娇色导航

? Hi everybody. Welcome to DEMO, the show where companies come in and they show us their latest products and services. I'm Keith Shaw. Joining me on the show today is Claire Vo, she is the Chief Product Officer at LaunchDarkly. Welcome to the show, Claire. ?

Thanks for having me. ? All right.

So quickly, what are you showing us today? And can you tell us a little bit about the name LaunchDarkly? ? Yeah.

So LaunchDarkly is a platform for software teams to release, monitor and optimize software in production. I've written a lot of code in my life, as I like to claim all of it is perfect, but that's not really true.

Some of it has bugs, and so as a software engineer, I want more control over how my code actually gets to production, how it works and when I can turn it on and when I can turn it off.

LaunchDarkly is a platform that gets high-velocity engineering teams all the tools they need to basically decouple release, upgrade software from the deploy process, and get things live. So LaunchDarkly comes from the idea of launching things but not live yet, and then turning them on when you're ready. ?

Who is this designed for. I'm imagining software developers. ?

Developers, software engineers and their leaders who really, really, really don't want their digital experiences to go down. ?

Big companies, small companies, midsized companies, everybody? ?

Everybody can use LaunchDarkly. If you're writing code that matters in production and you want it to work for your end users, LaunchDarkly is a solution for you. ?

Now I can't remember if we talked about this before the show, but would a company like CrowdStrike have benefited from something like this? ?

What I will say is that if you embrace the best practices of decoupling release from deploy, progressively rolling out, say, to 1% of the planet, as opposed to 100% of the planet, and putting the right runtime controls in, you're going to have a lot easier time than they did when something goes wrong.

?

What are the main problems that you're solving, and what would a company do if they didn't have something like this? ?

I don't know if anybody watching this has had this experience. When you release a piece of code, usually it works, and sometimes it doesn't.

And when you release that and it doesn't work, you want to contain the blast radius of that problem and turn it off as quickly as possible.

Without LaunchDarkly, you probably have to go through a whole deploy process trying to figure out how to rip that code back, do a revert, redeploy, make sure everything that isn't broken still works. With LaunchDarkly, it's just a click of the button. ?

So it's more than just being keeping your fingers crossed. ?

It is more than keeping your fingers crossed. ?

So let's go to the demo you've been promising me, a really cool demo. ?

So [for this demo], I am a software engineer at “Launch Airways.” Actually, I'm the VP of Engineering of software at Launch Airways, and as the software engineer I'm responsible for rolling out changes to our core booking site. As you can imagine, this has tons of traffic through it.

Many consumers use this to get home, get to their loved ones, and travel safely. ?

So this is a fake this is a fake website. ? Yes.

I have this idea as a software engineer that we can put a nice booking module on the home page that can let people get right into booking, as opposed to clicking this button and going into a booking flow, I can put it front and center.

So I spend some time working on that code, figuring out that feature, testing it locally, and I'm ready to go live, but I want to make sure that I have all the controls there.

I've tested everything, but we all know the only real test is test and production, and then I also have cross-functional stakeholders like product managers and executives and marketers who want to be prepared for launch, [they] don't want to have to time their stuff to when our deploy happens.

Really want more control over when this feature is released. And I see that you're a launch airways customer as well. You have a website up on the phone, you're going to watch along. So let's say it's the middle of the afternoon on a Wednesday.

We've decided that's a great time to release our flight booking module. Now, as a software engineer using LaunchDarkly, I have already done the work to put a couple lines of code around this new feature that tells LaunchDarkly do not release it until I'm ready, but we're ready.

And so if I didn't have LaunchDarkly, I'd have to go through a whole deploy process. It would be live. But instead, I can just click this button, turn on the flag, click Review. Make sure people know why I'm doing this. Turning on the booking module, save changes.

? Wow.

?

Okay, look at that. I wanted to go for the real reaction. Wow is right. Now, if anything goes wrong, customers don't like it. People start tweeting, I hate the purple of Launch Airways booking module. Well, then, no problem, I could turn that thing right off.

No C, no code written. You can see how fast it is.

So we have a globally distributed flag delivery network that can deliver any device, any web endpoint, and immediately, so you can imagine, if something goes wrong, seconds really matter, and we make changes in less than one second. ?

So it's not just turn on and off. That's kind of the crudest form of protection. What we really want to do is make sure that I can target. So let's say I have a particular segment that I think is good early access for this product.

It can be anything from a region to test users. Let's just say I want to turn it on for U.S. West. I can actually narrowly target this feature, so only folks in U.S. West, for example, will get the feature.

But your phone might not turn right, which is nice. And then it's really more than just these front-end experiences that we can do.

We know that software engineers work on more things than front-end experiences, and we know that those things have risks to them, so we can do things like guard your releases and watch them for regressions and metrics.

