The weekly report is the worst use of your team’s time
Someone opens 5 dashboards, copies numbers, puts them in a PowerPoint, writes the narrative, sends an email. Every week, 3-4 hours. Your best analyst doing scraper work. And to top it off, half the team skims the report.
Automatic reports generate that same report — numbers + narrative + recommendations — in 5 minutes, ready to send. Your analyst does real analysis; the report is put together by a machine.
How we built it
1. Data sources
GA4, Meta Ads, Google Ads, HubSpot, Stripe, Shopify, Mixpanel, your own DB. We connect the sources you already use via API. No migrating data to another warehouse — we read it where it is.
2. Templates with narrative
We design the ideal report with you: which metrics, which comparisons (week vs previous, vs last month, vs goal), how much detail, what tone in the narrative. The report comes out written as your analyst would write it, not as a dump of numbers.
3. Generation with LLM
Every week (or whichever period applies), a Cron runs the pipeline: data pull, calculations, narrative generation with Claude, PDF or Notion document assembly. The human validates; the system produces.
4. Automatic distribution
The report reaches whoever needs to receive it: email to managers, Notion to the team, Slack to a specific channel. PDF, HTML, both. Without anyone having to forward it.
Typical report types
- Weekly marketing: traffic, leads, CAC, ROAS, top creatives.
- Monthly executive: revenue, churn, NPS, milestones, risks.
- Daily ops: tickets, calls, day’s conversions.
- Campaign post-mortem: campaign close with ordered learnings.
- Investor update: for investors or board, with strategic narrative.
What changes when this works
- Your team recovers 3-4h/week per person involved in reporting.
- Reports always arrive on time — not “when someone has a spare moment”.
- Quality improves — more data, more rigorous narrative, comparisons that were never done before out of sheer effort.
- Decisions are made sooner because the information arrives sooner.
When we do NOT recommend this
- If every report you produce is a one-off investigative piece: there is no template to automate.
- If your team is 2–3 people and everyone looks at the same numbers in Slack: the formal report is unnecessary.
- If your sources are custom systems with no API: the integration is not worth it.
Stack we use
- Claude 4.5 for narrative generation.
- Python pipeline with Pandas for calculations.
- Direct API to GA4 / Meta / Google Ads / Stripe etc.
- Notion API + ReportLab for delivery.
We start with a diagnostic session
Before quoting for the full setup, we run a 90-minute session. We look together at your data sources, your current reports, and the audience that reads them, and we come away with an honest recommendation: whether this service fits where you are right now, or whether it makes more sense to start with something else.
We don’t charge for that session. If you’re interested, get in touch.