SaaS Factory - Updates, Announcements, and How We Teach Machines
Strap in. This issue has a lot to it.
Recap: Who Should I Unfollow?
I’ve been continuing to build out Who Should I Unfollow? and I’ve gotten some pretty good feedback, overall.
It’s not quite to the point of running itself, but it’s at least close.
The biggest improvement I made was the adding in of notifications. I now allow users to get notified once their account has been analyzed, with the options of email, tweet, and Twitter DM being offered, tweets being the default. This actually lets the user notifications be a marketing tool in addition to helping them come back and see their account analysis, since that was getting lost. And it might also act as a faster email capture mechanism, since I don’t collect it until the user signs up for a paid account.
That will help me follow up with users who don’t end up converting a little easier too. At least, that’s the current thinking. Once I see how it performs, I’ll be able to determine whether or not that’s actually the case.
I also built the first workings of an admin dashboard. This should make it easier for me to support users who have issues. I had a couple of bugs that I introduced while making updates, but luckily I had users who notified me. The hardest part is always the initial setup, so it’s not much for now, but I’ll be adding to it as needed. But the big win was actually just setting up the backend to support admin functions, because as I build out more products, I’ll be able to use those endpoints over and over again.
And that takes me to the most important thing: I’ve figured out where I want to go with things.
I was on a Twitter Space the other day that I attend weekly for startup founders, and I was talking about the progress I’ve had with Who Should I Unfollow? and shared a preview of another app I’ve got planned: Who Followed Me?
Notice a trend?
But as I was talking about it, one of my friends from the startup ecosystem asked me why I’m building out these small products separately instead of building one larger product with everything. And there are a few reasons, which I’ve touched on briefly before, but since I’ve gotten a number of new subscribers lately, I wanted to go over my ultimate vision in grand detail.
The Ultimate Vision
All of this started with Feather CRM, which has been pushed to the side for now. I’ve been looking at rebuilding it to replace the piece that was shut down, but I’ve come to realize that rebuilding the pieces that were based on the external services is going to be time consuming and isn’t really what I want to do. Instead, I’m going to take what I’ve learned and pivot a bit on the implementation. I was planning on integrating all of these services with Feather and bundling them up, while having little a la carte products built from the individual features. Those features can act as lead magnets and lead users to upgrade for the full experience.
I’ve been really focused on online relationships, how we form them, and how social media influences those relationships we create, and I’ve realized that social media is terrible for forming relationships by default. Instead of focusing on quality, they focus too much on quantity. It’s not about the specific followers, it’s just how many followers.
Relationships, Metrics, and Competition
And this means the metrics you see don’t actually mean much. They are inflated and you can’t tell the difference in likes. There’s a lot of collusion between large accounts to spread certain messages. I don’t think you have as much potential for viral messages on Twitter without a certain amount of support.
And then I saw this tweet:
Tony builds some really cool products for Twitter. In fact, he’s probably one of my biggest competitors in the space. He’s got some great UI/UX features that I’m a bit jealous of. I’m not a designer and I can’t match some of that. He’s also full time on his products and he’s been at it longer than me.
He’s an indiehacker through and through.
Then there’s another competitor:
Tibo is another person I’ve been following for awhile. He’s one of the founders of TweetHunter. They’ve got more of a big team and a bunch of momentum. They’ve been branching out into LinkedIn, which is on my roadmap eventually, but I need to get a foothold in the Twitter arena before I do that. They just acquired a viral post generator for LinkedIn and launched a new LinkedIn product on Product Hunt.
He and his team have a bunch of bigger accounts that they gave equity to early on in their journey and have been growing at a really fast rate. But they do all sorts of little launches like the one in that tweet to build up a user base and be visible. They get a ton of support and their products always do really well.
So why am I talking about competitors?
Rethinking The Idea Of Competition
There are a couple reasons. If you notice, I’m actually taking cues from both of them. I’m building all sorts of related tools like both of them and doing little related projects like Tibo (and honestly, Tony does that too. He has all sorts of mini launches). But I also can’t compete directly. They are both more established and have access to more resources than I do.
So I’ve got to think differently. I can’t stand toe to toe, so I’m going to have to get creative, which is where I can shine. And that’s not even including Hypefury, which is a really established competitor of mine that I actually pay for.
