Context
Most businesses run Google and Meta in separate tabs and reconcile them in a spreadsheet — which means nobody actually knows the blended cost of a lead. AdPulse is a privacy-safe recreation of a real build — the fix: one command center, every channel, and an analyst you can question in plain English.
The problem
Dashboards show what happened; owners need why and what now. The gap between a ROAS chart and a budget decision is exactly where agencies charge retainers — and where an embedded AI analyst earns its keep.
The approach
I built the unified performance surface: blended KPI cards (ROAS, spend split, cost per qualified lead), an 8-week spend-vs-conversions chart, and a campaign table ranked by return. Beside it sits the signature piece — an AI analyst panel that answers "why did cost per lead go up?" with a specific, campaign-level diagnosis and a reallocation recommendation.
What the build shows
The console reads 4.6x blended ROAS on $48.2K of tracked spend, identifies the exact campaign dragging efficiency (Lookalike prospecting, CPM +18%), and drafts the client-ready summary line — the full loop from data to decision to communication.
Recreated for confidentiality — a faithful, fully-working mimic of a real production build; client names, branding and data have been fictionalized.