Going v1.0: The Backstory and Future of Our Ad Network

A few weeks ago, we released the v1.0 version of our ad server. This release coincided with a huge upgrade to our contextual targeting, machine learning model. We wanted to take the time to share our story of how we got here, the details of both the software and movement we're building, and what's coming next for EthicalAds.

Our story

A few years back, we were building Read the Docs, a documentation hosting platform, but there wasn't any monetization. It was a service with millions of monthly visitors that thousands of projects relied on for their docs but run by folks with day jobs who wore pagers for their side project. It wasn't sustainable.

Today, we now have ~130 publishers and we gross just under $60k per month in ad revenue of which 70% goes to publishers

The obvious way to monetize was ads. However, we cared about the privacy of our visitors and we didn't want ad companies to track our users around the web. We went to a few ad networks and asked if we could proxy ad traffic or even just run ads without cookies. They laughed at us and told us to take it or leave it. So we built our own ad network including writing our ad server and client from scratch and did all our own sales and marketing.

Fast forward to July 2020, when our friends over at CodeFund shut down, we launched our ad network. Expanding our ads beyond Read the Docs had been on the roadmap but CodeFund shutting down greatly accelerated our schedule while also dropping a number of sites interested in privacy-preserving ads into our lap.

Today, we now have ~130 publishers and we gross just under $60k per month in ad revenue of which 70% goes to publishers. Most of our publishers are small sites or open source projects so to send them ~$40k/mo feels great.

Building a content-targeted ad network with machine learning

Firstly, we only show developer-related ads on developer sites. No ads about a product whose site you visited last week and nothing off topic. All our advertisers sell SaaS products for developers, cloud infrastructure, they're hiring devs, or something like that.

All our publishers (that's advertising lingo for sites showing ads) are hand approved. We make sure each site is the right fit for our advertisers and protect their brands.

"EthicalAds' targeting provides a better ad experience for our community members."

— Ido Shamun, Co-founder & CTO of daily.dev

Beyond the basics, we built a crawler that crawls the sites on our network to help target our ads. Using an ML classifier (built with SpaCy), it can tell if a page is about data science or about full stack development. This contextual targeting allows advertisers to reach the niche they are focused on and ensure the ads perform well. If you're really interested, you can try out our classifier, a part of our model, on HuggingFace, a site for sharing ML models.

After trying out our model and training it for a couple of months, we were finally happy enough with the performance to integrate it into our ad serving pipeline. We're really proud of this and that's why we went from v0.55.0 (yes, 50+ pre-stable releases) to v1.0.0.

What's next

Apart from privacy, one thing that separates us from the big ad players like Google and Facebook is that they can sell ads for almost anything you can think of. Currently, we have some traffic for a lot of developer niches, but outside of our main audiences and topics, it can be a challenge to run really focused campaigns on topics like "2 Factor Authentication" or "Web Development with Go".

However, as we get larger, our contextual targeting will scale well as our network scales and these highly focused campaigns will be possible. It's not an understatement to say that this kind of more advanced contextual targeting is the future of EthicalAds.

Another area we are looking at is to help sponsorships at large organizations deliver more value for sponsors. Currently, we're helping the Python Software Foundation to promote their sponsors across their ecosystem of web properties while also allowing them to report important metrics like how many times a sponsor's logo was seen and clicked. Check it our on the sidebar of the Python Package Index.

Join us for what's next

We're really excited about implementing new features on our ad network and continuing to iterate on sponsorships. Thanks again for being along with us on this journey to build an ethical ad network. Please email us if you have any ideas or feedback on our product or roadmap. We always love to hear from you.