- Segment helps companies organize their internal data about customer interactions to improve customer experience.
- The company started by launching an open-source library on Hacker News, which gained traction and attracted small companies as customers.
- They initially built a classroom lecture tool but realized it was a failure when students were more interested in using Facebook.
- They then decided to pivot and build an analytics tool, but struggled to gain customers.
- They developed a small library that could send data to multiple analytics tools, including their own, and open-sourced it.
- This library gained traction and showed potential for growth.

- The founders of Segment felt traction on their routing library and received pull requests, indicating market interest.
- They debated whether to build a business around their open-source library, Analytics.js, with one founder skeptical of the idea.
- They decided to create a landing page for Analytics.js and put it on Hacker News to gauge interest.
- The response was overwhelming, indicating product-market fit and a loss of control as the market dictated what needed to be built.
- The founders experienced another product-market fit moment when they discovered customers were using Segment's data for ETL pipelines.
- Segment's growth primarily comes from engineers and their infrastructure and processing capabilities are of interest to the engineering crowd.
- Most of the value is delivered through the hosted version, while some components are open source.
- The decision to launch early was driven by a divide among the founders and the need to get customer feedback.
- Product-market fit moments are unmistakable and result in a loss of control, with every metric going haywire and people talking about it.
- While some companies can iterate to find product-market fit, it is rare and can lead to staying in an iterative mode for years.

- It is dangerous to stay in an iterative mode for years without making progress.
- The Airbnb story should not be used as a reason to continue with a bad idea.
- It is important to have skeptics on the team who question and push for fast tests.
- Testing should be qualitative and involve talking to customers rather than relying on landing pages.
- Running a sales process and using qualification criteria can help determine if a product is a fit for a customer's problem.
- Messaging should be constantly refined and tested with customers.
- The best ideas come from going deep into a specific market and understanding its problems better than anyone else.

- The speaker discusses the importance of asking deep and detailed questions to understand customer needs and problems.
- They give an example of testing product market fit and how asking in-depth questions led to a better understanding of the customer's problem and potential solution.
- The speaker emphasizes the value of listening to customers and aligning with their vision.
- They also discuss the values and standards of their company, Segment, including karma, tribe, drive, and focus.
- The impact of GDPR on Segment's business model is discussed, highlighting how it aligns with their values and provides new features for their customers.
- The speaker shares a story about asking for more money during sales negotiations and the importance of pushing beyond comfort zones.

- The speaker shares a story about a negotiation where they were advised to ask for $120k but ended up settling for $18k.
- They believe that it is difficult to offend people with a high price in person or on a phone call, but it is not recommended to share pricing information via email.
- Y Combinator (YC) was helpful for the speaker's company, particularly through demo day and the founder network.
- The speaker emphasizes the importance of having a clear product vision and prioritizing customer feature requests based on that vision.
- They have learned a lot about finance and the importance of hiring the right people, including a part-time CFO and HR advisor.
- The speaker believes that there are many exciting opportunities in the startup world, such as in-space rocket engines and industrial inspections by drone.
- To learn more about Segment, visit segment.com.

Segment is a company that helps businesses improve customer experience by organizing their internal data about customer interactions. They initially started with a failed classroom lecture tool but later pivoted to building an analytics tool. They gained traction with an open-source library and experienced product-market fit moments. Segment's growth primarily comes from engineers and they offer a hosted version of their product. The speaker emphasizes the importance of understanding customer needs, refining messaging, and aligning with customer vision. They also discuss the impact of GDPR on their business model and the lessons they learned about finance and hiring. Y Combinator was helpful for their company and they see exciting opportunities in the startup world.

Peter Reinhardt - https://twitter.com/reinpk - is cofounder and CEO of Segment - https://segment.com/ 

Segment helps companies capture data from every customer touchpoint and send it to the tools where it can be used most effectively.

They were part of the YC Summer 2011 batch.

The YC podcast is hosted by Craig Cannon - https://twitter.com/craigcannon

***

Topics

00:00 - What is Segment?

1:30 - Segment’s first customers

3:05 - Their YC application

4:00 - Going through YC

5:30 - Realizing their first product didn’t work

10:30 - Launching Analytics.js

11:45 - Experiencing product market fit

16:55 - Debating whether to launch or build out the product

19:15 - Evan Farrell asks - You mentioned in the SS lecture that you had to totally pivot to Analytics.js to find PMF, is it possible to purely iterate on something people kinda like to find PMF, or should it be clear from the outset if a new idea is something people want?

20:30 - The importance of having a skeptic on your team

23:30 - Customer interviews

26:30 - Benjamin Liam asks - How did they know they have the right messaging to explain their product?

28:00 - Idea generation

32:45 - Danny Prol asks - What values and standards do you have in place for your team at Segment? And how do you actively build that culture into your company?

