Sagar Batchu

So those are two learnings that we had. I think the takeaway for me there was really optimizing for tech stack internally that allows for onboarding as many customers as possible with as little support. And then the second one, making sure we identified like which features were the kind of killer features for a customer.

So those are two learnings that we had. I think the takeaway for me there was really optimizing for tech stack internally that allows for onboarding as many customers as possible with as little support. And then the second one, making sure we identified like which features were the kind of killer features for a customer.

So those are two learnings that we had. I think the takeaway for me there was really optimizing for tech stack internally that allows for onboarding as many customers as possible with as little support. And then the second one, making sure we identified like which features were the kind of killer features for a customer.

When you're going to put out a UI to your customers, like you really care about design and quality.

When you're going to put out a UI to your customers, like you really care about design and quality.

When you're going to put out a UI to your customers, like you really care about design and quality.

After that kind of experience, when we pivoted into SDKs, I think one of the things that happened was that we started getting into the hands of so many users. A lot of that feedback came naturally.

After that kind of experience, when we pivoted into SDKs, I think one of the things that happened was that we started getting into the hands of so many users. A lot of that feedback came naturally.

After that kind of experience, when we pivoted into SDKs, I think one of the things that happened was that we started getting into the hands of so many users. A lot of that feedback came naturally.

One of the things that we invested in on the own was like a kind of support process that allowed us to get unadulterated feedback from our customers directly to us and everyone developing on the product. We didn't buffer that with support or any kind of function like that. So we literally had Slack channels with every customer as they ran into questions and issues, iterated directly.

One of the things that we invested in on the own was like a kind of support process that allowed us to get unadulterated feedback from our customers directly to us and everyone developing on the product. We didn't buffer that with support or any kind of function like that. So we literally had Slack channels with every customer as they ran into questions and issues, iterated directly.

One of the things that we invested in on the own was like a kind of support process that allowed us to get unadulterated feedback from our customers directly to us and everyone developing on the product. We didn't buffer that with support or any kind of function like that. So we literally had Slack channels with every customer as they ran into questions and issues, iterated directly.

And we very much built a kind of culture around forgiveness or ask for forgiveness, not permission. And so everyone developing the product had full autonomy to make the design decisions and fly forward. really take customer feedback, implement directly, and ship again.

And we very much built a kind of culture around forgiveness or ask for forgiveness, not permission. And so everyone developing the product had full autonomy to make the design decisions and fly forward. really take customer feedback, implement directly, and ship again.

And we very much built a kind of culture around forgiveness or ask for forgiveness, not permission. And so everyone developing the product had full autonomy to make the design decisions and fly forward. really take customer feedback, implement directly, and ship again.

So we got into this really iterative process early on with a dozen or so customers in SDK product, where every week, every day really, we were doing multiple releases, having them use it. And I think what was really key was identifying customers that were willing to make it.

So we got into this really iterative process early on with a dozen or so customers in SDK product, where every week, every day really, we were doing multiple releases, having them use it. And I think what was really key was identifying customers that were willing to make it.

So we got into this really iterative process early on with a dozen or so customers in SDK product, where every week, every day really, we were doing multiple releases, having them use it. And I think what was really key was identifying customers that were willing to make it.

This only works if you have engineers at the customer who are in Slack with you, going back and forth, really diving deep into things like, How do you want an SDK to represent union types? That's like a tough type safety problem. Or how should file streaming be handled for different encodings in an SDK?

This only works if you have engineers at the customer who are in Slack with you, going back and forth, really diving deep into things like, How do you want an SDK to represent union types? That's like a tough type safety problem. Or how should file streaming be handled for different encodings in an SDK?