Who? Where? Why? Using data to understand your audience


Andreas Constantinou speaking at DevRelCon London 2020

How can you use data to better understand your developer audience and inform your developer relations strategy? In this session from DevRelCon London 2019, Andreas Constantinou, CEO and Founder of Slashdata, walks through his company’s approach.



Andreas: There was a question earlier about how many developers are there in the world. So, it’s this question and other questions like this I’m here to answer. Unlike many of you, we don’t build developer communities, we measure developer communities. And we have lots of data on developer communities. I’m gonna share quite a lot of that.

So firstly, Slashdata, we measure the entire developer population. I’ll tell you how in a bit. So, we measure how many developers there are, what profiles they have, segments, and so on. We measure what tools they use, why they adopt these tools and why not, how they respond to marketing activities from hackathons to marketing support, and where they’re going next. And I’m gonna share quite a bit of that. The way we do that is we run two global surveys a year. We’ve been doing that for quite a while, a few years now. We’ve run 18 surveys globally and we on board about 40,000 developers annually to these surveys, and we measure developers from across the spectrum from mobile to machine learning.

Recently, we started looking at things like app extensions for third-party ecosystems like CRM and ERP and voice platforms and so on. And I have lots of data points on all of these. I’m very proud of our customers as well. Nine customers in the top 20 global brands. And so, firstly, the question asked by the Google guy earlier today, how many developers and how big the developer population is? We look at a range of sources.

So, this is kind of first starting with a methodology. So, we talked about just over or just under, sorry 20, no, 19 million developers as of the beginning of 2019, it’s of course slightly bigger now. And we look at several sources. So, one is areas of developer activity like Stack Overflow accounts and GitHub accounts and npm accounts and areas, of course, sources for labor statistics, such as US Labor Statistics and EU statistics, and on top of that we overlay our own data.

So, by looking at all the five different areas of sources, we come up to an average and we build on that. So, we have seen, of course, that developer population grow. Nothing big there or nothing new, but it’s been growing about 20 percent annually and it will continue to grow, slowing down a bit.

We forecast the growth to go to about, what is it? 45 million in 2030, so about twice the size, and this includes all sorts of developers, pro, and hobbyists, and machine learning, and AR/VR, and web and mobile and I’ll talk more about this, but this is the kind of top level picture. Now where our developers activate, of course you cannot talk about a web developer. A mobile developer, they actually do all of these things at the same time. Based on our data, we see that, on average, they are active in 2.7 of these domains out of 10 domains. They are active in 2.7 of these domains. What’s interesting here is, of course, Web Apps has been around for a long time.

Cloud development desktop hasn’t been around for a long time, but desktop is close at growing the slowest. What’s growing the fastest is, as you’d expect, machine learning. And there’s ML and AI has more than twice the number of developers than AR/VR and to me it’s interesting because they’ve both been hyped for roughly the same amount of time, but AR/VR hasn’t seen the commercial traction in terms of end-user devices that are cheap enough to generate the developer traction that you’d expect. So, ML has been growing much faster and we’ll see later in the breakdown of how many pros versus hobbyists there are, which is also another interesting way to see things, and I’ll also talk about games in a bit.

Now, in terms of languages, languages are important because you need to know what languages support for your toolkit or what language binding to create and so on. JavaScript has been there at the top just under 12 million developers total. The next two languages, Java and Python, are interesting because they’ve all been driven by technology paradigms. So, Java has been driven by Android, not so much the back-end developer, because, back-end development, because they has been superseded by the technologies and Python has been driven by machine learning.

And then visual development tools, which is kind of akin to the local paradigm has been there for a long time, roughly at that position. Again you see 2.7 languages spoken in average at the same time. That number has been decreasing. Interestingly, we see some more stability and JavaScript is at the top and I’ll tell you a bit more about where it’s trending. So, this is the ranking so forget like absolute numbers, but this is the ranking of languages you will have seen other rankings from like CHILL, B, or Red Monk, or Stack Overflow, or GitHub, Octave or something they call it.