37:00 - Ashwin Doke asks - How has GDPR impacted Segment's business model?

39:15 - Andrew Pikul asks - Any advice he has on asking for more money than you're comfortable asking for. 

41:45 - Juan Carlos Garza asks - How did YC help you to where Segment is right now?

43:15 - Juan Carlos Garza asks - In an early stage, what's the thin line between ignoring a customer suggested feature or moving a customer requested feature to the core of your application?

44:45 - Biggest learnings since YC

45:50 - Important hires at Segment

The average person probably doesn't know what Segment is.

So Segment helps companies give their customers a better customer experience, and we do that by helping them organize all of their internal data about all their interactions with the customer.

So for example, if you go to the bank, they interact with you at the ATM, at the teller, via a phone call center.

They're interacting with you across this huge surface area.

And they need to be able to coordinate that interaction.

They need to know that if you encountered an error on the ATM, the teller needs to be able to say, I'm so sorry you encountered an error.

And so what we do is we help bridge that gap of having a single record of all those interactions with each customer.

Because previously, companies would build all this in-house, or not at all, maybe.

1 is they would build it all in-house, exactly, and there'd be a rat's nest of data pipelines from 1 place to another.

And so the engineering team would spend all their time building these data pipelines rather than actually building things for the customer.

The other world is 1 where it really used to be a one-on-one relationship with a bank branch manager, for example.

And they might keep the information in a CRM.

But if you are someone who would be an optometrist, You go in, they have your past orders, et cetera.

But now the world is moving much more to a Warby Parker kind of world, where you're not interacting with a person.

It's like not the right technology for understanding what interactions you're having with a customer.

And instead, it's all these different digital channels.

And your first customers, were they large scale companies like banks, or who did you get in the beginning?

Ryan Young At the very beginning, we actually launched as an open source library on Hacker News.

And it took off there, blew up basically overnight.

Peter Gabriel So to be fair, it was kind of like the long road

No, no, let's not skip the year and a half of dark times.

Yeah, yeah, we shouldn't totally go over that.

But okay, yeah, so anyway, you blew up overnight

Once It went live on HN, it blew up overnight.

And so our first customers were the folks hanging out on Hacker News.

It was basically small companies for the most part, founders who were looking for better ways to instrument their web applications and mobile apps with this sort of analytics tracking.

And so the initial growth was, that was just completely unpaid customers, right?

So they're just using the open source, whatever you put up on GitHub, right?

Walker So the strange thing about the open source library, which is sort of this data router, so you put in 1 piece of data, customer did X, and then we turn around and transform it and fan it out to all the different places downstream.

And what's funny about the library is if you want to turn on a new tool, if you want to send the data to a new place, you need to recompile the library and redeploy it to your website.

So the open source library doesn't quite actually solve the problem, which is really a marketer or a product manager wants to use a new tool.

So what happens is you really want to use the hosted version.

So almost no 1 actually uses the open source version.

There's just no way to make a good open source product around it.

We actually applied as a classroom lecture tool.

So the idea was to give students this button to push to say, I'm confused.

And the professor would get this graph over time of how confused the students were.

We were college students at the time and we had a bunch of professors who were excited about it at MIT and elsewhere.

I'll never forget our, in our YC interview, we were pitching this and PG was getting pretty excited.

And then he turns to Professor Miller from MIT, who we had talked to and had given us the original idea for it, and says, hey, would you use this?

And Professor Miller says, nah, probably not.

But we just rolled with the punches and said, yeah, well, you know, we talked to like 20 other professors and they were all excited about it.

We went through the whole YC with this idea, built it out, hundreds of thousands of lines of code, super complicated classroom lecture tool product, had like presentation view, and people could ask questions, it was very complicated.

We actually even raised money at demo day with this idea about 600K.

Finally we deployed it in the classroom as the fall semester got started after demo day.

All students opened their laptops and they all went straight to Facebook.

So the way we discovered this is we were standing in the back of the classroom just counting laptop screens.

So we'd be looking over the shoulders of the students, you know, 123, and we discovered at the beginning of class, about 60% of students were on Facebook, and by the end, about 80% were on Facebook.

In other words, they were supposed to be using your desktop app the whole time.

That's right, and the professor at the beginning of class had been like, can everyone please get out their laptops, we're going to use this new app.

sudden, all these students are distracted by Facebook.

So we had accidentally put in an attention bomb, if you will.

Yeah, and you hadn't brought on any users during 1.

And you weren't testing during the batch?

We tried testing, but there's not that many classrooms that are in session during the summer.

Right, the school year starts in September.

So we had beta tested in a few summer computer science classes at both Stanford and Berkeley.

But There was always technical difficulties and other things that sort of prevented us from getting a real sense of what was happening.

