Building an EdTech-Startup
My attempt to improve our education
This is how I built my first startup, an EdTech Startup.
The beginning
I studied business administration. As a tutor I started my first business. A website where students at my university could prepare specifically for their math exam.
This project was successful, reaching more than 1000 students in 1,5 years and generating a revenue of around 6-10k€ per year.
I basically created the coolest student job ever. I only had a few work hours a year and enough money as a student.
The idea
"What I have is the perfect side project for a student. I should make this available for other students as well."
I pitched the idea in an entrepreneurship class and received an email from Emil, a fellow student interested in joining me.
We met, got along well, and started working together.
After some time to validate our idea, we decided to go for it.
The plan
Judging from an extensive testing phase, a survey we conducted, and the success of my first project, mawiwi.de, we believed in hyperspecific online courses for students.
From the profits of our MVP test, we acquired another website in our area that was already generating revenue through recruiting and ad deals with large companies.
Our goal was to enable students to easily create courses for their fellow students and earn money with it while we would take a cut. The differentiator was our software, allowing students to develop courses much faster.
When a math course was created at a new university, all materials from other courses could be reused and adjusted directly on our website, from interactive tasks to flashcards and even videos. This was possible through our Git-like version control system and easy-to-use material editors.
The launch
After a few months of development, we launched our website, studymaniac.de, in November 2022.
Two months in, we had around 250 paying users and good feedback.
But there was one problem. While demand was there, we couldn't find enough students to create courses.
After some exhausting months, Emil decided to leave the project.
Why did I keep going?
Digital education is my passion. In January, we had around 1000 daily unique visitors in Münster and Frankfurt alone, mostly from 3 courses and word-of-mouth.
I conducted a survey amongst the users, with the following results: Platform rating: 4.8/5 Price rating: 4.7/5 Content rating: 4.5/5
The feedback was the reason I could not stop, but I had to find a way to solve the supply problem.
The pivot
At the end of 2022, GPT-3.5 (in form of ChatGPT) was released, a new Large Language Model (LLM) that had enormous abilities. With my data science and NLP background, I monitored the progress closely. On the 14th of March 2023, GPT-4 was released, and I knew what to do.
After running some tests, I concluded that AI-generated courses may solve our supply problem. If not now, then in 2 years.
From the money we made in the first months, I hired motivated students to help me create the first AI-generated courses.
In a few weeks, we had the first prototype ready. We developed a process that was able to:
- handle multiple large documents and segment them into smaller parts that are assigned to topics
- generate content for each topic, including interactive tasks, flashcards, explaining documents and videos
- find the best YouTube videos for each topic
All the described steps cost us around 20€ per course and took around 20 minutes of generation time.
We tested it with the first students, and the feedback was mixed. All were surprised by the quality of the generated content, but it was still far away from our human-made courses. So, we made some adjustments and launched the first AI-generated courses for a few euros.
This showed that students are indeed willing to purchase AI-generated courses, as we labeled them as such.
I also quickly launched an update that enabled students to give feedback directly inside the courses and improve them independently. Similar to Wikipedia, everyone can edit all materials, create new ones, or suggest the deletion of existing ones. For each change, a voting process is started amongst the peers of the course. This was a measure to fight the quality problem that generative AI poses.
In a few days, we sold more than 100 courses. The feedback, again, was mixed, and time was running away as exams were approaching.
The results of the pivot
I felt quite some pressure and lost hope at times. I desperately wanted to validate my hypothesis about AI-generated courses in the short time frame until the next exam phase in Germany. At the time, we were 4-5 people working on the project, mostly relatively inexperienced.
It was a very exhausting time for me, and I needed a week off after the exams. On the one hand, our results were promising, but the feedback after the exams was that the quality was not good enough. Because we labeled the courses AI-generated, our customers were not too upset about it. After quite a few feedback cycles, I felt that we had hit a plateau in quality, and it seemed that we would not be able to improve it significantly in the short term, at least not for our target customers, who were economics and computer science students.
While I was proud of our technical achievements, I was not satisfied with the results as I also felt this was different from my vision of how I wanted to improve education.
Whats next?
We arrived at a very exciting platform - a platform that was super optimized for multiple creators to work together. Editors for every material type, a version control system, a voting system, and pretty much everything essential for a learning platform.
I believe there is no alternative to Wikipedia that offers learning materials and courses to actually help students pass exams.
Or even broader, help people pass tests. Be it pupils, students, or professionals.
I know that Studymaniac can be this platform due to its technical foundation!
Here, you can find a detailed post about the technicalities of the platform.
The exit
In March 2024, I sold Studymaniac to Ecoreps, a company that specializes in digital education for students and has been around for years.
I am very happy with the result. The negotiation process went smoothly and the buyers will continue to operate the business and serve my customers.
I am now looking for new challenges and am excited to see what the future holds for me.
Thank you for reading this! If you have any questions, feel free to reach out to me at jannis@studymaniac.de