This is a screenshot of the (fateful) blog post that launched Heroku in 2007. This is the entire thing. It’s just 4 paragraphs, two of which are about logistics.
And yet within two weeks, they had 1,000 signups. 6 months later, they had 10,000. There was no marketing team, no comprehensive launch strategy, no HackerNews[1] voting ring, and no content-rich website full of sample use cases. All there was, was a new product that solved a specific problem (deploying apps) for a specific set of people (Ruby developers), a problem that had not as of yet been solved really at all.
Want to tap into a bit of nostalgia? Check out some old Heroku blog posts. Start with this one about failed Mongrel starts that begins with a picture of some empty Monster cans. This post shares some excitement about a Japanese app that uses Heroku. Another post is a 3 sentence commentary on someone else’s quote. Some posts, like this one about launching a Google Group, are literally one line! Think any of these would get past a modern marketing department today? And yet, by 2010, Heroku powered more than 100K apps and got acquired for <$200M by Salesforce[2] with only ~30 employees.
Nobody would dream of launching and building a developer tool like this today. Growing adoption of a Heroku-type product in 2024 is a mad dash of content production, social media thought leader posting, launch weeks, LinkedIn comment chains, races to the HackerNews front page, SEO, drip campaigns… and don’t even get me started about “building in public.” Gone are the days of whimsical, few-sentence announcements and blog posts that start with “behold.” It seems like to get developers to use your new widget today, you need to do more – orders of magnitude more – than in 2007.
So what happened?
A big pond full of both small fish and big fish
What happened, I think, since 2007 is that the second part of the Heroku equation changed. Heroku was, to quote myself from a few minutes ago, “a product that solved a specific problem (deploying apps) for a specific set of people (Ruby developers), a problem that had not as of yet been solved really at all.” Today, most of these problems have been solved in some way or shape[3].
There are tens of thousands of developer tools out there, many (most?) completely free and open source. Any conceivable issue you may have, from something as fundamental as “how do I stream data in real time?” (Kafka) to “how do I make cute little arrows show up in my Terminal” (Oh My Zsh) – well, there’s a tool for that. Actually, there are probably more like 4-5 tools for that. You need to do more than Heroku today because in 2007 there were 0 other Herokus and today there are about 1,000,000[4] other Herokus.
If you don’t believe me, I brought some data along. 10 years ago, in 2014, there were just ~6 public companies like Heroku: software businesses with technical buyers like developers, data teams, SREs, etc. And one of them was Oracle, which went public back in 1986, well before your author was even born. 10 years later there are more than 30 of these types of companies, from databases (MongoDB, Snowflake, Couchbase, Elastic) to DevOps (Splunk, Datadog, JFrog) to plain old infrastructure and developer tools (Hashicorp, Atlassian, Twilio, Mulesoft, etc).
Though many developers might not always think of public markets as an important input to what happens with their tools, they very much are! Because if you started a developer tools business in 2007, there was a very, very small number of successful companies you could look to for inspiration, or model as an eventual financial outcome. Your potential investors (and employees) are judging your chances of success by how many other people have had a similar success, and there just weren’t that many. Today, everyone knows that developer tools are a legitimate market category that can drive shareholder value. And by the way, all of these new public developer tools companies are full of, you guessed it, developers who will eventually go and build their own developer tools.
Speaking of developers, there are a lot more of them around to create new tools than there used to be. In 2021 more than 100K students graduated with an undergraduate degree in something Computer Science related (or at least as close as public statistics can get us to that definition). That is 2x (!) the number that graduated with those types of degrees 10 years prior, and it has been steadily climbing every year.
It is true that more developers means that the total addressable market for Herokus has increased by quite a bit: software is more important, there are more developers, and there are more buyers. Digital transformation! But even if the total spend on developer tools has doubled, or tripled, or even quadrupled since Heroku was created, it still doesn’t come close to matching how many more tools there are. Why are there 65,000 GenAI projects on GitHub? I don’t know.
I snooped around on YCombinator’s website and scraped a bit of data about their historical batch composition. The percentage of YC batches made up of technical tools has grown a good deal since 2014, from around 10% to around 18-20%, and then up to 30% last year (although I attribute most of that to the AI craze). In other words, there aren’t just more developer tools than there used to be; they’re arguably getting funded at higher rates than other categories.
And then there’s the simplest statistic I can give you. Pitchbook, which readers in venture capital will already know of as the painfully expensive but ruthlessly reliable[5] investment data provider for private markets, tracks around 1,950 Seed or Series A rounds over the past decade for startups in the U.S or Europe targeting a technical audience. So if you want to put a number on just how many more Herokus there are out there, this is (the low end of) it.
