A Slovak technology news portal replatformed: Core Web Vitals pass rate from 43% to 93%, bounce rate from 83% to 27%, and 1.3 to 2.3 page views per session.
TOUCHIT is one of the largest technology news portals in Slovakia, and the brief was not the usual one. They were rebuilding the site from the ground up, and they wanted the technical side of that rebuild owned rather than reviewed at the end.
So I supervised the third-party contractor and read what they delivered, commit by commit. At this traffic that is not pedantry. One uncached template or one query nobody thinks about is the difference between a page that renders and an origin that falls over at peak.
Performance is decided before anyone writes a template. Wait until development is finished and you are negotiating changes against a budget that has already been spent.
So I sat with the design team from the first week. Every major decision had to survive the same question: what does this cost at first render, and what does it cost on a phone on a weekday evening. That is a much cheaper conversation to have over a mockup than over a staging build.
The share of visits with good Core Web Vitals went from 43% to 93%.
LCP was the metric failing, and TTFB was why. The server took long enough to answer that nothing downstream could recover it: no image format, no lazy-loading, no critical CSS. So the work did not start on the front end. We rewrote most of the codebase and moved the infrastructure underneath it.
The first graph is the old site against the new one. Some countries still show weaker scores, which has no bearing here. The portal publishes Slovak content for Slovak readers.
The second covers the two weeks after migration. The wobble in "Needs Improvement" is me tuning CDN settings in production, which is the only place CDN settings are genuinely tested.
Two numbers moved, and both are numbers this business is actually run on.
Before migration it sat around 83%. The day after launch it was under 35%. It has held at roughly 27% since.
The articles did not change that day — same writers, same stories. The design, the templates and the stack underneath all changed at once, so I would not put the whole drop on speed. What I can say is that TTFB and LCP were the constraint we removed, and the number moved the day the new stack went live.
The portal is ad-funded, so page views are revenue. Readers were viewing 1.3 pages per session before the migration. Immediately after launch it jumped to 3.1, which was people exploring a new design rather than a number to plan against. It settled at 2.3 pages, a 77% increase on where it started.
The last year has not been kind to publishers. AI Overviews and the wider change in how results are assembled have taken organic traffic off the whole sector, and TOUCHIT sits in that sector.
What has held is everything downstream of the technical work. Readers stay longer, engagement has not regressed since launch, and organic traffic flattened out instead of continuing to fall. In a year like this one, flat is the result.
Enterprise Technical SEO Consultant
I am a developer-led enterprise technical SEO consultant. 10+ years building the web before I started fixing it, and I still ship production code today. I read the responses your site actually sends, tell you which findings are worth a sprint, and take the fix to your developers myself.
One specific problem, taken apart, with a position on what to do about it — argued against the primary documentation and against what I see in audits. Then three stories from the last two weeks, picked and explained.
A new episode every two weeks. Unsubscribe in one click.
