Counter began with a few questions about my own websites. The interesting part turned out to be deciding how much I really needed to know, and building something that stopped there.
Why did you decide to build it?
I wanted to know which games were being started, how often they were finished and which pages people were looking at. Fairly ordinary questions, really. I didn’t need to follow individual visitors around or bring a whole marketing setup into a couple of personal projects.
I’d used Google Analytics and Tag Manager before, but they felt like more than I wanted for this. Building my own tool gave me a chance to start with those questions and work backwards. Admittedly, building an analytics platform is quite a long way round wanting less configuration, but I was interested enough to find out where it would go.
How did you choose where to start?
I kept the first step narrow: could a site report that something had happened and have it recorded correctly? A game starting was enough to test that idea. The dashboard and the more interesting questions could come once that basic part worked.
I also wanted it to stay easy to add to. When I make another game or add a feature, I should be able to choose what is worth counting without setting everything up again somewhere else. Starting with one useful action gave me something I could check before building more around it.
Why was privacy so important to you?
Because the questions didn’t require me to know who anyone was. I wanted to understand how the things I’d made were being used, and I didn’t see a reason to collect more information just because I could.
So Counter doesn’t use analytics cookies or build visitor profiles. That means accepting some limits: ten game starts could be ten people, or one person coming back several times. I’m comfortable with that. Broad country totals give me a little context, but I don’t treat them as a reliable description of the person playing.
It still matters what each site sends, so I review that rather than relying on the word “privacy” in the project description. The details someone enters in my contact form, for example, have no business ending up in the analytics.
Why not keep all the information?
I couldn’t see much value in keeping every individual action indefinitely. Recent detail can help me look at what has happened around a change, but further back I’m usually interested in the overall pattern.
The approach is to turn older records into daily totals, with records becoming ready for that after 30 days. That isn’t a promise that every record disappears at an exact deadline; the process still needs checking, including what happens when something fails.
What mattered to me was making that a deliberate part of the project, rather than collecting everything now and leaving the question of what to do with it for later.
What did you want the dashboard to feel like?
I wanted to open it, ask a question and get somewhere without spending half my time arranging the screen. The first version involved too much moving between views and choosing the same filters again, which rather defeated the point of making a simpler tool.
I changed it into one workspace so I could choose a site and a time period, then narrow things down without losing that context. For example, I can look at game starts and then focus on a particular game, rather than begin again on a separate screen.
It’s a tool for my own use, with protected access rather than a public demo. That helped me concentrate on whether it was useful in practice instead of trying to make it serve every possible audience.
How did you keep the build manageable?
I treated it as something I’d need to look after alongside my other projects. Keeping running costs modest mattered, as did avoiding more infrastructure than I needed. I chose a setup that fitted the services I was already using, while leaving room to add the dashboard and other parts as the project developed.
Codex carried out a substantial amount of the implementation, testing and deployment work. My part was setting the requirements, working through the choices and reviewing what came back. Having something to use made that much easier: I could see where the dashboard was asking too much of me, or where a seemingly useful addition would collect information I hadn’t intended to keep.
That back and forth was more valuable than trying to write a perfect specification for the whole thing at the start.
Where is it now, and what have you taken from it?
I use Counter across KJ’s Games and this website. It gives me counts of the actions I’ve chosen to measure, including game activity and visits to the case studies, without needing a history of individual visitors. It isn’t a perfect record, and the numbers can’t tell me whether somebody enjoyed a game, but they give me something useful to investigate.
The part I’d carry into other projects is being willing to question the requirement as well as the solution. “We need analytics” could have become a very different project. Asking what I actually wanted to learn helped me decide what to build, what to leave out and when a working version still needed another look.