However you measure it, there are simply orders of magnitude more technical tools out there than there used to be.
A random walk down Market Street
Perhaps you find the above data compelling and perhaps you don’t. So let’s try a different angle. Imagine you’re a developer starting a new app from scratch, or maybe you’re an engineering manager evaluating new tools for your team at a larger company.
What do you use for observability? Well, you’ve got Datadog, New Relic, AppDynamics, Dynatrace, Chronosphere, and Honeycomb, not to mention options from the major cloud providers, open source like Prometheus, and about 100 startups. Need log search? Well for that you’ve got Elastic, Splunk, Sumo Logic, Solarwinds… the list goes on. But hey, you already know this because you see their ads everywhere.
What do you use for a database? You can start with Supabase, Neon, Electric, Prisma, Aiven, RDS, Azure SQL, Google Cloud SQL, Heroku Postgres (hey), DigitalOcean Postgres…and that’s just for Postgres! There are databases with ORMs, databases with branching, databases that are free, databases with horizontal scaling, and my personal favorite, databases with built-in admin panel UIs. And if you’re an airline, you’ve always got Oracle.
Maybe you want to build some generative AI into your product. Remember when OpenAI was the only GenAI model provider in town? Now you’ve got Meta, Mistral, Anthropic, Databricks, Gemini, Stability, Mosaic, and Xai. You can deploy your models on Cohere, Together, Fireworks, or AWS Bedrock. And you can store your embeddings in Pinecone, Chroma, LanceDB, Zilliz, or Weaviate, which you might not even need because you can also use Postgres, Redis, Clickhouse, Elastic, or Cassandra.
Tool sprawl is not just for major categories like observability or databases. When I was working at Retool, we waded into a (relatively) new category we were calling “internal tools.” There were tons of existing options for admin panels, like the built-in admin panel for Django, ForestAdmin, Internal.io, Unqork, and Appian. But we felt we were differentiated (which I think we were). And then within a year of us having a modicum of success, the clones started to pop up: Appsmith, Superblocks, Dynaboard, Airplane, Budibase, Basedash, and (somehow) more, all for a market that hadn’t even had a proven exit yet. All founded by developers, all funded by VCs, all believing they had a unique angle.
So put yourself in a developer’s shoes (or perhaps you already are in them). In 2007, there weren’t that many tools out there. So when you saw a new one that solved a problem for you, you jumped at it! When the new tool released a new feature, or fixed a bug that had been bothering you, you were very interested! All they needed to do was just say that they had done these things. They didn’t need to tell you a whole story about how they at Initech have been about user trust since day one and they’re excited to introduce their new evolution of the V7 product and they think you’ll love it.
In 2024 though, there are so many tools out there that you’re drowning in them. Your entire day is using and evaluating and discussing and configuring and deleting and reinstalling and migrating tools. So when you see a new one…well, maybe you’re just a little less excited.
This of course doesn’t mean that it’s not worth starting a developer tools company, or that you’re doomed to drown in a sea of lookalikes. Quite the contrary. Great, new, enduring developer tools are created every month – just think about how much software in your team’s budget comes from companies that were created <5 years ago. I’ve spent my entire career working at these kinds of companies, and things are even more exciting today than they used to be.
What it does mean, though, is that getting your developer tool to market today requires very different tactics and philosophies; and that the old modicum of “if you build it, they will try it” (or something like that) is no longer true. Good product and engineering, to some degree, are table stakes. Everything else – specifically go-to-market – has never been more important. Founders who can put their creative energy towards sales and marketing, instead of treating them like a necessary evil, are going to have the best chance of standing out from the pack.
I will close with this. That initial Heroku blog post from Adam had a somewhat profound line at the end of the first paragraph:
And if your product offers something of real value, awareness in your target market will grow steadily and strongly over time.
If you want to know what changed since 2007, it’s this. Brilliant, product-minded developers start new companies every day that offer something of real value. Awareness in their target markets does not grow though, because about 5-6 other brilliant, product-minded developers also started new companies around that time, or a few years earlier, or later, that offered something of real value too.
There sure are a lot of developer tools out there. So what are you doing to stand out?
(1) Which, by the way, was only created 8 months earlier, in February 2007. ↩︎
(2) Whether you consider this a good or bad outcome in hindsight, it is nonetheless reflective of growth that any founder today would happily sell their soul for. ↩︎
(3) And yet most developers would still tell you that everything they do seems highly inefficient, which is a topic for another time. ↩︎
(4) For those curious, there were more than 200M public repositories on GitHub at the time of writing. ↩︎
(5) When I was a VC at a small firm back in the day, Pitchbook was known to cold call VC firms to confirm individual deals and deal sizes. ↩︎