In this episode, Frank and Andy speak with Dan Burcaw on Entrepreneurship, Using AI to Stop Customer Churn, and Deploying Code onto Nuclear Submarines.
The following transcript is AI generated.
Hello and welcome to data driven.
The podcast where we explore the emerging fields of data science, machine learning, and artificial intelligence.
In this episode, Frank and Andy speak with Dan Burke or Dan is a serial entrepreneur who has founded four companies each on the forefront of a major technology wave, open source software, the smartphone.
Cloud computing and now machine learning.
Currently he leads Nam Eml, a company focused on helping app developers start and grow mobile subscription businesses.
If you follow Frank and or Andy on social media, you certainly have heard them bang on about their secret project.
I will drop a one word hint here foreshadowing.
Now on with the show.
Hello and welcome back to data driven.
The podcast where we explore the emerging fields of data science machine learning, an artificial intelligence, and if you like to think of data as the new oil, then you could consider us Car Talk.
Because we focus on where the rubber hits the road.
So with that as my guest on this pandemic road trip, that hasn't happened.
By my copilot here is Andy Leonard.
How you doing Andy?
Hey, I'm doing pretty good Frank how are you?
I'm doing well, I'm doing well.
I had a kind of an architecture session this morning, so that went really well.
It was an interesting conversation and I love doing those.
Those are always fun.
Yeah, so I'm proofing the next book.
Proofing is the absolute last chance to remove all of the typos I've left in.
As I've gone through the last three full edit sessions and there's still some there.
Frank, I'm convinced that the next book is going to have, you know, have a fair share of those.
What I'm really concerned about.
Is making sure that the demos work an yeah that's you know it's it's tedious and it's the LastPass so you know it's like is this over yet? Yeah, I'm sick and tired of reading this guy's writing and it's me so.
That was the hardest part.
Like when I wrote a book on Silverlight an aside from it being about Silverlight, the hardest thing wasn't so much writing, it was having to go back and re edit my own stuff and like.
You know, and I would look at it and be like man like I'm a terrible or.
That's I have said over and over again to my computer monitor who wrote this crap.
By a friend if you live.
But Fortunately for this is a second edition, so an it's one of those second editions where I kept the first 10 or 11 chapters.
I I changed from my writing language.
I wrote it like three years ago.
And I really this grew out of a series of blog posts that I wrote back in 2012. It was all in VB back then, Visual Basic. And so I wrote it that way in 2017 and for the 2nd edition I went back and updated all of that. That's really the only thing I changed was I went to C sharp.
An I kind of needed to because the rest of the book was going to be in C sharp anyway.
And so yeah, that's that's kind of how it went.
And for anybody listen, it thinks wow, Andy is smart.
He's written a book about C sharp.
He must know C sharp really, really well.
I say throughout the book I am not a C sharp developer.
I feel like I'm working my way up to being a noob, but but.
Don't you work classes?
I do wear glasses.
So you can see sharp.
They took me awhile.
Do you have your sound effects running from I?
Do were back in Zend Caster.
So for folks listening like I don't remember this being on the live stream.
If it's not, we're doing this the old fashioned way right then, and don't worry, Andy and I've been live streaming a lot, which you probably noticed, but today we have a very special guest, don't we, Andy?
Yeah yeah, Dan Burke all is awesome.
He's a co-founder and CEO and I hope I say this right, is it?
Is it nami? Nami ML Dan.
00:04:07 Dan Burcaw
Yeah nami. Like tsunami.
Ah OK, I got it right the first time NAMI AML and it's a really smart service for monetizing digital products with subscriptions.
And just he's had a whole ton of experience working in, you know, in marketing for the Oracle Marketing Cloud, working with the mobile product for that.
So pretty smart Guy joined joined Oracle back during the acquisition of Push IO and.
Push IO was a leading mobile messaging provider as well.
And he served there as a Co founder and CEO.
There's a bunch more in here about Nan, an it all kind of boils down to super smart, successful guy.
We've had a little bit of banner before we click the record button an I can attest to.
That is really enjoyable conversation.
I look forward to this show.
Thanks for being here, Dan.
00:05:05 Speaker 1
Really happy to be here.
00:05:05 Speaker 1
Really happy to be here.
00:05:06 Speaker 1
Thanks for having me.
Awesome, so you're a serial entrepreneur and you founded a bunch of companies.
Um, but my favorite part of the bio I read on you was that.
You wrote software that ended up on a nuclear submarine.
00:05:23 Speaker 1
Yeah, that's right.
00:05:26 Speaker 1
It's it's hard.
That that totally away I was like what?
00:05:29 Speaker 1
It it's it's hard to even tell that story sometimes because it's so unbelievable.
00:05:35 Speaker 1
I 17 years old at the time.