I don't know, we didn't really get tests rolling until the fall.

Okay, and so what was the come to Jesus moment where you realized you had to change the product?

Standing in the back of a BU classroom as an anthropology class and I remember arriving at the 60% and the 80% and we went up and apologized to the professor and walked out.

And so at that point you're just like okay we got to kind of shut this down or figure out what to do with the money.

Well, we had just gotten wires for these checks, for this money, like literally a week before.

So we called back all the investors, and we were like, well, it turns out this is a terrible idea.

So what do you want us to do with the money?

And almost all of them said, well, we invested for the team, so go find something else.

Next day, we're like, well, what are we going to do?

And we realized we should have been able to see some of this digitally.

You couldn't see it in the analytics metrics at all.

So we decided, hey, let's build an analytics tool.

Let's build a better web analytics, mobile web app analytics tool to compete with Mixpanel and Google Analytics.

The idea was to give really advanced segmentation because we also wanted to understand how some computer science classrooms at MIT were using it differently than anthropology classes at BU, et cetera.

And we couldn't do that analysis in the tools we had.

So that was the idea, was to build an analytics tool.

We spent about a year building the infrastructure necessary to do the analytics.

And really, we're not succeeding in getting any customers during that time frame.

So you had, I mean, did you have a whole onboarding process?

I was going on little sales trips, I was meeting with people.

Basically what would happen is people would say, well I already have this other analytics tool installed, so like, I'm not that interested.

You definitely, at the time, you were not 10x better than Google.

We were at parity in certain dimensions and we were exceeded in other dimensions but they just weren't the dimensions that mattered apparently.

So you just kind of like blow up overnight with Hacker News.

This is just the analytics tool, which we realized in December 2012 was failed.

So we're well over a year into the analytics tool.

We now have tried 2 ideas, both have failed.

So we decided to have office hours with YC again.

We come back, walk around Little Cul-de-sac by YC with PG.

He comes to a stop and says, So just to be clear, you've spent half a million dollars and you have nothing to show for it.

And it was a good sort of come to Jesus moment, though.

And you got to pause there and then rewind all the way back to the first week of YC.

And it was in that first week that we'd been like, well, we should have analytics tools on our classroom lecture tool.

So we Googled analytics, and we found Google Analytics Mixed Panel and Kismetrics.

And we're looking at them, and we're like, we don't know which 1 of these things we should use.

They're kind of all similar, but Google Analytics is a little more marketing-y, Cosmetics is a little more revenue-y, Mixpanel's a little more product-y in terms of the sorts of insights they can give you.

And at the end of the day, they all collect the same data.

They all collect basically who is the customer and what are they doing.

So we decided to write this little tiny abstraction that could send data to all 3.

It was like 50 lines of code among the hundreds of thousands for this classroom lecture tool.

And then we decided, we'll just send data to all 3, we'll look at all the tools, and we'll just pick the 1 that we need.

So it'll give us a lot of optionality basically for free.

And then we forgot about it for like 4 months.

4 months later we clean it up a little bit more, 4 months later it cleans up a little bit more.

As I mentioned, now we're struggling, at that point we're struggling with this question of, well I already have Mixpanel installed so I don't really want to use your analytics tool.

He's like, remember that little library that sort of abstracted away the differences between these tools and let us send data to all 3?

What if we added ourselves as the fourth service that it could send data to?

And then every time someone has that objection, we hit them back with an open source library that they can use to send to us and them.

That seems like a clever growth hack, it gets us around this problem.

So we did that, cleaned it up, open sourced it, and people started replying like, oh this library's great, we'd love to use it.

A couple weeks later we'd follow up and be like, hey, we saw you're using the open source library, but all you have to do is copy paste our API key so that you can use our analytics tool.

No, like, So we started just feeling like there's some traction on this little routing library, which is maybe up to like 25 stars on GitHub or something like that, not much.

But some people are issuing pull requests and it's the first time we've ever felt sort of like pull from the market, if you will.

We weren't just pushing a boulder uphill.

We had this conversation with Paul Graham.

The next day we sit down, this is our second time with this, we're like, okay, we have 100K left in the bank, what's our final shot?

And my co-founder Ian is like, you know what, I think there's a big business behind Analytics.js, which is this routing library.

And I was like, that is literally the worst idea I've ever heard.

Like, it's 500 lines of code, you're grown a little bit, it's 500 lines of code and it's already open source.

I have no idea how you build a business around that.

And so we fought about it all day long, the 4 of us.

I went home and I was trying to figure out how to kill the idea.

So I wake half the night, finally figured it out, came in the next day, I was like, all right guys, here's what we're going to do.

We're going to build a beautiful landing page, really going to pitch the value of this Analytics.js open source library, and it'll have a signup form at the bottom so that we can get people to sort of express interest.

