflow, which enables you to do end-to-end machine learning and data science on that data.
一旦您拥有了这些数据，Delta 技术会将其质量、可靠性和性能提高，从而使组织中的任何人都能够使用这些数据。现在您可以对您的数据进行治理，确保其安全。在此基础上，我们有一个名为 MLflow 的工具，能够在这些数据上进行端到端的机器学习和数据科学分析。
And now we've also built something that we just announced a month ago, which is SQL Analytics, which enables you to do the data warehousing workloads, BI and other SQL-based workloads. That now covers the end-to-end journey that typically an organization has to go through to be successful with their data. But for your company, wherever you are, that journey is a different one, it's not a long-term adventure, but think it through in terms of what's the end-to-end journey that they have to go through to be successful.
我们现在也建立了一个月前宣布的SQL Analytics，使您能够处理数据仓库工作负载、BI和其他基于SQL的工作负载。这现在覆盖了一个组织通常必须经历的全方位的成功数据之旅。但对于您的公司，无论您身处何处，这个旅程都是不同的，它不是一个长期的冒险，而是要考虑到他们必须经历的全方位的旅程，以获得成功。
How many steps are there? Are there ones where you would have a competitive advantage if you could do two of them in one platform? Could you do it better? Because I don't need to switch between two products with different formats and different ways of connecting them and integrating them, and is there a way you can do that better? That could be huge competitive advantage for you. And Enterprise will want to buy that, because Enterprise today are sort of inundated with complicated software stacks with four data warehouses and five copies of this and eight copies of that. And it's very hard to govern and manage all of that. So they are always open to, if you could consolidate, a few of them. A little bit patient.
有多少个步骤？如果你能在一个平台上完成其中的两个步骤，是否有竞争优势？你能否做得更好？因为我不需要在两个格式不同、连接方式不同、集成方式不同的产品之间切换，你是否有更好的方法？这可能对你来说是一个巨大的竞争优势。企业也会愿意购买这种产品，因为现在企业被四个数据仓库和五个这个、八个那个的复杂软件架构所淹没，很难管理和控制它们。因此，他们总是愿意如果你能够将它们合并成几个产品的话。有点耐心。
Yeah. I think it's a good segue, in fact, to another question that's been coming in on the machine learning tooling space. What you're doing is sort of much broader in many ways. But the machine learning tooling space in recent years has become very complicated with lots of startups. There's startups for building models, startup for testing models, startup for hosting models, producing, putting them into production.
是的，其实我认为这是一个很好的过渡，可以谈一下机器学习工具空间方面的另一个问题。你所做的事情在许多方面都比较广泛，但近年来机器学习工具空间变得非常复杂，涌现了许多初创企业。有些初创公司用于构建模型，有些是用于测试模型，有些是用于托管模型，生产并将模型投入生产。
And at the same time, every major tech vendor, including the three cloud service providers, also has its own ML tooling platform. So do you have any specific advice for founders starting companies in this category? Yeah.
同时，每个主要的技术供应商，包括三个云服务提供商，也都有自己的机器学习工具平台。所以，您有什么具体的建议，供那些在这个类别中创业的创始人参考吗？是的。
简单说就是各大科技公司都有自己的机器学习工具平台，针对这种情况，您是否有任何建议给想在这个领域创业的人们？
I think MultiCloud is the future. Even the cloud vendor, some of them are now releasing. I think Google released the BigQuery Omni, which is BigQuery should work on other, and so on. And you'll see that from the other cloud vendors too. So that's an advantage you have on your site, which is MultiCloud will be a thing. And a particular cloud service that only works on that cloud is going to actually eventually become a huge disadvantage for that cloud vendor. It isn't today, but eventually it will become.
我认为多云是未来的趋势。即使云供应商们也开始推出一些多云产品，例如Google发布了BigQuery Omni，它可以在其他云平台上运行，其他云供应商也会相继推出类似的产品。这为我们带来了优势，因为多云将成为一种流行趋势。如果一个云服务只能在某个特定云平台上运行，那么它最终会成为该云供应商的巨大劣势。现在还不是，但最终会成为。
So that's a good silver lining. Second, I think the higher up the stack you move, the less competition you will see from the cloud vendors. The reason is this. The cloud vendors are starting with storage, compute, and networking. And they have to nail that. They can't lose that war.
因此，这是一个好的转机。其次，我认为您在技术上越向上移动，就越不会有来自云供应商的竞争。原因是这样的。云供应商从存储、计算和网络开始。他们必须做到精准。他们不能输掉那场战争。
But that's only three or four things. Core ingredients. On top of that, you cannot rebuild many more things. And on top of that, even more, right? So the tree structure just explodes above it. You move two, three levels up.
但这只是三四件事情。核心成分。在此基础上，您不能再重构更多的东西。再加上更多的东西，对吧？因此，树形结构就在其上方爆炸了。您要向上移动两三个级别。
意思：这个句子是在讨论一个问题的复杂性，指出仅有几个核心因素并不能解决问题，而且解决问题涉及到的层面会越来越多。同时，这个问题是很复杂的，因为涉及许多层面和关系。
Now there's like literally thousands of apps, or tens of thousands of apps. The cloud vendors won't be able to do all of that. So it's moving up the abstraction. Yeah, so every hop you move up in this sort of software stack, it's going to be harder for them to dominate all of that. So that would be my advice.