The difference here is that we’re not just looking at web searches or, you know, GitHub accounts and so on and so forth are mentions, but we’re looking at the actual use, stated use, of developers of these languages over time. So it’s, in my mind, the most reliable metric on actual language use. So, JavaScript has been there for a while, for two and a half years. This is the breadth of the data point. Everything else is more or less stable in terms of ranking.

So what’s interesting is Kotlin has been increasing very fast, which is also now Android’s. It’s a first-class citizen for Android so it has a similar kind of relationship to Java as Swift has to Objective-C and as we see here Swift is clearly winning in terms of popularity, like number of developers speaking the language versus Objective-C and interestingly for me Lua, which was very popular in games and other applications, has gone down significantly.

That’s it for languages. Now, I want to look at how you segment developers because developers, of course, you know, there are 19 million of them, but they’re very very diverse as is the human population and the people in this room. So, how do you segment developers? And the more sophisticated your developer relations or marketing program is, the more sophisticated you will be in segmenting developers.

So, we run this survey not on developers, but on developer program leaders. This is the people who attend our future developer summit who are the director and VP level so kind of the decision makers and developer programs. This is a recent study from about 45 developer program leaders. We asked them, how do you segment your developer population? And of course only a small, very small part, 11 percent said they don’t, but how do the others do that?

So, by target industry, by Pro versus hobbyist, target audience stuff you would expect. Now, there is a long tail here. Some of them I would agree with in using. Some of them I would not and I’ll tell you which ones I would not agree with. So, there’s a lot of technology and commercial criteria in segmenting developers and I’ll tell you why we think this is not the right way to segment them. By the way, you get access to the slides. Don’t worry about taking pictures.

Technology has a habit of becoming obsolete. This is mobile technologies which no longer exist. Also, a developer using one technology as a professional will use a different technology as a hobbyist so, you know, how do you segment them? Which bucket do they fit? Same thing applies to languages. You would use one language for one project and another program language for another project. Also, again, for areas of involvement, if you look at are they web developer a game developer, well they’re all of them at the same time because this is only a very small percentage of developers who speak, or sorry, who work in these domains exclusively.

So, they work in multiple domains at the same time. You cannot use that as a means of developer segmentation. Of course, you have a choice. We took the choice of not taking any single one and we have, let’s say, we’re lucky to have the data on a very large number of developers every six months and data across all of these metrics and we chose these metrics here. So, demographics, involvement area, industry, and so on and so forth, decision making, which is very important. You can see the deltas if you factor developer decision-making in and then we did not include revenues or technology or language choices as I mentioned and I mentioned why and we learn clustering analysis and we developed a set of personas.

Now, these are not universal personas. If you run your own analysis or if we run this analysis on your own developers it will look different, but we ran this on the entire developer population. We came up with six segments and colored these into six personas and this is the ones we came up with. And yet, this is unsupervised so we didn’t tell it where to look. We ran the clustering analysis. This is what it found.

Young learners are those who are relatively inexperienced. They are mostly motivated by fun and sense of learning. If they work in a company, they work in a very small company. The young professionals are more motivated by money and, you know, business growth. Very little software experience.

The middle standards is where you start to see developers getting more experienced. They don’t have any experience in emerging sectors, mostly mobile or desktop and they will be motivated by money and business goals. Then you’d have the emerging extenders who are, and we call them, emerging because they are interested, usually, in AR/VR or IOT or machine learning. Not necessarily data science because there’s a bit of a gap there and they will work in teams up to 50 people. They do have commercial goals.

And then, where you get to the very the top pros, these guys are 35 and above mostly, they have titles like CIO, CTO, the DevOps specialist, architect, and so on. And they have the largest influence in tooling decisions and then finally have not necessarily hobbyists, but the loners like those who work on their own or work in very small teams up to five people with just one or two developers at the most. Again, this is our model by taking the entire developer population with all the profiles and coming up with six segments if we run it on a different type of the developer population or a sub-segment. Let’s say AR/VR developers only will come up with a different result, but it’s a way of, I think, reflecting on using clustering analysis on the developer population.

