The Spaghetti Detective

The Spaghetti Detective was in beta testing phase for 8 months (a lot longer than I expected) before I finally felt I was ready to launch it and start charging my users.

The reason why it took 8 months to get it out of beta phase was that I didn't feel the Machine Learning model was good enough. The product was to alert users with 3D printing failures. But my users were still getting a lot of false alarms. As a matter of fact, at the point of public launch, my users were still getting more false alarms than real ones.

This made my feel "ashamed" about the product and had a very low confidence level that users would pay for the product because the accuracy still looks terrible. Then in October, I scheduled interviews with about 20 users to hear what they have to say. It turned out most of them don't mind false alarms. Some of them went as far as to say false alarms mean "the AI is trying to help me". That finally gave me the confidence to pull the product out of beta and start charging.

When I launched it, I sent to all beta testers to give them 25% discount but only if they use it within 3 days. I believe the extra incentive and sense of urgency worked well.

My product is subscription model. In a previous similar product, I made monthly subscription as the default option, and gave only 17% discount to the annual plan ($50/year vs $5/month). 80% of my users ended up with monthly subscription. The problem with that is the churn rate was pretty high. There were subscriptions being cancelled almost every day. For The Spaghetti Detective, I gave 30% discount to annual plan ($36/year vs $4.5/month). 90% of them took the annual subscription! Now I have 1 year when I don't have to worry about churns!

The reason why the revenue dollar amount after launch looked pretty good was mostly because 90% of the payments were for the whole year. MRR was only $250, a much modest level. Nevertheless, this launch was still a lot better than the launch of the previous product, for which the revenue of my first month was merely $50.

The beta product at the point of launch was just a very rudimentary Machine Learning model with pretty modest accuracy. I sent emails to a few thousands of users of my previous product, which was also 3D printing related. And I got more than 1000 signups in just a few days.