Blog
Measured results and the arguments behind them. Two kinds of writing: what I found when I measured something, and what I think it means.
Writing
Flat for four months, then a third higher
A 14,000-product ecommerce catalog, reworked over six months. Traffic stayed flat for four, then climbed a third. What the shape of the line actually shows.
Own measurement · anonymized client project. Search Console, Web search, weekly, September 4, 2025 – August 16, 2026. Conversion rate from GA4, organic sessions, June–August 2026 against June–August 2025.
Clicks fell 14%. Rankings didn’t move.
A B2B SaaS site kept its impressions and its positions, and lost 14% of its clicks over six months. What that changed about which pages are worth improving.
Own measurement · Search Console + Bing Webmaster Tools, six-month window · anonymized client project
When site speed stopped being a technical detail
An ecommerce store where load time and completed purchases correlated at −0.81 — strong enough to move speed out of the backlog and into the revenue conversation.
Own measurement · GA4 → BigQuery, mobile sessions, anonymized client project
I built my own SERP tracker
Most SEO platforms are priced like a toolbox when the job needs one screwdriver. What it cost to build a focused rank tracker instead, and who shouldn’t.
AI search isn’t an SEO problem. The buyer journey moved.
B2B companies notice AI search as lost clicks. The bigger change is that the early part of the buying process now happens somewhere you can’t measure.
Next step
Want this run on your own numbers?
Reading about a method is one thing. If you’d rather see it applied to your site, send me your URL and one sentence about what isn’t working.