Tutoring platform, part 2 of 2
The business I shut down
I built a better product. Not enough people ever saw it.
Context
After the first version failed, I changed how I was building, not just what I was building. Instead of learning a subject fully and then executing, I started from one question: what is the smallest version of this that gives a student real value? I broke the build into pieces, the user flow, lesson delivery, onboarding, log-ins and payments, scheduling, and learned only what the next piece required, using it the same day. Every week I released an update to a small group of students and collected anonymous feedback.
By the measures I was watching, the product got better, and for a while it did take some of the load off live teaching.
The problem
Usage didn’t follow. I had assumed the hard part would be building the product. I was wrong.
A platform only makes sense at scale. It needs a steady flow of new students, and I had no real plan for where they would come from. My tutoring had grown on referrals, one family telling another. That works for a tutor with a limited number of hours. It doesn’t fill a platform.
Meanwhile the costs didn’t wait. Servers and the other running costs kept rising whether students came or not.
What I did
- Went looking for users. I contacted high school clubs, offered a free first month, and added a referral program so existing students had a reason to bring friends. None of it produced the steady traffic the platform needed.
- Looked at it honestly. Revenue wasn’t covering costs, and I couldn’t see what would change that.
- Shut it down. I closed the platform and took the loss.
The product
- Yes: Lesson content students asked for
- Yes: A clear sense of progress
- Yes: Simple access
- Yes: Weekly releases and anonymous feedback
Getting it to people
- No: A plan for where new students would come from
- No: Steady traffic beyond referrals
- No: Revenue that covered servers and running costs
What I learned
At the time, I thought growth meant making the product better. I eventually learned that a good product without distribution is often just a good product that nobody sees.
I had put almost all of my effort into the part I knew how to improve, and almost none into the part I didn’t understand: how people find a product in the first place, what it costs to reach them, and what it costs to keep something running while you wait.
Two things survived. One was a way of working: smallest useful version, learn what the next problem needs, test it on real cases, repeat. I used the same loop three years later when I built operating systems at JuicyBite with no engineering background.
The other was a set of questions. At Bondit, the first things I looked at were the sign-up funnel and what it cost to get each user. At JuicyBite, it was distribution: which stores, which channels, who decides. Those weren’t new interests. They were this failure, a few years later.