So let's say, instead of working on a front-end feature, I'm refactoring a core database query that's a much more complex, much more impactful change.

I can actually do the same thing that I did with that front-end feature for this back-end feature, this database feature, I can watch and monitor it for metrics like latency and errors, and actually release it guarded.

What that means is it will slowly release it over time, but watch for regressions in those particular metrics that are important to me, and in this particular release, I wasn't such a good software engineer, and you can see latency in my new database actually spiked by new queries spiked.

I don't want that going to users -- slower websites means less dollars, less flights, and so LaunchDarkly can actually automatically roll that back, and send me a nice little alert that says, hey, Claire, nice try. But your query wasn't perfect.

Let's give it another go, write the code and re-roll it out. So this is like a co-pilot for software engineers rolling out anything. And this could be a database change, it could be a front-end change.

You can track engineering metrics, business metrics, whatever matters, but it's the protection you need to make sure that the feature that you're building actually works. ?

I think that LaunchDarkly is really great for teams that are rolling out software, big changes or little changes, and want to make sure they work well.

But we're not all playing for like it's just not broken, and so we also have experimentation capabilities that allow teams to actually optimize those experiences.

As I said, you know, maybe people like the purple banner, maybe they don't we want to really test and optimize to figure out what's going to help people book flights and trust our brand, and so we can do things like test with test copy, test functionality, test changes.

And what I like about our experimentation capability is, while this is a pretty simple example of testing, headlines, testing text, because this is built for software engineers, by software engineers, you can actually do experimentation on lots of different things.

So that can be you can experiment on different back-end API structures. You can experiment on front-end experiences. You can experiment on a single page flow versus a multi-page flow. Whatever you want, if you can code it, you can run an experiment, right? ?

I like your three, little text there for each of the different experiments. ?

Well, we do have, if you want early access to it., We do have an AI, of course, I have to say AI. I'm from San Francisco. We have an AI variation builder, so if we're not feeling particularly creative, we can generate some of these variations for you.

And then, of course, what really matters, just like are things going wrong, is are things going right, and so you can start to see the results.

Higher intent copy for this particular experience, performed a lot better than being funny, even though we really like I know it's a real bummer. I've been doing A/B testing and experimentation for over 20 years, and I'm always disappointed that my favorite usually doesn't win. ?

So this is the expectation that's better, too, to have real stats than gut feeling or intuition. ?

100% I mean, how many of us have had, you know, a boss say, I just know it's gonna work, and now we're giving you the tools to really determine not does it work functionally.

We give you those tools, but does it work for the business and does it work for your users? And you know, just like code, the best test is to test in production with experiences. The best test is to really test it with users. And I love all this.

I think it's really fun. I was a LaunchDarkly customer for over four years before joining the company. Earlier this year, it is just tremendously useful tool for software engineers. It's something we use every day to release code, target them, change how they operate, experiment on them.

And then one of the other things that I was really excited about joining LaunchDarkly this year, and why I chose to come and lead the technology organization is I had a hypothesis that there were AI tailwinds on this company, and that there was a real applied use case for AI.

And so I don't know how much building you've done with these AI models or LLMs, but my experience has been that the concepts that we've been talking about today, controlling the release of code, configuring it in runtime, on fly, having fallbacks and fail safes on your models, testing things, all that matters even more if you're building customer experiences with these non-deterministic models, because quality is harder to predict, and customer experience is very, very important.

?

So how long would it take a customer to set this up? Is this a long install process? Is it just click a button and it’s done? ?

We’ve done somewhere between. It's got to be between three minutes and 14 seconds. It's Pi. So it'll apply. It'll appeal to the nerds. No, we actually revised our Quick Start experience for first time developers using LaunchDarkly.

Cody, our Senior Director of Innovation and I went on stage in May and did it live, wow. And we did it in sub-four minutes from sign up for LaunchDarkly account to that magic moment where I turn on code ?

Do you offer a free trial? ?

We have a free trial, and then we have a great developer plan. So if you're a startup out there or have a project that you just want to see what these capabilities look like, we have a plan for you.

And then, of course, we are supporting some of the largest enterprises in the world. It's been really fun not just to build this stuff for software engineers and startups.

It's really great to see this being used at global scale, with 10s of thousands of features going live every day, ?

You've got a lot of other cool features too. We don't have a lot of time to go through all of them. So where can people go for more information? Launchdarkly.com, and make sure they don't go to the fake airline site. ? Oh no, no.

Launch Airways is not real. We are a software company, not an airline. ? All right.

Claire Vo, thanks for being on the show. That's all the time we have for today's episode. Be sure to like the video, subscribe to the channel, add any thoughts you have below. Join us every week for new episodes of DEMO. I'm Keith Shaw, thanks for watching.