Substack | The Big Bet to Dominate Publishing
And how Substack could use A.I. to supercharge their platform.
Hey, I’m Jacob ✌️ Welcome to all of the product people who subscribed since last time! If you’re not a subscriber yet, I hope you’ll consider joining this band of like-minded folks who are all learning how to build better products. Questions or feedback? Drop ‘em here.
The Creator Economy
When Chris Best, Hamish McKenzie and Jairaj Sethi co-founded Substack back in 2017, their vision was to create a paid publishing platform for writers.
Even though blogs and email newsletters had existed for decades, digital-native writers (like me... sort of?) were still hacking together their own string of tools to publish and get paid online.
In 2014, I started a blog that I published 3 times a week. The creator economy was starting to blow up and your email list was the goose that laid golden eggs. I had heard stories of people making $50-100K off a single launch email to their list. I just had to build a list and I would be printing money.
To get set up, I had to:
Set up a website. Okay, easy enough so far. I just set one up on Squarespace for $99/yr. Piece of cake.
Set up an email newsletter. MailChimp was blowing up around this time, so I just set that up real quick. Free until I hit 2,000 subscribers.
Capture emails to keep readers coming back. Seems straightforward. I just dropped a form block on the Squarespace site and integrated with my MailChimp account. Boom. Done.
Except, not so fast... this was where I slowly began to realize the breakdown. You see, the blog would generate some SEO over time which would drive top-of-funnel traffic. But only a small percentage of those readers would subscribe to the newsletter which meant, if I had a way to measure my churn rate, it was probably something like 98%. People would stumble across the blog, close the tab and I'd never see them again. This is where a lead generation tool would come into play to incentivize those readers to subscribe to my list.
Create lead magnet. Create a lead magnet PDF and offer it for free using a tool like SumoMe in exchange for an email address. Okay, maybe that boosted my conversion rate to like... 7-8% instead of 2%? Still not great. But I trudged forward.
Cross publish to blog and newsletter. Okay. Time to start writing. I was putting out 1,500-word articles 2-3x a week for years. And every time, I'd have to publish the new article to my blog, copy and paste (and often, reformat) that same article to MailChimp to send it out to my list.
A little extra effort. What's the big deal? Well, my email analytics and blog analytics are totally separate and are super confusing to aggregate. My blog was tracking a million metrics through Google Analytics including total page views and unique visitors. My email list was tracking some other stuff including total opens and unique opens. I knew those numbers were important but I had no clue how to extract a single, valuable, trackable metric from it all.
And monetization? That was a full-time job by itself. I sold physical and digital products, tried affiliate links and even made some headway toward building a course before I gave up.
Writing was fun for me and I wished I could just get paid for that. But everyone said writing was just a way to attract an audience you could sell a course or an eBook to. Paywalling your writing was a death wish. If your writing couldn't be indexed for SEO, how would you ever grow your audience? Also, there wasn't any good way to paywall your writing anyway.
So after 3 years of not making more than a few hundred bucks, I decided to shut it all down. Coincidentally, that was in 2017. The same year Chris, Hamish and Jairaj started Substack.
So close 😅
The “Paid Publishing Platform” for Writers
Their premise was simple. Writers should be able to:
get paid for their writing.
own their audience.
not have to get a CS degree to do it.
So they started with what I believe was the perfect combination:
Email + Blog + Payments
By creating an email tool that also gave you a hosted blog, aggregated your analytics and offered a simple monetization tool, Substack finally made writing on web for a living an accessible dream.
It solved a few very poignant problems for digital-native writers:
List Building: No longer did you have to create a stupid lead magnet or add those annoying pop-up modals on your website to get people to subscribe to your email list. With Substack, your blog WAS your email list.
Multi-channel publishing: Anytime a writer would hit "Publish," their content would be simultaneously published to their blog AND sent to their email list. This meant more eyeballs, more consistently WITH the added benefit of SEO.
Easy Analytics: Though I would actually appreciate a few more analytics tools in Substack, you can't argue with the power of simplicity. For every article, Substack gives you one big number to focus on: Total Views. Is it going up or down? That's all that matters.
Monetization: Writers no longer had to write a book or be a reporter to get paid. They could publish to the web and let people subscribe to their work! And it would only take a few minutes to set up.
How to not lose writers to competitors…
The big challenge that Substack is faced with is helping their users grow and monetize their writing while not losing them to competitors with different value props. There are a host of alternatives from ConvertKit to Revue, to Ghost. The most recent (and IMO, the most threatening) rival is a company called Beehiiv.
It’s a growth-focused newsletter tool built by the folks behind the ridiculously successful business newsletter, MorningBrew. The engineers that built their referral system set out on their own to make that same system accessible to other newsletter creators. They've since included an ad network for easy ad monetization alongside their subscription offerings to give writers a suite of tools to get paid.
With competitors moving swiftly to steal marketshare, Substack has a decision to make. Will they copy the features of their competitors or double down on their value prop of empowering digital-native writers?
Since the beginning, Substack’s founders have touted that the MOST important thing is to empower writers to write. And I think I agree. I believe that
…the best leading indicator of growth is quantity of content.
…the best leading indicator of monetization is quality of content.
Substack's challenge is to help creators be MORE prolific WHILE maintaining a high bar for quality. A tall order to fill.
Here’s what they did 👇
Substack’s “Blurred Lines” Strategy
They leaned into this writer-focused strategy in a few ways including a ridiculously powerful recommendation engine, but I only want to focus on what I believe to be one of the biggest opportunities for Substack to win:
An excerpt from an article on Substack in The New York Times paints a accurate (though partial) picture of their current challenge:
“The good news for the company, five years old this summer, is that it is still growing. […] But to maintain that growth, Substack executives say, the company must offer more than newsletters.”
