How Building Better AI Makes My Marketing Easier

Hello! What's your background, and what are you working on?

Hi, my name is Przemek Chojecki. I was born in Warsaw, but am currently living in London. I did a PhD in mathematics in Paris, France, then I worked as a Research Fellow at University of Oxford. I went back to Warsaw for three years and basically switched my career from academia to business, and from pure maths to machine learning.

So far I've managed to fail two startups for two different reasons. Contentyze is my third startup and it definitely is looking the best so far.

Basically, we're building a content generation platform with a goal of turning raw data into compelling content — be that content marketing copy, automated news or e-commerce product descriptions.

home

We launched the beta version of the platform in April and we currently have two to three people signing up each day. The service is currently free, but we make money thanks to our content, as we use it ourselves to experiment with engaging stories. We manage to make between $4,000 and $5,000 through ads and affiliate links.

Eventually, we plan to start charging a standard monthly subscription fee as well as for additional requests a user might have for AI-generated text.

What motivated you to get started with Contentyze?

I always loved content in a broad sense. I've read a lot, I've written three novels (one of them is available in English!) and was generally quite engaged in culture. On the other hand, my second passion is productivity. I love the feeling of getting things done and thus I love automating as much as I can.

Since 2019 I've been heavily experimenting with various forms of writing, such as running various blogs, the biggest one being my channel on Medium with 3,000+ followers. Having a technical background and using my love of automation, I started tinkering with ideas on how to automate my content creation process. That's how Contentyze came to be. And it's greatly helped create more content.

Even if you fail now, the skills you're acquiring will be valuable in the future.

Tweet
Share

The very first validation came from the semi-automated media outlet PetaCrunch. The goal was to automatically interview founders of recently funded startups (global startups which had raised $1,000,000+ rounds) and my team of AI journalists managed to interview 1,000+ executives in fewer than three months. That was the point when I decided I wanted to go all-in with (semi-)automated content creation.

What went into building the initial product?

Contentyze came to be in January 2020, and at the beginning it was basically me tinkering with various machine-learning models. Then I hired my friend, a Python back-end developer who helped me set up a platform, and then I also hired my master student, who focused on training more deep-learning models. Together with another full-stack developer, we're a team of four.

We're totally bootstrapped, and funds for the project came from me personally. The first version of the platform was ready after four months, and after six months the new flashy front end was ready. In the next two to three months we plan to launch new amazing features. Starting in September we plan to focus on approaching clients on a larger scale.

So far, all our efforts have gone into just building the product, which I've then used directly in my content-creation efforts. In a sense I've built Contentyze for myself. I'm happy it works, and now I'm exploring how helpful it might be for others as well.

Before March 2020, it was a bit scary as I had almost no revenue outside of individual consulting gigs I was doing as a data scientist. Fortunately, since April I've managed to better understand how to monetize content and it's been steady growth since then.

We're currently in a comfortable situation where we don't need funding for our development. Nevertheless, I'm talking to various VCs, as I found their remarks valuable and I'd appreciate a business partner in someday scaling our operations one hundred times.

What's your tech stack?

Our platform is based on Python and Django. We're using Python because that's the language of choice for doing machine learning.

We're happy to be a part of the Microsoft for Startups program, which allows us to use cloud computing for free. That's a huge deal for us, as we already spend $1,000+ monthly on various computations, and that's a number that tends to grow.

We're constantly facing new technical challenges as text-generation algorithms are quite tricky to control, especially if you want to have high-quality text output.

How have you attracted users and grown Contentyze?

Contentyze is all about content creation, and because of that content marketing is quite easy for us. We literally create from dozens to thousands of content pieces per day through various platforms, so it's becoming easier and easier with time to discover us.

automation

We also spend some funds on advertising through Facebook, which is working fine for us.

Our strategy is about growing through organic traffic, and we plan to do that through creating content using our own platform. I believe this will get us really big thanks to the cycle of more-content-gives-more-users-gives-more-revenue-gives-better-machine-learning-models-gives-even-more-content.

I definitely recommend content marketing for any starting entrepreneur, even if they don't have a product yet. Starting a Medium blog or a newsletter really costs nothing but time.

What's your business model, and how have you grown your revenue?

We're quite unique in the fact that we have two viable business models.

The first model is standard. It's just SaaS revenue: a monthly subscription to our platform or API requests, and payments per request. We will incorporate this model in September or October. For now, the platform is free.

The second model, which right now brings between $4,000 and $5,000 monthly, is indirect, via using our platform to generate content which is then distributed through various blogs and other channels. This is particularly interesting because it pushes us into the direction of becoming a new vague media company, which from my perspective is very interesting.