00:05:38 Speaker 1
The company that I cofounded was building a flavor of Linux.
00:05:46 Speaker 1
A flavor of Linux that was designed to run on Apple Macintosh hardware.
00:05:52 Speaker 1
And at the time.
00:05:54 Speaker 1
Then the the reason for that was that Apple was using the power PC chip power PC chip in that moment of time. You know, we're kind of talking in the late 90s. Early 2000s had fantastic price per performance per Watt, which is a metric that a lot of folks in the kind of high performance computing world look at when they're trying to figure out.
00:06:18 Speaker 1
How to build these kind of supercomputer clusters?
00:06:21 Speaker 1
And so it just happened at that moment in time, the Mac would had had the best price performance per Watt because of the chips that they.
00:06:29 Speaker 1
We're using and so we we ended up doing a deal with Lockheed Martin and the US Navy to build a cluster of Macs running Linux.
00:06:45 Speaker 1
That were deployed across the US Navy nuclear sub fleet for the purpose of doing sonar image processing, yeah.
00:06:55 Speaker 1
The the the software that I wrote was related to.
00:07:00 Speaker 1
You know how folks on the boat would have to manage these units if there was issues, how would you know?
00:07:07 Speaker 1
Kind of the maintainability repair ability was a big issue when you're actually out at sea and trying to have this stuff run in kind of a mission critical fashion so.
00:07:17 Speaker 1
We ended up.
00:07:17 Speaker 1
I mean it was this was such a crazy project because the hardware was modified hardware.
00:07:22 Speaker 1
It wasn't off the shelf Apple hardware, it was Apple Hardware and then we did a bunch of things to it and then it was Linux and then it was some custom software that made the whole thing operate an.
00:07:35 Speaker 1
So it's it was.
00:07:37 Speaker 1
It was a nutty project, an I'm.
00:07:40 Speaker 1
Looking back on it now, I'm surprised that it had ever shipped quite frankly.
Spoken like a true engineer, right?
You're always you always look at your flaws and like Oh my God, that's actually running.
So so you you where did you go after that?
'cause it says you know you're a serial entrepreneur and so how did you get into?
I don't want to steal.
Kind of our pre canned questions Thunder but.
Tell me how did you get into A&ML? Or were you doing ML on those on those retrofitted Max?
00:08:18 Speaker 1
No, we weren't.
00:08:19 Speaker 1
We weren't, but but you know, I think that part of the hype that world of high performance computing where a lot of our customers were, you know, national labs or defense oriented things.
00:08:30 Speaker 1
I mean, part of the appeal of what we were offering in that period of time was that they were running algorithms an doing some of this stuff.
00:08:39 Speaker 1
You know, obviously ahead ahead ahead of their time and they need it.
00:08:43 Speaker 1
There wasn't the cloud computing yet, so they were literally just trying to assemble the biggest.
00:08:49 Speaker 1
Supercomputers using off the shelf hardware that they possibly could so we weren't writing the algorithms.
00:08:55 Speaker 1
We were more enabling these algorithms to be run, but I would say the Fast forward is that in terms of my career, is that working on that led to?
00:09:09 Speaker 1
Being involved in sort of the mobile ecosystem from the launch of the App Store and the iPhone back in 2000, seven 2008 and in a way it was very very similar to what we did with the submarines. Because you were dealing with constrained hard.
00:09:25 Speaker 1
Where you always had to care about performance and battery life and battery life, less so on the Subs.
00:09:31 Speaker 1
But some of the same sort of constraints where you're trying to get the best performance you can out of these things and operating in that mobile landscape and building apps for some of the largest consumer brands.
00:09:46 Speaker 1
And then you guys mentioned in the in the intro about push IO.
00:09:49 Speaker 1
This mobile messaging company that we.
00:09:51 Speaker 1
Built, we ended up at Oracle building this mobile marketing engine is part of the Oracle Marketing Cloud an and one of the things that we saw there is that now.
00:10:03 Speaker 1
Fast forward to kind of more modern times and there's such a prevalent use out there of.
00:10:11 Speaker 1
Technology like email, you know email marketing systems and push notifications in the world of mobile in order to tackle kind of a fundamental problem that exists with some of these products, which is the user.
00:10:28 Speaker 1
Download your app, let's say, and they use it and then and then they churn, and then they abandon an and you as a publisher of a product like this, is one of the battles that you're trying to fight is how do I get them back into the experience and so are sort of observation is we were.
00:10:48 Speaker 1
You know, done our tenure there and we're looking to do something next.
00:10:52 Speaker 1
And new was a couple of things.
00:10:55 Speaker 1
The first thing we saw was that with the iPhone 10, I think it was.
