For most of my career, reporting meant a Friday afternoon of copying numbers from Ads Manager into a spreadsheet, formatting it, and hoping the client didn’t ask a question the data couldn’t answer.
After eight years of that, I got tired enough to fix it.
The problem wasn’t the data
Every platform gives you the numbers. The problem was the assembly - pulling from Meta, TikTok, sometimes Google, normalizing different date formats, different metric names, different attribution windows. Then making it readable for someone who doesn’t live in these platforms.
The work wasn’t analysis. It was janitorial.
What I changed
I stopped treating reports as documents and started treating them as systems. Consistent templates. Named columns that never change. A single source of truth per client that gets updated, not recreated.
The actual insight layer - what’s working, what isn’t, what to do next - that’s what deserves the time. Not formatting.
The AI layer
Recently I’ve been experimenting with using LLMs to draft the narrative section of reports - the “so what” paragraph that most clients actually read. Feed it the numbers, get a first draft, edit for accuracy and tone. It’s not replacing the thinking. It’s replacing the blank page.
Still a work in progress. But the Friday afternoons are shorter now.