Now, a trend and insight into trends and you know there’s tons of stuff I could speak about here, but I chose some trends regarding cloud, which is probably one of the most interesting areas. Now, we look at into cloud, but this is firstly the broad picture. So, this looks at all the 10 development areas that we track. The first column reflects, I think, almost exactly, but not exactly, the population sizing that we showed before, but then the interesting thing is if you break down into professionals versus hobbyists.

So, you see a few things here. Firstly, back-end and web apps is where you have the most professional developers. In hobbyists, there are an exception, which is interesting, followed by, as for third-party ecosystems, things like what I say P is for ARP, what Salesforce is for CRM, what slack is for messaging systems, what Alexa and Siri are for voice platforms. So, this is mostly pro developer based.

On the contrary, you have those which are mostly hobbyist populated things like games. Now, why so many hobbyists developer in games? Because games is dominated by the Triple-A Studios then make most of the money or if not all of the money and everyone else is, you know, making a go at it, but only as a hobbyist not having any serious revenue intentions or hopes.

AR/VR is interesting. So, AR/VR is one of the most hyped technologies still and you see new devices coming out from Facebook and others, but still only a very small, sorry, very small percentage of professionals and actually two years ago when I was doing a similar presentation, it was at 20 and 80 percent and it has just moved slightly up to 26 and 74 and this, in my mind, is the mobile paradigm, which didn’t take place in AR/VR, which is that immobile we have now, you know, two billion devices plus, actually four billion devices, two billion Android, but the same thing did not happen for AR/VR. We didn’t go into this massive availability of AR/VR devices. They’re still computationally intensive, they’re still expensive, they’re not affordable, developers professionally will not be interested until there is an addressable market of users to target. And we’ve seen that again and again and this has not yet happened for AR/VR, but it has happened for machine learning to a much larger degree. An IoT, we look at it very differently. So, if it’s Industrial IOT, many more pros as opposed to consumer electronics things like, you know, smart speakers at home, relatively few pros compared to hobbyists.

Now, let’s look into specifically cloud. So for each type of development, we look at the types and which solutions they use, which technologies they use. So, this is for cloud containers is unapologetically at the top and it was number two until recently so it surpassed DBaaS as the top cloud technology for cloud developers.

All the sample size that you see here are fairly substantial so this is four thousand developers, and we have data every six months, followed by DBaaS and PaaS and then you see the cloud orchestration tools. Things like Kubernetes is back here, but it will move up as given now that containers is number one.

Now, let’s look specifically at Kubernetes. I had one more slide because it’s an important technology. I am not showing all the data here, in fact, I was struggling to get the slide from the analyst thing because they are very precious. I’m very cautious on what’s behind the paywall and what’s not. So, Kubernetes is number one here. It’s the only, the interesting thing here with this slide is that it’s the only technology area in cloud where Amazon is not number one and, of course, this is, in a sense, vendor agnostic.

Now, you know, informally it was founded to avoid an Amazon monopoly, but now, of course, Amazon is building on top of Kubernetes and creating its own kind of variant on top, but it’s the only two things. It’s the only technology area in cloud where Amazon is not number one and that there is such a big difference between number one and number two. And, of course, you also see very very few 10% of developers are not aware of what Kubernetes is. This is cloud developer specifically.

Let’s look at some other stuff. Globalization and how you, or if you should, globalize your program and that means localize it globally. So, firstly, the survey that we run reaches all four or five major continents. So, Americas, Europe, Africa, Asia. We have quotas to do that when you reach different communities. We reach over 70 different communities and channels from paid advertising to partners to actually help us get sample from all of these regions.

We also run, again, surveys to developer program leaders asking, you know, developers are out there. They have very different needs and wants based on their region. Vastly different needs and wants, but do you actually change your program? Do you change your outreach? Do you change your support activities? The money you invest? And so on by region. Now, it turns out that half do and half don’t. Now, that half that do are in very different stages so 10% only localized content, which is you know the the easiest thing to do. Another 10% has a centralized team, but they differentiate services by region and another 19% have regional teams in place, of course, this is the most expensive. You need a substantial size developer program, like in the 20 plus size, to have regional teams in place, but for me it’s striking that half of developer programs don’t have local strategies. Final bit for today is the marketing and relations side of developer programs.

