iy. / Notes & Experiments
A running log.
Write-ups from real projects — what worked, what didn't, and the data behind both. No filler, no recycled listicles.
8+posts
3categories
Lead scoring for service businesses.
Most lead scoring models are built for e-commerce. This one is for a service business with long sales cycles and offline conversions — the kind of thing that breaks standard funnels. Here's how I built it at Elithair using Python, BigQuery, and a bit of patience.
08 min read →2026.05.07Data AnalyticsServer-side tracking in a cookie-less world.GTM + Stape.io + 100% GDPR compliance. What the setup actually looks like in production.06 min2026.04.18MarketingROAS 18.85: dissecting a high-performing account.Not a template — an account-specific breakdown of what drove exceptional returns for a UK e-commerce client.10 min2026.03.29DS / MLPandas vs Polars: a practical comparison.I ran both on the same 3M-row dataset. The results were not what the benchmarks suggest.05 min2026.03.11Data AnalyticsAutomating multi-domain reporting with BigQuery.Forty-plus domains, one pipeline. How I reduced weekly report prep time from hours to minutes.07 min2026.02.22MarketingProgrammatic in 2025: what attribution shows.Three campaigns, three different attribution models. The differences were large enough to change budget allocation.08 min2026.02.04Data AnalyticsBuilding a GDPR-compliant analytics stack.What "compliance" actually means in practice, and the infrastructure decisions that followed.09 min2026.01.20DS / MLStreamlit dashboards for analytics teams.No frontend team, no waiting. A working internal dashboard in an afternoon — the template I now reuse.06 min