Streamlining Analytics, Migrating with AI Pt. 5
Post-migration, I knew some jobs were left untouched. Not because of the amount of work involved — again, I don’t think the workload has meaningfully increased. For the first project afterward, I wanted to focus on a part of the website I hadn’t touched in a while: analytics.
For some context, I had used Google Analytics on Mad Tea Party for years. When privacy compliance became an issue, Google was less than kind toward small websites or bloggers. At the time, paying for a CMP felt counterintuitive: every third-party option I found through Google’s list was paid, and it was difficult to find one that served customers outside Europe and North America. Designing a pop-up UI was the cherry on top of the cherry on top.
Simple Analytics sounded promising: a free hobby plan, no cookies, and European hosting, with a pitch that it knew European privacy rules. The 30-day history sounded restrictive even then, but the bargain was clear: modest statistics without a consent pop-up. It solved the immediate problem while adding another account, provider, and a few more paragraphs to maintain in the Privacy Policy.
After the broader migration, Cloudflare handled the domain, DNS, delivery, and security for the website. Having seen what Cloudflare offered, I began the side project by asking Codex for a candidate list that met my criteria. I expected third-party services that pulled Cloudflare data into their own analytics. Instead, Codex suggested Cloudflare’s in-house tool and found that it fit the common privacy requirements I was trying to cover: cookie-free analytics without another consent pop-up.
Planning the analytics change was less about migrating data than observing a handover: the old tool had to stop cleanly and the new one had to start cleanly. As before, execution was mostly automated. There were no surprises. I only had to check in with Codex as each waiting period ended and mark the next item off the checklist.
The visible gain was a streamlined dashboard and six times as much traffic history. The operational gain was one less account and a shorter explanation of what happens when someone visits the site. Cloudflare’s familiarity does not make it safer, but it does make the disclosure less alarming. On a blog, ‘I use Cloudflare, a well-known company, for delivery, security and analytics’ is easier to understand than either ‘I use GA4 with a Europe-based CMP’ or ‘I use a European analytics service that aims to replace Google Analytics without compromising privacy’. It’s a rather ridiculous proposition: introducing another service before the website has even introduced itself.
Part 4 ended with documentation as the hedge against technical debt to AI. This is one benefit of keeping AI in the loop. Well-written documentation for an active project serves as both a human interface and a token-saving measure. With that documentation in place, AI fast-forwarded a plan that had been sitting on the back burner: researching it, laying it out properly, then executing it. The result was not operational bloat; it removed a seam and streamlined the operation.