Every month, 280 hours of senior time went into screenshot work. Strategists had stopped doing strategy.
The agency had 18 retainer clients, each on a monthly performance report — branded slides, narrative, recommendations. Each report took ~16 hours: pull data from 5 platforms, screenshot Looker dashboards, paste into Slides, write narrative, route for review, fix typos. Repeat × 18.
The pain was twofold: senior strategists were spending two full days a month on assembly per client, AND the report quality was inconsistent because tired humans miss anomalies. Recommendations were generic. Clients had started to notice.
We're a strategy agency that ships PowerPoint. That ratio is wrong.
One pipeline. Five connectors. Per-client narrative with anomaly detection.
We connected GA4, Meta Ads, LinkedIn Ads, Google Ads, and a custom client CDP into a single Python pipeline. The pipeline runs on the 1st of every month, computes month-over-month and YoY deltas, surfaces statistical anomalies (e.g. CPL drift > 2σ from trailing 90-day), and writes a structured narrative — but the narrative is a draft, not the final word.
Account managers review the draft, override or add commentary on the 3-5 things that need a human touch, approve in Slack. The slide assembly happens automatically against per-client branded templates. Total touch time per client: ~1.5 hours instead of 16.
Four weeks. Three clients first, then all 18.
We piloted with 3 clients first — the simplest, mid-complexity, and most-complex retainer. After two reporting cycles, we onboarded the remaining 15. Account managers moved from 'this is a glorified spell-checker' to 'I'd quit if you took this away' within two months.
Client-side reception was the surprise: clients explicitly noticed that recommendations were more specific and that anomalies were flagged earlier. Three clients increased their retainer in the quarter after the pilot.
After eight weeks live across all 18 retainers.
- Monthly report assembly time−89%16 h → 1.5 h per client
- Total agency hours / month−89%280 h → 30 h
- Anomaly catch rate+3.4×Statistical detection vs eyeballing
- Retainer expansion in pilot quarter—3 of 18 clients upgraded
- Time to first full-coverage run—3 weeks
Same backbone, expanded into mid-month performance reviews and account-strategy decks.
After the monthly reporting was stable, the agency commissioned phase 2: a mid-month performance check-in (lighter, faster) and an annual account strategy deck builder that pulls the last 12 months of monthly reports into a forward-looking strategy doc. Same pipeline, different consumers.