If you’ve heard me speak before on this topic, you’ll find that I don’t really believe in competition. I prefer collaboration over competition. That means I’m not too worried about coming into a space with these more established companies. And they aren’t even aware of me, so I’ve got that advantage. But I can certainly learn from them and see what they do that works well.
And more importantly, figure out where they are weak. That’s an opportunity to fill in the gaps. That’s why competition doesn’t matter as much as people think. There are always tradeoffs that have to be made and that means that they can’t fit everyone’s use case.
See, here’s the problem with all of them. They aren’t focused on relationships as much as they are the numbers. They are trying to make numbers go up, which is the typical preoccupation for companies. And it’s probably the easier route, because it’s a metric that all users understand. You help their numbers go up and they’ll help yours go up.
There are a number of ways to do that. You can help users write tweets that perform well. You can help them be consistent. You can help them do all sorts of things that will lead to numbers going up.
And to some extent, they’ll get better at what they do.
But nobody ever asks about whether or not those numbers need to go up. What do they really track?
It’s easy to track numbers. It’s harder to track relationships. Because there’s not really a usefulness to getting relationship numbers up too high. As those numbers go up, the quality of relationships goes down. From the beginning, I had Feather set up to focus on quality of relationships over quantity.
I’ve created a number of excellent relationships this way. And I’ve been spending a lot of time helping people around me set up their own content distribution pipelines, funnels, and systems.
Let’s shift gears a minute and talk machine learning.
Machine Learning 101
When you’ve got a large dataset and you want to train an ML pipeline, what’s the first thing you want to do?
You want to identify your goal. The algorithm has to be able to identify certain traits. There are two ways to do this: unsupervised machine learning and supervised machine learning.
Unsupervised learning can identify patterns and groupings in the data without human help. These algorithms are ok, but most aren’t as useful. That’s at least my view of it, because there are a ton of factors that go into data correlation and many times, existing data capture systems don’t usually have enough context and therefore relying on patterns uncovered my unsupervised algorithms may or may not be useful. They are probably more useful for identifying places you might want to dig deeper.
But what about supervised machine learning?
These algorithms use labelled data to train on, and then based on the groupings they observe, can repeat that. This is a really powerful idea because if you can identify patterns and trends, you can train these algorithms to recognize them quite well. But the hardest part: labelling the data and ensuring that you are training on high quality data.
Let’s put this in different terms. What if I wanted to teach a random person on the street a subject like Quantum Physics? Sure, I could find some other random person to teach them the subject, but they likely won’t know it. Or if I wanted to do it like an unsupervised ML algorithm, I could get 1000 people to try to teach them, and hope that they are able to figure it out from the bits and pieces everybody teaches them. That doesn’t sound great, does it? Probably not going to be very successful.
But instead of having 1000 people try to teach them, what if we examined every person in that group and happened to find a professor of Quantum Physics? By labelling that person as qualified, we could then have that person teach the random person on the street Quantum Physics. Maybe they still don’t quite succeed, but there is going to be a much higher chance of success.
We tend to behave much the same way when we are learning, right? We don’t want to learn skills and concepts from random people, we want to learn from people that we think know what they are talking about.
How do we identify those people?
Building Trust Online
That’s the trick, isn’t it? Right now, we’ve got all sorts of people online who are acting like they know what they are talking about. And they use all sorts of marketing tactics to convince you of that fact.
But this is a bit odd isn’t it? That doesn’t really prove that they know the skill, it’s just proof that they understand marketing. I wrote about this a few months ago when I explored the next iteration of the Creator Economy: the Expert Economy.
As people have started adapting to the glut of people who are using marketing tactics to sell you their course on this, that, or the other, we’ve started looking at other signals to identify who knows what. One of the easiest things to look at: follower counts. If 50,000 people follow this person, they probably know what they are talking about, right?
Ok, so here’s my plan: create 50,000 bots and have them all follow me. Then I’ll be an expert right?
You see the problem. It’s not that hard to fake that level of data, right? Hell, I could connect each of those bots up to GPT-3 and get them to tweet all sorts of stuff too. They could look completely legit.
Which is where trust comes in. Trust is going to be so important in the future as the main indicator of your level of success. People are becoming less and less trusting, and where scarcity exists, that's an opportunity to provide a ton of value.
