I was debating writing about this. I’m not a huge fan of growth hacks on principle, but that said, I’ve come to a big realization.
I’ve got to get better at using my skills effectively, and one of those skills is automation.
But how can I find the balance between automation and focusing on the people I’m working with? That’s something I’ve struggled a lot with. I’ve seen a ton of people using AI in all sorts of ways to personalize their outreach, but that feels off to me. I like to focus the personalization in more authentic ways as I reach out to people. And yet, the numbers make personalization more difficult, if I want to reach out to more people.
So what should I do?
Let’s take a step back and focus on somewhere I struggle. I have a really hard time doing cold outreach, so during a coaching call I had with one of my podcast guests (side note: episode just came out. You should listen to it.), I told him that one of the biggest issues I have is a resistance to cold outreach.
It’s something I’ve tried really hard to overcome. I think it has to do with the many times I’ve gotten my nerve up, sent out a bunch of messages, and done so badly or with no results. The pain of rejection is real.
I know I’ve got to overcome that, so how can I do so?
Caelan gave me a great tip. He said, “What if the cold outreach wasn’t so cold? What if you asked them to give you a few minutes of their time for a whitepaper you were writing?”
A lightbulb flipped on for me. From there, I could start building the relationship, offer to reach out to the recipient if they were interested with the finished whitepaper, and then continue the conversation if they were interested in tools as well.
So, now I’ve got a strategy. But how can I figure out who to reach out to?
Back to the growth hacks.
I consider myself to be pretty decent at the whole Twitter thing, although I’m not as effective as I know I could be. So I asked myself, “How can I use Twitter to identify podcasters?”
The first thing I did was looked through my database of Twitter users and ran a query for all users who had “podcast” in their bio. That got me over 3000 accounts, but I wasn’t completely happy with that. I wanted to know about people who were actively publishing podcasts, and I didn’t want to try to run a ton of analysis on them.
How can I identify podcasters who are publishing new episodes?
I can search for “New Podcast Episode” and pull a list of all people who tweet that.
But I don’t want to just search that constantly and send a bunch of DMs manually.
Automation to the rescue. I went to my N8N instance and set up a quick automation that runs every 4 hours and pulls the last 50 tweets that include “New Podcast Episode”.
Great, now I’m ready to go. I queued up a batch of Twitter accounts and ran into my first problem: the majority of Twitter accounts I’d found had their DMs closed.
In my first batch of 50 accounts, I ended up sending out 6 DMs.
And got 1 response that made me rethink my approach.
“As I get a lot of people asking me for feedback and stuff. I ask the question, what's truly in it for me? And it's not me asking for money, not in the slightest. But what, in a sense, do I get for my hard-earned insight?”
So, I explained a little bit about what I was doing and offered them free usage of the tools I’m building.
And then I got this back:
“I will respectfully decline, and I'll tell you honestly why. I work as a digital marketer. You could have hooked me if you said or mentioned 3 things about my content you liked with references in your intro to me. I want to know if you spent "your" time learning about who I am. Been getting a lot of this of late and giving my time up for that. I want to feel it's well earned. No offense, but I hope this helps when you reach out to the next person. That said, nothing but lots of love and luck in your venture!”
Well, shit.
The more I thought about this response, the more I realized it was totally on point.
I’d feel the same way. But if I want to reach out to hundreds of people, I can’t listen to/watch all of their stuff if I want to get other things done. So I went back to the drawing board and thought about what it was I was actually trying to do.
Something I’ve been thinking a lot about is the future of trust on the internet, especially with the rise in AI. Instead of using AI to reach out to people with personalized messages that can reference their content and entice a response, I tried to be honest in my message, at which point I failed miserably, because there are a ton of people reaching out with messages and everyone is tired of it. Hell, I’m tired of it, even when it’s something that benefits me!
I still like the idea of collecting data from podcasters. Ultimately, I’m still trying to help them.
I’ve just got to overcome this problem:
Why should people trust me? They shouldn’t. If I want them to do what I’m asking of them, I’m putting a lot of work on their plate. They have to go to my profile, decide whether or not I’m reasonably not scammy, then fill out the survey, and eventually end up trying out a product.
That’s a lot.
How can I bypass that level of work to get to a place of trust with people?
And now let’s talk again about growth hacks.
Hacking the Growth Hacks
There are a lot of ways to hack trust. And let’s be honest, that’s what we want to do, although I think there are good ways to do that and bad ways to do that.
Here’s the new approach that I’m coming up with literally as I write this. I think there’s a place to capitalize on the trust I’ve already earned in two places:
People who already trust me and want me to succeed. I’ve been putting as many good vibes into the world as I can, and this is a place where people can help me out if they want.
People who vibe with my ideas. I know not everyone will, because I’m breaking away from a lot of standard practices and what we’ve seen as traditional success.
I think that’s where some big opportunities lie. Think about the trust you’ve already established and ask yourself if there’s a way those people can help you achieve the outcomes you want.
Can you come up with ways to incentivize them to reach out on your behalf?
Those ae the ideas I’m playing around with right now.
Setbacks
Here’s an email you don’t want to see:
Especially when it’s followed up a little bit later by this:
In a nutshell, Pinecone is a vector database that I use to store embeddings, which allows me to store numerical representations of podcasts that I can use as the basis for searching within them.
And Pinecone accidentally deleted all of my data. Almost 1M records that were deleted.
Luckily, I also stored the embeddings in my Postgres database, so I didn’t need to re-run all of the transcripts through GPT to recalculate them. But it still took about 4 days to reload everything. I probably could have done some better parallelization, but I just wanted to run them as easily as possible.
The app still isn’t performing as well as it did before though, so now I’m wondering where the data got messed up.
So right now, I’m evaluating a different vector database: Chroma. I’ve chatted with one of the founders a couple of times, and I’m really excited with where they are going. And it’s open-source, which I’m a big fan of. Eventually, I’ll switch Choose Your Algorithm over, but I’m starting with a bit of a new project.
Modeling The Future
Something I’ve realized with Choose Your Algorithm: most people don’t get it right away. I’ve tried to model it a bit with podcasts, but that hasn’t been quite as successful as I wanted and it blends into a few other products.
So I’m doing something a bit different: I’m building out my own AI.
It’s going to be everything I believe to be possible with it. And it will be a testing ground for new ideas and approaches. Right now, it’s a super basic chatbot that has two main features: it can answer questions about the future based on my blog Hitchhiker’s Guide To The Future and it can tell you that it will ask me, request your email, and save that request to a database. It’s a way to offer a mechanism for me to find out what people want to know (therefore driving content ideas) and will allow me to deliver answers in a personalized way.
Coming soon: podcast recommendations, more information about the project, other blogs, probably this newsletter, etc., etc., etc.
Changes
I’ve been spending some time thinking about my goals and how everything I’m doing fits together, and one of the things I’ve realized is that this newsletter doesn’t quite match what its initial goal was anymore. I’m at the point where I think I could launch a new product each week, but that doesn’t mean that I want to do so.
I’ll continue writing, but I'm planning on experimenting with formats in the coming weeks. I’ve got an idea that I think will help me publish more regularly and build up some of the stuff I need for the AI site to function well.
Stay tuned!