Transitioning out of M-KOPA
Friday was my last day at M-KOPA. I’m going to miss the data science folk, the larger analytics organization and all the great PMs and engineers I got to work with. I’m also going to miss the data with all its odd quirks and hidden stories. There were four largeish themes of my career at M-KOPA.
A big part of the work I did in the early days was to help set up our pricing experimentation infrastructure. Large scale matched, clustered field experiments. Dosage response type experiments. Experiments where there was no way to avoid selection. There was even a cheeky multi armed bandit jankily put together to optimize marketing SMS (most definitely overkill). It was so many of my dreams come true.
Once that was sorta put together, I focused on building out continuous cash forecasts. There were multiple use cases for this, the most important of which were to estimate losses. So it was really, really important to get it right. Man, was this a journey. Figuring out a good representation of the financial lives of our customers to be able to model cash flows took months. But when we finally did crack it, it was magical 🪄. It worked well enough to be used by finance for loss accounting, sales for optimizing stock allocation, credit for pricing etc. One solid set of models can go surprisingly far.
The third theme that’s sort of interwoven through all my time at M-KOPA was just fun rabbit-hole analytics. These were bursts of focused energy on specific problems that resulted in analytics products. Fraud, COVID (yeah we did actually create an index to measure how COVID was impacting us), measuring agent performance, lead metrics for different commercial goals. A fun mix of problems and statistical techniques.
Finally, I held positions of leadership within the data organization. This included people and technical leadership. Hiring, growing teams and building ML infrastructure to play a tiny part in M-KOPA’s incredible growth story. This is where I learned the most from the excellent engineering leadership at M-KOPA. I stood on the shoulders of our staff engineers even as I learned to be one. I’m so grateful to have had access to them.
I’m very excited for M-KOPA’s future and can’t wait to hear all about what the data science folk continue to build.
What’s next for me? I’m at the Addis airport on my way to Senegal. I’m very caffeinated and super pumped to learn about a new product in markets I haven’t worked in before. It’s a product I’ve admired for some time now so I feel incredibly lucky to have an opportunity to contribute to it. More on that soon!