现在有成千上万的应用程序，甚至可能有数万个应用程序。云厂商不可能处理所有这些。因此，它正在向抽象化方向发展。是的，所以每个软件堆栈中向上移动，他们要控制所有这些将变得更加困难。那就是我的建议。
This is not categorical, but I'm a little bit worried for people that are straight up in the core storage or network in the very like, that you should probably assume that the cloud vendors will nail. Because there's three of them that have infinite funding. And that's literally the only thing they're supposed to do. It's not like you really do have infinite funding, especially with the stock market is going these days. Exactly.
这不是绝对的，但我有点担心那些在核心存储或网络方面直接从事工作的人，你应该假设云供应商会占据主导地位。因为有三个云供应商拥有无限的资金。而这确实是他们唯一应该做的事情。但实际上你也不可能拥有无限的资金，尤其是现在股市的情况。正是这样。
I want to wrap up maybe with one last question for you. As you take a step back on what has been a remarkable journey over the last decade, what's perhaps the one insight or the single biggest lesson you have when you look back? If you were giving the 10 year younger version of Ali advice, what would you tell Ali to 10 years younger?
我想用最后一个问题来结束，问你一个问题。当你回顾过去10年的非凡旅程时，也许你会有一个洞见或者最大的教训是什么？如果你给10年前的阿里一个建议，你会告诉年轻的阿里什么？
I would tell him a few things. One is cliche, but the other one I think people might like hearing. The first one I would say is that focus on people that you can trust and finding leaders that you can trust. It's very hard. It's these kind of marriages, finding the people that you can work together really, really closely and figure this stuff out. There aren't enough of them. And even if you figured out the three or four people that you work well with now in two years or three years, when you reach the next stage of your company, who's going to be the next person to do that there?
我会告诉他几件事情。其中一个是陈词滥调，但另一个我认为人们可能会喜欢听到。我首先会说的是，注重与能够信任的人合作，并寻找可以信任的领导者。这非常困难。这就像找到可以共同工作并找出解决方法的人类婚姻，这样的人不够多。即使你找到了三四个你现在与之合作良好的人，在两三年后，当你的公司进入下一阶段时，谁会成为下一个人呢？
So spend much more time on that, much earlier and building trust with those. The second thing I would say that I would focus on much more these days than I did in the past 10 years is I would be more uncompromising in the sense that there's a lot of shortcuts you can take to get revenue. Don't do them. The boards will push you. You want to hit those revenue numbers, but just don't do it. If you're building for a vision of the cloud, just do that. For instance, at Databricks, we built a cloud platform. But then I approved one exceptional deal, which was on prem. Just one.
所以，要花更多时间早期与那些人建立信任。其次，过去10年里我更加注重的事情是更加不妥协。虽然有很多捷径可以获得收入，但不要采取这些方法。董事会会推动你去达到收入目标，但不要这么做。如果你正在打造云计算愿景，就坚持这个方向。比如，Databricks公司打造了云平台。但我批准了一个特殊的交易，这笔交易是在本地进行的，只有一个。
There was only one of them. And it just took so much energy from me and the people involved and supporting it. Don't take shortcuts like that. It wasn't worth it. The important thing I would say is just if I look back at the person 10 years ago, is trust yourself much more. There's a lot of herd behavior. Everybody runs in that direction. But nobody's like, OK, but who's leading us in that direction? I don't know. We're all running this way. Let's go. Think from first principles and stick by it and believe in yourself.
只有一个人。但是它花费了我和参与支持它的人太多的精力。不要走捷径。这不值得的。我想说的重要的事情就是，如果我回顾10年前的自己，就是要更加相信自己。有很多从众的行为。每个人都朝着同一个方向奔跑。但没有人说过，那个方向领导我们的人是谁？我不知道。我们都跑这条路吧。要在首要原则上思考，坚持自己的信念。
And if you think big enough, you might be able to actually pull it off. So I have confidence. That's probably the three things I would tell myself.
如果你有足够大的想象力，你可能真的能做到。因此我很有信心。这可能是我告诉自己的三件事。
That's it for this episode. You can find past episodes and subscribe to future ones on iTunes, Stitcher, and SoundCloud. BWB SEO is brought to you by Foundation Capital, an early stage venture capital firm with 27 IPOs, including Netflix, Lending Club, TubeMobile, and Sundra. I'm Art Shogard, a general partner at Foundation Capital. I'm passionate about helping BWB entrepreneurs who are trying to solve hard problems. So this podcast speaks to you. If you are interested in growing from a technical founder into a business leader, drop me a line. Thanks and see you next time.
本期节目就到这里了。你可以在iTunes、Stitcher和SoundCloud上找到以往的节目并订阅未来的节目。BWB SEO由Foundation Capital赞助，Foundation Capital是一家早期风险投资公司，共有27个上市公司，包括Netflix、Lending Club、TubeMobile和Sundra。我是Art Shogard，Foundation Capital的常务合伙人。我热衷于帮助试图解决困难问题的BWB企业家。所以这个播客适合你。如果你有兴趣从技术创始人发展成为商业领袖，请给我发信。谢谢，下次再见。