MVP, not POC — I left Excel

I had this project in my head for a long time. Excel was not enough to keep control of my investments. I wanted management, privacy, control and to step a little outside the hold Google Drive has today.

I use it to track assets, positions, and the history of buys and sells, all in one place, in a way that actually works for how I want to see it.

I built this for myself.

How I usually work

I start with a plan: requirements, a tentative stack, and the Excel I use today the one I want the system to replace. I hand that to the agent.

I usually ask for phases. That way I can review step by step and, if something does not fit, stop and sort it out. The cost is real: a mid-flight change forces you to reread the plan from that point forward. Sometimes the change was never in the plan. Sometimes it was, but in a later step. Either way, you have to look again. That does not only cost time. It also burns tokens and the memory of the session you are working in.

The experiment

This time I took another path. I reviewed the plan, made a few corrections, and decided to run it all at once. I wanted to see how it behaved if I did not keep cutting it into stages.

I went to get tea. I sat down to watch it work. Time passed, not much and the project was ready.

I ran docker compose up and watched the images build. I opened the browser and boom: there it was, what I had asked for.

Not everything came out the way I expected. Some UI features did not match what I had in mind. In others, the agent assumed and picked its best option. I had to review and ask for changes again.

The turn

I started with a routine inspection: how you navigate the site. Then I used each feature.

That is where the magic showed up again. I stopped thinking only as an engineer. I started thinking like a PO. I saw potential, possible features, changes I already wanted in. I was making a second cup of tea and thinking about how to improve the product what else would help someone using it even though, for now, the user was me.

POC vs MVP

For a long time I have heard that we should build a POC (Proof of Concept) to validate whether an idea, or a given technology, is viable. Almost always that idea comes from someone who makes the decisions. We question it, we ask why, and we end up building it because it is our job, period.

This time that person was me. I understood that role better: the one who usually asks us for “things.”

A POC reduces technical uncertainty: can this be done? An MVP (Minimum Viable Product) reduces value uncertainty: does this help?

I had no doubt the stack could be done. I had doubt I could leave Excel. With this MVP I found that a simple version covers most if not all of my needs.

For this kind of project, thinking in MVP rather than POC is more useful. Important features make the product more robust. They do not necessarily show the real value: how it solves for management, privacy, control.

What is next

Now other problems show up, and these are production problems: how to handle backups, where to store them, software quality, how to deploy this on one of my servers.

That does not invalidate what I already have. The experiment showed another way of working. It became clear that if I define things and keep clearing up doubts mine or the agent’s at any given point, we can work from a more robust plan, closer to the product we are building, with the acceptance criteria we actually expect.

I no longer need that spreadsheet. Right now I have a system that makes me happy, and it keeps growing as I need other features.