Put it up on Hacker News and we'll see what happens.

I was thinking like, this will just totally kill it.

So we'd build a landing page, put it up on Hacker News, and this is when we have this year and a half in the making overnight explosion.

Yeah, and that kind of segues into your whole start-up school talk, right?

And up to that point in your life, had you launched anything where the market, I mean, this is a Marc Andreessen quote, because he kind of coined the term, it pulls it out of you, right?

And it feels very much like losing control, right?

Like previously you're like building a thing and you roll it out and you're building a thing and you're pushing it out.

And all of a sudden you like put a thing out there and people start running away with it and using it in ways that you didn't necessarily expect.

And you're sort of like, what, what, what, what, it's just in beta, like stop, stop, we need to fix these other things because otherwise, it's like this feeling of losing control and almost like the market is dictating to you now what the rules of the road are and what needs to get

Yeah, would you differentiate that from overwhelming demand for 1 particular feature?

Versus like we're just gonna take this and use it however we want but there's a ton of demand there Would you separate those 2 things?

Not not really I think Those people always want more features Yeah, but the the thing that flipped was people would previously tell us they wanted a feature, but not use it.

Whereas now people were using it and they would want a second feature.

And the, it's a super important distinction.

I think a lot of founders get caught in this sort of like all the death spiral of user feedback, where they keep going and showing someone their product and asking them for feedback.

They give them some feedback about how they could make it better, but they don't use it.

And then they bring it back with those fixes and they ask if this is better, and it's just like a death spiral where it never gets anywhere.

But once someone starts using it, they'll have more requests and that just means they're going to pay you more over time.

Paul Wright, yeah, I like how you put it in the lecture where you're basically like, if you have to ask yourself, it's not product market fit.

Paul Wiltz Yeah, and this was now 6 years ago?

Walker Yeah, and since then we've had a few more secondary product market fit moments.

Walker About 2 years in, we discovered that all of our most valuable customers were sending their data to an S3 bucket.

Which is basically, they were keeping log files of the raw data, which was a little weird.

Because typically what people were using the data for was to send to an analytics tool, an email marketing tool, and a CRM, and a help desk, places where a business person is deriving value.

So we went on this sales trip to New York, myself and our first salesperson, Raph.

And We met with 5 customers that were using this S3 bucket.

We just asked them, like, oh, what are you doing with the S3 bucket?

The first customer was like, well, we have a data engineering team that's taking the data out of the bucket and converting it into CSV files, and then they're uploading it to our data warehouse, which is a Redshift cluster.

So basically they were using it as the initial input into an ETL pipeline.

Went to the next meeting, second customer is like, well, we have a data engineering team who's taking the data out of S3, converting it into CSM.

And then the third, fourth, and fifth ones all said exactly the same thing.

And that was the point at which I started becoming a conspiracy theorist.

It seemed like some pre-meeting had happened.

But they were all doing exactly the same thing, so it was really obvious.

We just built a way to load data directly from Segment into a Redshift cluster.

You really felt that it was product market fit again.

Ryan Neuhofel Yeah, it was very explosive.

So we'd grown revenue from 0 to 2 and a half million in the first year, and then we launched this Redshift connector, and the next year we went from 2 and a half to 10.

Walker That's right, it was 1 step too far for them to realize that we could do it easily.

Their mentality I think was that, oh, Segment is a way that I integrate marketing tools.

And so a data warehouse isn't a marketing tool, it's a BI tool.

Surely Segment can't integrate that, it just didn't click.

Wiltz Our go-to-market model is primarily through engineers.

We talk a lot about the way that we've built our infrastructure over time.

We obviously process a lot of data so there's a lot of interesting infrastructure problems.

I think now we're processing hundreds of thousands of user actions per second.

So there's a lot of data going through there.

That's generally interesting to the hacker news and engineering crowd.

And yeah, typically an engineer is the 1 that brings us in.

Sometimes a really technical product manager, but it's someone who's like, yeah, this is going to solve this weird rats nest data pipeline problem that I've got.

A good portion of it is open source, but most of the value that we deliver is actually by running the hosted version.

Right, because at the end of the day it's not just the developer.

You're saving the developer's time, but it's these business people that really need it.

Yeah, and frankly most of the complexity is hidden away in how you actually operate and scale a data pipeline that is processing the data.

So our JavaScript library is open source, our iOS SDK is open source, our Android SDK is open source.

You can sort of collect the data from anywhere and those collection libraries are open source, but the sort of core infrastructure pipeline is not.

Okay, and so right before you guys launched on HN, or this like small tiny micro launch or whatever it might be, Were there other avenues that you were considering pursuing?

Like in that debate in the day before, were you thinking about other stuff?

See all Y Combinator transcripts on Youtube

Peter Reinhardt on Finding Product Market Fit at Segment