There’s talk in Web3 spaces about tracking reputation on the blockchain. And there’s some potential there, but I don’t think we necessarily need that right now. Because, again, it’s easy to fake. Most people don’t understand enough about blockchain to do the level of due diligence necessary to validate that on-chain reputation, so scammers will abound. Scarcity of understanding is an opportunity for scams, which leads to a scarcity of trust.
Are the pieces coming together yet?
With Feather, I was trying to figure out what trust looks like. I wanted to identify the data and label it. Who’s building trust, how are they doing it, and what do those accounts look like?
And that led me to realize something very important: trust means different things to different people. Everybody uses Twitter differently, with different levels of trust as a baseline, and different levels of trust with others on Twitter. And that level of trust determines so much of how they interact.
With all of that outlined, I’d now like to share my pivot:
Feather CRM is going to become Twitter Garden.
I’ve got a landing page built with a waitlist signup, but it turns out that Carrd won’t let me publish to any domain that has another big domain in the name. So I can’t push the site to twittergarden.com. Might have to build a custom landing page instead… Will need to figure that out, I guess.
The goal: give people the tools to create the Twitter experience they want, one that they can be proud of. A garden serves many different purposes: it can nourish, by letting you grow food. It can just be beautiful, a place that you have designed to be looked at and admired.
It can be a jungle, although that’s kinda what Twitter is by default, so people who want their Twitter to look like that aren’t my audience.
What if your Twitter experience was something you could be proud of, something you could share with others, and something more meaningful than “just another social media site”.
That’s my goal here. I see so much potential for Twitter, we just have to put in the work, weed out the bad actors, prune the inactive accounts and relationships, and plant seeds that can grow into something meaningful and beautiful.
So that’s my plan. I’ve got a long way to go, but I’ll be starting to design the initial pieces and connecting the tools I’m building into that garden context. They’ll be available separately, but also together if you want to build something more.
Announcement #2 - It’s “Book” Time!
I’ve been throwing around ideas for a book for a long time. I’ve known that I want to write a book, I just couldn’t quite narrow down exactly what that book would look like.
That changed a few days ago, when I realized I had the perfect title and I needed to just start.
But why is “book” in quotes?
What is a book? You probably have a pretty standard book in mind, but I’m exploring what a book really is, and I’m going to be sharing the process of compiling one in public (not really a surprise, is it?).
Introducing: The Hitchhiker’s Guide To The Future
The Hitchhiker’s Guide To The Galaxy was a book that really made a huge impact on my life and this is a way for me to pay homage to it while also providing a guidebook to the future for people who don’t know what’s coming.
The writing process is going to be extremely interactive, and much like Douglas Adams, I’m going to be exploring the content through different media formats. There will be the written content typical of books, but also a game of sorts, a podcast, and potentially other things that I’ll announce later. Still ironing out all the details, but it’s going to be fun. That’s my number one objective with it.
I’m going to be exploring new technologies as I write about them, being as hands-on as possible.
Each technology will be the focus of a season of the podcast version. First technology up: blockchain.
I’ve been listening to a podcast all about tokens and how they are used and I’ve wanted to explore some things related to how tokens can be used, how DAOs can be structured, etc. It’s not going to be a grifty Web3 project, but instead, I’m going to try to figure out how to reward my audience for doing things that help me write the book/host and promote the podcast, etc. Still exploring exactly how that will work, but again, should be fun!
Ok, that’s probably enough for this issue, before it ends up crossing the 3000 word mark. Clocking in at just over 2800 now.
Factory Production Stats
Let’s take a look at where things stand now. I’ve got Who Should I Unfollow? live and generating revenue. I’m also calling Friend Content bot a product that’s live as well. Spent some time tweaking Who Should I Unfollow? and am ready to start marketing it a bit more now. Gotta get some more traffic going to it as I prep to launch TwitterGarden in the coming weeks. Might throw in a launch of Who Followed Me? as well, as that’s something that will be a pretty fast launch I think, based on what I’ve already built.
Current Products Live: 2
Total Revenue: $94
Recurring Revenue: $24 MRR
Paid Users: 26
Subscribers at $5/month: 3
Subscribers at $1/month: 9
1-time payments of $5: 14