We've seen a steady growth of revenue in this second model. Basically around $300-$700 growth month to month:

  • June: $4,000+
  • May: $3,500+
  • April: $2,800
  • March: $2,400
  • February: $1,800
  • January: $1,500

However, these numbers don't show our margins well, because much more work went into manual content generation (read: me writing content by myself) in January and February than in May or June, when we already had various processes set up.

Month Revenue
Jan '20 1500
Feb '20 1800
Mar '20 2400
Apr '20 2800
May '20 3500
Jun '20 4000

What are your goals for the future?

We want to finish development on our platform by September. Contentyze is available right now for anyone to test, but not all the features (like automated distribution to WordPress or social media) are implemented yet.

Up to now we've been bootstrapped, but I do hope to raise a VC round in Q4 2020 or Q1 2021 with a goal of boosting growth. We do have some scaling challenges. For instance, how computation-heavy some of the machine learning models are, and how difficult it is for first-time users to start using the platform right now. We want to correct both of these things.

So these goals are more on the product side. On the traffic and revenue side, I'd love to reach 1,000 users in 2021, and reach $10,000/month in revenue. To do that we'll have to scale smartly.

What are the biggest challenges you've faced and obstacles you've overcome?

I guess the largest mistakes I've made were certain decisions in my previous two startups. Fortunately, Contentyze is growing fast from month to month and it seems alright now.

This is the first product startup I'm running by myself (my previous startups were run with co-founders), and I love it. But of course, that might change with time. It's only been six months since we started with Contentyze.

Have you found anything particularly helpful or advantageous?

I love quick iterations: clear goals with deadlines. This is not always possible with machine-learning models but it definitely works with software-engineering parts. Fast execution is key.

There's no single road to running a company, and no single template for a successful company.

Tweet
Share

After running two startups previously, the crucial trait I've developed is confidence in what I do. In my prior attempts, I was always looking into what other people were doing with their startups or businesses and trying to implement everything I deemed valuable. Right now I'm in the happy spot where I'm fully aware that there's no single road to running a company, and no single template for a successful company. The key to success is dedication, patience, and being 100% sincere with oneself. You can't lie to yourself that everything is going great when it's not.

What's your advice for indie hackers who are just starting out?

Patience is vital. Work hard, but work smart as well and think about the direction you're going. In the end, it's the long term that counts. Even if you fail now, the skills you're acquiring will be valuable in the future.

On the other hand, you should also be critical of your goals, evaluate them constantly, and measure whether you're going in a good direction. It's easy to get stuck on a single project just for the sake of the time you've already spent on it, without looking realistically at future prospects. That's why I'm a huge fan of measuring everything I can when it comes to business, from users and revenue to the skills I'm acquiring.

editor

My final piece of advice is to talk to people and seek their feedback. Don't do everything people tell you to do (as there will always be many contradictory opinions), but be open to any criticism. I tend to write down all my feedback, and if any points get repeated I take a longer look at them and try to understand where they come from. This way I'm in the constant iteration loop and can make my product more appealing to others.

Where can we go to learn more?

First of all go to our website, where you can sign up and test the platform for free. I'd love to know your feedback!

Otherwise, the best ways to follow me are Medium and Twitter.

Let me know if you have any questions regarding Contentyze and how to use it!

Przemek Chojecki , Founder of Contentyze

Want to build your own business like Contentyze?

You should join the Indie Hackers community! 🤗

We're a few thousand founders helping each other build profitable businesses and side projects. Come share what you're working on and get feedback from your peers.

Not ready to get started on your product yet? No problem. The community is a great place to meet people, learn, and get your feet wet. Feel free to just browse!

Courtland Allen , Indie Hackers founder

  1. 3

    Great work! Are you using GPT-2 for the model?

    1. 1

      Thank you, one of our models is finetuned GPT-2. :)

      1. 3

        Nice! And I imagine it'll get even better with GPT-3 if you get access :)

        1. 1

          Definitely! We're on the waitlist for GPT-3. :) But also the large part of what our users do is templates and scripting as you can just import a spreadsheet and make a bunch of stories out of it.

          1. 2

            Nice. I like the interface and focus on automation. Even if you get 'competitors' or users have ways to generate the content from other models, you're setting yourself apart with the ease of use and script/automation/integration functionality. :)

            1. 1

              Exactly! My ultimate goal is to make writing simple for anyone. :)

  2. 2

    Inspiring story and product. Thanks for sharing.
    One question: what does "generated texts" mean in your pricing? What is a text?

