Today, I’m excited to announce a new project.

UserDetective is an easy-to-install churn prediction and customer retention tool for SaaS businesses. It uses the data science, machine learning and SaaS-specific skills I’ve developed over the past decade to help subscription businesses keep their customers.

The project is in the very earliest stages. I’m writing about it here so that I can describe what it’s like to build a SaaS product with machine learning at its core. As products like this will become more and more prevalent over time, I hope that these posts will add value to those considering a similar idea.

It might be silly to announce so unfinished a product on a blog (even one with very modest traffic like this), but over the past decade I’ve had a handful of product ideas that I haven’t followed through on.

They never stuck because they were all forced ideas. Dreamt up simply because I thought I needed to have a product business going.

This one is different for a couple of reasons:

  1. Other people are involved - I’m working with a developer and a marketing / growth expert. This is new. With every other product idea I’ve had, I’ve done everything alone. That can really strain your passion for a project.
  2. The idea arose organically, - a combination of a long-term interest, solid recent experience, and a good working relationship with the other people involved.
  3. I have (a few) readers - putting the idea out here like this might help keep me focussed and honest. I’ve adopted a kind of radical candour approach to this blog thus far and I hope to present the journey here in real-time rather than looking back with rose-tinted (or otherwise-tinted) glasses.

So far we have a landing page, a predictive model, the beginnings of a robust event tracking stack, and a logo. I’m going to cover the various decisions taken and technical challenges encountered as frequently as I can manage, but I’d love to hear any thoughts, feedback, or questions you have about building a SaaS business based on machine learning.