We hit 50,000 users on a Tuesday. I remember because our server costs tripled overnight and almost gave me a heart attack. There was no big celebration, no champagne. Just a frantic Slack thread with my co-founder that was basically a string of fire emojis and panicked keyboard smashes. That’s the reality of scaling an EdTech platform. It’s not the clean, predictable growth chart they show you in business school. It’s chaos. Beautiful, terrifying, educational chaos.
Scaling our AI learning platform taught me more than my MBA ever did. We weren’t just building another content delivery system. We were trying to create something that actually adapted to individual students, a private tutor in your pocket. And along the way, I learned some hard-won lessons—the kind that only come from seeing your assumptions get completely wrecked by real users.
This is the stuff they don't teach you in school.
1. Personalization is a Bug, Not a Feature (At First)
Everyone in EdTech talks about “personalization.” It’s the holy grail. But here’s the thing: when you start, true 1:1 personalization is a trap. We spent months building this incredibly complex algorithm that analyzed dozens of data points to create a unique learning path for every single student. It was a technical masterpiece. And it was a complete failure.
Why? Because at 500 users, you don’t have enough data for the algorithm to be smart. It ends up making weird, unhelpful suggestions. The “personalized” path felt random and students hated it. We were so focused on the AI that we forgot about the user experience. The big lesson here was to start with cohorts. We grouped students into a few key personas (e.g., ‘the crammer,’ ‘the diligent weekly learner,’ ‘the easily distracted’) and built semi-personalized paths for them. It was 80% of the value for 20% of the effort. Only after we hit about 10,000 users did our fancy algorithm start to have enough data to actually become useful. Don't build a Ferrari engine when you only have a go-kart chassis.
2. The “Correct” Answer is Often the Worst Teacher
Our first version was a glorified quiz machine. Question, multiple-choice answers, and a big green checkmark for “Correct!” or a red X for “Wrong.” Engagement was terrible. Students would get a question wrong, feel dumb, and churn.
I honestly had no idea what I was doing at first. I thought accuracy was the goal. But learning isn't about being right; it's about understanding why you were wrong. The breakthrough came when we redesigned the entire feedback loop. Instead of just saying “Wrong,” we started explaining the distractors. We’d show a student, “Hey, you chose B. That’s a common mistake people make when they forget to account for X. The reason C is the correct answer is because of Y.”
Suddenly, getting an answer wrong became a learning moment, not a judgment. Our engagement metrics shot up. We learned that the most valuable real estate in a learning app isn't the question, it's the explanation that follows a mistake.
3. Data Can’t Measure Inspiration
As a data-obsessed founder, I wanted to quantify everything. Time on page, questions answered, videos watched, you name it. We had dashboards for our dashboards. But we were missing the most important metric: inspiration. That “aha!” moment when a concept finally clicks.
You can’t track that with an event listener. We had to get qualitative. I personally emailed or called hundreds of our first users. I didn’t ask them what they wanted. I asked them to tell me a story about a time they felt frustrated and a time they felt brilliant while using our platform. Those stories, those messy, unquantifiable narratives, became our product roadmap. It’s how we learned that students needed more context, not just more content. It’s a lesson I’ve carried into my work as an investor, which I’ve written about in my guide to angel investing.
4. Your Biggest Competitor is YouTube, Not Other Apps
We used to obsess over our direct competitors. We’d track their feature releases, their pricing changes, their marketing copy. It was a complete waste of time. Look, I get it, you have to know the market. But the reality is, for a student struggling with a concept, their first stop isn’t another specialized learning app. It’s YouTube. Or TikTok. Or a group chat with friends.
We’re not competing for a budget. We’re competing for attention. And you don’t win the attention war with more features. You win with speed, clarity, and trust. Can we give a student a better, faster, more reliable answer than a 15-minute, ad-filled YouTube video? That became our north star. It forced us to be ruthless about simplifying our user interface and getting students to their “aha!” moment as quickly as possible.
5. The Ethical Tightrope is Real
When you have data on how 50,000 students learn, you have a lot of power. And with that comes a massive responsibility. We could see which students were falling behind, which ones were struggling with anxiety, and which ones were likely to drop out. The temptation is to intervene, to automate, to “fix” it.
But you have to be so careful. An AI-generated “You seem to be struggling” message can feel incredibly creepy and demotivating. We learned to use our data to empower human educators, not replace them. We built dashboards for teachers and tutors that flagged at-risk students, providing them with conversation starters and resources. The AI’s job was to find the signal in the noise; the human’s job was to provide the empathy and connection. It’s a partnership, not an automation problem.
6. Content is a Moat, But Community is a Fortress
We spent a fortune creating high-quality video lessons and interactive content. It was our moat, our defensible advantage. Or so we thought. The truth is, content is a commodity. Anyone with enough capital can hire experts and create a course.
What they can’t replicate overnight is community. The moment we added a simple, moderated forum where students could ask each other questions, the platform came alive. Students who had mastered a topic would jump in and help those who were struggling. It created a virtuous cycle of learning and teaching. The platform was no longer just a tool; it was a place. This is something I see all the time in startups, as I discussed in a post about scaling startups.
7. The Goal Isn’t to Finish the Course
This was the hardest lesson for me to learn. I was obsessed with completion rates. It seemed like the most obvious measure of success. But we were wrong.
Nobody cares about finishing a course. People don’t sign up for a learning platform to get a certificate of completion. They sign up to solve a problem. They want to pass a test, get a promotion, or build something they couldn’t before. Our job wasn’t to drag them through 40 hours of video. Our job was to give them the specific knowledge they needed to achieve their real-world goal, as efficiently as possible.
We shifted our entire focus from “course completion” to “goal completion.” We started the onboarding process by asking, “What are you trying to accomplish?” And then we tailored the content to get them there. Sometimes that meant they only completed 20% of the “course.” And that’s a win. They got what they needed and moved on with their life. They were successful, and they loved us for it.
The Real Takeaway
Building a learning platform for 50,000 people was a humbling experience. It taught me that the shiniest tech is useless without a deep, almost obsessive focus on the human beings who use it. It’s not about the elegance of the algorithm, but the quality of the learning moment. It’s not about the data you collect, but the inspiration you create. And it’s not about getting students to finish your course, but about helping them start their next chapter.
Frequently Asked Questions
Are these recommendations still relevant in 2026?
Absolutely. While specific tools and tactics change, the underlying principles remain consistent. I update my thinking regularly based on what I'm seeing in the market and across my portfolio companies.
How were these items selected?
Each item on this list comes from direct experience, either from building my own companies or from patterns I've observed across the 200+ startups I've invested in. I prioritize practical, actionable items over theoretical concepts.
Which item on this list has the highest impact?
It depends on your stage and context, but in my experience, the items near the top of the list tend to have the broadest applicability. That said, sometimes the less obvious items create the biggest breakthroughs for specific situations.
How do I know which items apply to my situation?
Start by honestly assessing where your biggest bottleneck is right now. The items that address that specific constraint will give you the highest return on your time and energy.