    1. 1

      Thank you!

      A generated text is any instance of clicking on Generate or Preview in Templates.

  3. 2

    Cool stuff and nice to meet someone with such a similar background (Warsaw -> London -> tech startups)! I also worked with text generation tech before using transformer and RNN models; not simple to set up end-to-end but once you get a commercial product working the results can be quite astonishing.

    What made you decide to promote 3 separate products at once (templates, writer, SM)? My hunch would be that one is hard enough to promote :)

    Regarding your landing page, have you also thought of either providing samples of the text your software generated and/or adding simpler visuals about how it exactly works? I'm thinking something like "templates + data + Contentyze => story" each accompanied by a real life sample. The reason I'm suggesting thinking along these lines is that it could help your potential buyers understand the problem & solution more viscerally. Not sure they'd otherwise make the effort to dig through the writing and connect all the mental dots.

    1. 1

      Thank you! :)

      Having '3 products' in 1 is in my mind a neccessity if you want to beat competitors and offer an entreprise-level product for really generating engaging content at scale.

      Actually we have samples, tutorials and real life examples in Solutions: https://contentyze.com/solutions.html

      You can choose between Real Estate, E-Commerce, Finance, Media, Marketing and each use case has a tutorial in Contentyze using real data. :) Definitely I want to have even more explanatory texts and videos on how to use it, but this should be a fairly good start.

  4. 2

    Congratulations! I tried to sign up but the page doesn't work: https://app.contentyze.com/accounts/signup

    1. 1

      Still causing sign up issues....

    2. 1

      Sorry for the problem, we're actually deploying the new version right now. :) It should be back to normal within an hour.

  5. 2

    Great work! Are you open to feedback? Because I just tried the editor to generate content around Lead generation. The first try was for it to generate short text. However, I noticed that the content generated sound so much like a poem, not really marketing content. On the second instance, I used a like on Lead generation from my site's blog and requested extractive content. And it pulled the entire content. Is this how it's supposed to be used?

    1. 1

      Sure, I'm open to feedback. :) Regarding your trials, I don't know what you've put as an input exactly but:

      • when generating text from an input, you should put a full sentence in or a question like "How to generate leads online?"; you'd have to generate the same thing couple of times to get a great text
      • by requested extractive content you mean Extract Rewrite? Then yes, it should basically pull the entire content and make it shorter. You should try Abstract Rewrite for a more different text.
  6. 2

    This is super interesting. I think services like yours will have a massive impact on the internet as a whole, especially when they use GPT-3 and beyond.

    Many of us with a background in writing and content production feel threatened by automated content services. I think it's a valid fear, although reading some of your auto-generated blogs on contentyze.com, there's a long way to go before you can pass the Turing Test, so to speak.

    Many auto-generated texts make grammatical sense but feel weird to read because they don't really have an argument, or central idea; they don't go anywhere. It just reads like text filling up a word count. Your post here https://blog.contentyze.com/content-48/ doesn't really answer the question posed in its title, is structured strangely, and cuts off before it ends.

    I think for a few years we'll be in a hybrid spot where human editors are necessary to make AI writing actually flow properly. After that, who knows?

    Congratulations on your business - I don't think using it to replace a content marketing department is a good idea, but I'll certainly be using it to perhaps solve writers' block or save a bit of time. If I can write 50% of an article and leave the rest to a machine, maybe that could work.

    1. 2

      Definitely, I agree with you - the goal here is not to replace writers or content marketing departments but rather boost them and save time, so that they can focus on more creative things.

      A great example here is media industry, which implements automation for a couple of years now. They don't want to replace journalists, just make them not work on boring, repetitive texts like: weather news, traffic or stock updates.

  7. 2

    I love this story. I'm about starting my journey as an indie hacker building my first SaaS. This really helped my nerves.

    1. 2

      Good luck, SaaS companies are a lot of fun!

      1. 1

        Thanks. Do you have more resources I should consume on the subject?

        1. 1

          To be honest, not really - it's mostly trial and error, reading various articles and talking and watching interviews with other entrepreneurs further along the way. :)

          1. 1

            Thanks.

  8. 2

    What a fantastic journey! It is very inspiring for someone like me. I also gave up academic research for industry, aiming to be start my entrepreneurial journey. Thank you for writing this article.

    1. 2

      Thank you! It was quite a bumpy ride. It's never easy to make this switch and go from academia to business. But it was definitely worth it. Good luck for your journey! :)