00:11:01 Speaker 1
00:11:02 Speaker 1
00:11:03 Speaker 1
And that was using algorithms running on the device, so the benefit was you could unlock the phone very, very fast, but also it had some privacy characteristics where Apple doesn't need your face and the kind of the point cloud representation of your face to be up on their servers somewhere.
00:11:21 Speaker 1
That was really intriguing to us.
00:11:23 Speaker 1
The other thing was that we saw that the app economy, so to speak, was in transition from kind of the early days of where it was paid downloads.
00:11:33 Speaker 1
Transitioning to kind of in app purchases, which the game ecosystem has really been been focused on to trying to create more durable, sustainable revenue models through subscription.
00:11:35 Speaker 1
Which is right?
00:11:46 Speaker 1
And so how we sort of arrived at focusing a lot on data at NAMI.
00:11:53 Speaker 1
Is that it?
00:11:54 Speaker 1
It seemed to us like there was.
00:11:56 Speaker 1
If we we would, we really were excited about an idea that if we could.
00:12:02 Speaker 1
Help the guy.
00:12:04 Speaker 1
00:12:04 Speaker 1
App publishers a mechanism to send way fewer push notifications.
00:12:10 Speaker 1
An email messages because they had a technology stack that could allow them to detect in the experience, right directly on the device that somebody was showing signs of churn, or that somebody was showing some.
00:12:25 Speaker 1
Early intent that they might be a a candidate to be a subscriber, and so just that idea that maybe there's a way that we could be part of cutting down the messaging load by making the actual experiences smarter and more intelligent about what users are doing was where we.
00:12:42 Speaker 1
What sorts of signals?
That you can collect given specially with Apple's kind of enhanced privacy policies that they've been been doing.
What sorts of signals kind of indicate churn?
00:13:00 Speaker 1
So you know it is.
00:13:01 Speaker 1
It's a great question.
00:13:02 Speaker 1
When we started out we were thing.
00:13:04 Speaker 1
Gain, we're going to collect all this crazy stuff.
00:13:07 Speaker 1
I mean, we were even thinking at one point in the early prototyping that you know, maybe maybe, what carrier the user is on is some signal.
00:13:16 Speaker 1
Maybe the device form factor, whether it's the really expensive version of the phone or the lower you know there's all these things that we were thinking about, but.
00:13:24 Speaker 1
00:13:27 Speaker 1
And we're not my cofounder and I are not experts in this field, so one of the things that we did was we recruited our CTO who has a PhD in applied math and had been building data science animal models, kind of in production at, you know, in the real world.
00:13:46 Speaker 1
Applications of places like the Los Angeles Times and Tribune Publishing and one of the first things he told us when he came in was guys like, wait, you're trying to?
00:13:55 Speaker 1
You don't need all these days.
00:13:56 Speaker 1
Points a lot of what you're trying to collect just isn't isn't going to move the needle, and So what.
00:13:57 Speaker 1
00:14:03 Speaker 1
It really gets to both on the so we look at, you know from subscriptions we're looking at.
00:14:07 Speaker 1
Kind of two things.
00:14:08 Speaker 1
One is what are signals that show that somebody might be have a propensity to purchase.
00:14:17 Speaker 1
And then, secondarily, that early turn detection, or kind of likelihood to churn.
00:14:23 Speaker 1
And it turns out it's it's pretty simple on some level, because it's really about the behavioral signals around engagement.
00:14:34 Speaker 1
So are they using the app or are they using the app a lot?
00:14:38 Speaker 1
Or they did they used to use it a lot and now they're not using it as much so those are kind of the key?
Signals, so you're not popping up little boxes and saying, do you want to keep using the app?
Check yes or no.
00:14:54 Speaker 1
No, I mean it's it's.
00:14:55 Speaker 1
It's funny, you know.
00:14:57 Speaker 1
I have a friend that has that has a company that that powers some of that around the ratings right?
00:15:05 Speaker 1
Do you want?
00:15:05 Speaker 1
To make this.
00:15:06 Speaker 1
Yep, and you know they have a really fascinating take on it, which is that.
00:15:11 Speaker 1
Because whenever I see one of those, I hit no, I don't.
00:15:14 Speaker 1
I you know, I just like want to dismiss it an.
00:15:18 Speaker 1
Yeah, he's got a strong viewpoint that by by asking a user a binary question it provides them better data for what they're trying to do around kind of customer sentiment an so I just thought I was fascinated by that because whenever I see one of those ratings popped up so I just wanted to like.
00:15:41 Speaker 1
I want to say no, even if I like the experience on some level, I have a visceral reaction that's just like I leave me.
Well, I wonder if that's
Well, it's always when you're sorry, ID.
That's OK, go ahead.
It's always when you're trying to do something or the kids are screaming like do you want to write this?
It's always when you're.
Like no, I want to use this stupid app like even if I like it.
But what I find myself doing and I've as I'll say not now like like remind me