We measure that quite a lot so let me tell you how we measure that. We measure 20 different attributes, from documentation and sample code, to marketing support, online events, video streaming, tutorials, meetups, events, online documentation, Twitter, and so on. So, we mention all these activities and what we measure is we ask developers, what’s important to you? For the programs or technology brands that you use, how well are they doing in what’s important to you? And how satisfied are you?

And here is some of the results from that series of studies. So, firstly, what’s important? So, for me, this is a key slide because it says if you have X budget, if you have 1 million, if you have 10 million, you have, you know, quarter of a million, whatever it is. How do you split it? And this is what developers say time after time, what is important to them? And number one is going to stay there forever, it’s documentation and sample code. This is where you start. You cannot not have that. This is kind of the bread and butter of a developer portal when it starts integrations with development tools. We’ll see a bit about that in a bit. Tutorials and how-to videos. This was at number four and now it’s number three and before at number three we had Stack Overflow and other answers in public forums, but it seems tutorials and how-to videos are getting better and better. I’ll show you some rankings as well in that. And then the extended course, training course, is tech support and official forums.

Now, let’s look at how the developer programs, at least the big brands, are comparing. So, here we look at something like 20 plus brands and we look at both east and west. We look at technology in terms of software and hardware. We have a varied mix. We measure three things. This is developer satisfaction and this is a composite index. This is for the support activities you care about like documentation or hackathons or events or meetups or whatever it is. How well are the programs you are using performing on those activities? So, it’s satisfaction for the things developers care about. It’s not just a raw NPS.

And then, how often are you using it? Like, are using it daily? Weekly? Monthly? And so on. This is the engagement. And the bubble size is the adoption. How many developers are using Internet on these programs? And you see these bubbles there move over time so this is as of six months ago data. Microsoft used to be at the top, now it’s superseded by Mozilla and Google. Mozilla, despite being sort of web-only, it has been doing extremely well in terms of developer satisfaction because they are the role model, among other things, for documentation and sample code. They’ve been there for a while. Google, of course, very large organization investing heavily in developer relations, no surprise.

Specialized tools for game developers like Unity, you have clusters here like big boys for Facebook and Amazon. However, who are not doing as well, and they have been in this middle cluster for a long time, here you have some hardware companies, and you have some traditional enterprise companies like IBM, Twitter, of course, used to be much higher, but they rebooted their developer program. They started from scratch pretty much.

And then you have some interesting companies here like Red Hat, who’s now IBM, Digital Ocean has been doing extremely well and has moved up the ranks here. Another insight, I cannot show you all the data points here, but I wanted to show you how the programs differ.

So, here we have just five important of these attributes like dev tools. Dev tools, you see, Microsoft and another company here doing very well and everyone being far behind whereas information via newsletters, blogs, and social media, there’s not much differentiation. If you ask developers how important is information via, you know, newsletters, blogs, they won’t care as much. It’s one of the lowest ranking, but everyone has it because it’s considered like a must-have, but you know having best-in-class integration with developer tools is much more important and, as we saw before, it’s one of the core attributes. Documentation and sample code. It takes a lot to get documentation right.

There’s many ways, many technologies, many kind of approaches to how you do documentation, all the way from classic like DRI documentation to more peer support documentation and certification is increasingly important as we saw. Again, big gap in how the programs are delivering it so there’s best-in-class leaders, but there’s also, you know, very perceptible changes and differences in how the developer programs are delivering.

I think that’s done for today. That’s me for today. If you want any of these data points just email me. I’ll send the deck over. We’re also creating a deck for a few of these stats so you can, you know, embed in your presentations if you’re doing a business plan or a marketing plan or just want to present at a conference. That’s it for now, thank you.

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.