—Tiffany Hsu, Substack’s Growthspurt Brings Growing Pains
They started blurring the lines between a blog and an email newsletter by introducing a simple primitive: one-click publishing.
You write. You publish. Substack handles the rest.
Send via email? Done.
Cross-post to blog? Done.
Share on Twitter? Done.
And in the last year, we've seen them begin to blur the lines across other forms of content as well. That simple hub and spoke model of creating multiple pieces of content with a single push of a button was a powerful unlock for a lot of creators. And I believe they're leaning into that.
🗣 Unlocked: Text-to-Speech Voiceovers
They've supported podcasts and audio embeds for a while, but notably, in June, they released text-to-speech voiceovers. With ZERO additional effort from you, Substack offers an audio version of your article for readers to listen to while they drive to work, go for a run or clean the house. It just happens with that one push of a button.
💬 Unlocked: Threads
Then again, in August, they released Substack Threads, a way to engage your readership through a Tweet-length prompt and then use the Substack app to participate in a Slack-style conversation. Once again, allowing writers to create new types of content with that same single publish button.
Substack could have built a powerful podcast editor or a general-purpose messaging tool, but they didn't.
I believe it's because they are leaning into the beloved simplicity of raw creation and 1-click publishing. The point isn't infinite flexibility. If it were, there are plenty of other more powerful tools in the world.
The point is to let creators to create without worrying about the complexities of getting their creations out into the world.
Substack’s Big Opportunities
Now, with the understanding that I don't work for Substack or have any inkling of what their product strategy is, I do have an idea that I believe would give Substack a massive competitive advantage.
1. A.I. writing tool
If you missed my article last Tuesday on A.I. writing tools, perhaps start there.
But to summarize, I tried three A.I.-powered writing tools - Jasper, Copy.ai and Lex - and gave my bull and bear case for whether or not it'll replace human writers.
I think we can all agree that it's impressive tech. But one of the biggest downfalls is that tools like Jasper don't have the correct training data to understand MY writing style.
My writing is pretty casual - I tend to write like I talk. I overuse ellipsis to communicate a pause and dashes instead of semi-colons. I’ll throw in filler words like "um” or “yeah” because I think it’s funny to read. And I’ll even start a sentence with “and.” Sorry Mom.
But you know who does have that training data?
At the time of this writing, I've published 21 long-form articles on Substack. Other writers have published many, many more. By layering those articles on top of the base data model of GPT-3, I think Substack could begin to unlock a powerful A.I. writing assistant right inside of their editor.
This A.I. writing assistant would not only help writers create better content more often, it would also widen Substack’s competitive moat the longer a writer uses it. With each article, the A.I. is better trained and can offer more useful suggestions. Writers who get used to being prolific with that tool in their tool belt will find it difficult to jump ship and lose that finely tuned assistant.
A.I. can already write like human. The next evolution is going to be helping it write like a specific human.
Of course, even without the benefit of modeling the writing to fit my personal style, there’s a lot that AI can help with in the writing process:
Coming up with topic ideas
Coming up with good titles and sub-titles (interesting title formats are mostly universal)
Constructing outlines before you start writing
Extrapolating out a bit, I think Substack could lean into their 1-click publishing primitive and offer supporting content written by A.I. to augment the article, including:
Tweets to promote the article
Community Thread topics based on the article
Free or paid subscription calls to action
Intros and conclusions
To summarize, Substack has opportunities to use a writing assistant to:
Leverage the base data set of GPT-3 to offer an A.I. powered writing assistant in the native Substack editor.
Use the author's existing body of work to layer on more personalization to the A.I.'s suggestions.
Lean into the hub and spoke model by generating short, powerful supporting content to help the writer promote the article and foster engagement.
2. A.I. voice tool
Since we're already a few miles down the tracks on the crazy train, let's just push it a bit further.
Remember the A.I. generated podcast where Joe Rogan interviewed Steve Jobs that I mentioned in Issue #3 of Product Nuggets?
While I realize that it was trained on thousands of minutes of audio from both Joe and Steve, I wonder if there's some sort of middle ground.
Let's say I recorded a voiceover for the 21 articles I've already published. Substack has the words I wrote and the audio of me saying those exact words which is super helpful for training a text-to-speech model. So now it can pick up on things like cadence, pauses, pronunciation, accent, inflection and emphasis. Could it then take the existing bot voices that read my articles currently with the text-to-speech function and slowly begin to refine them until they sound more and more like me over time?
I'm gonna cut this short, because I'm not well-versed enough on this topic to avoid making a fool of myself. But it's fun to dream, isn't it? :)
This entire article was written and scheduled to go out but then last night I was listening to the latest episode of Lenny’s Podcast with Sachin Monga (Head of Product at Substack).
And somewhere right in the middle of me shaving I hear Lenny say…
“Something that I suggest y’all look into a little bit is OpenAI assisted writing.”
Lenny! Did you scoop my article??
What can I say… great minds think alike.
Psst, Lenny, if you’re reading this hit me up on Twitter!
It’s time to click “publish”
To bring it home, I believe there are opportunities for Substack to develop their creator tools even more. Tools that help digital-native writers expand their content empire with little to no additional effort. Tools that lean into that 1-click publishing primitive and make it dead simple to create better content more often.
In combination with their ever-growing network effects powering their recommendation engine, I believe they have a bright future ahead.
P.S. — Do you have a Substack publication? Drop a link in the comments!
❤️ Smash that heart!
If you enjoyed this article, smash that heart icon to show some love! 🙏
Super interesting perspective -- having the data for a personalized assistant that gets better over time (and to keep writers motivated to train it) - very cool possible play.