The CPG reporting demo
Harvest Crisp Foods, a brand that does not exist
Every number below is synthetic. The structure is real: it is what I would build for a frozen-food brand with several co-packers, national retail and a board that asks questions.
- OverviewAre we on plan this month, and where is the gap coming from?
- Sales & DepletionWhat shipped, what sold through by retailer, and what that means for weeks of supply on shelf.
- Production & SupplyWIP by co-packer and site, finished goods by DC, plan vs actual, and what is short next week.
- Data SourcesWhere each number came from, when it last loaded, and what failed validation.
- AI InsightsPlain-English questions on top of the numbers, plus the Slack alerts leadership would actually receive.
Two rules. Every number traces to a data source, a load time and a row. The AI layer sits on top of the reporting and never replaces it.
This demo is the pitch. The app I run for a bottled water brand is the proof.
Rebuilt here with synthetic data: weeks of supply by SKU, production runs, plan vs actual, P&L vs budget, a sync log. Same structure, made-up numbers.
How this maps to a real build
Five boxes, one line each
- 1
Partner CSV, PDF, email
Co-packers, distributors, SPINS and retailer portals keep sending what they send today.
- 2
Parsed and validated
Each file is read, checked against the schema, and every rejected row is logged with a reason.
- 3
Postgres warehouse
One database on infrastructure you own, with the history kept so trends are real.
- 4
Metabase self-serve
Dashboards the team can edit and extend behind company Google login, without waiting on me.
- 5
Slack and AI on top
Threshold alerts, a board pack that builds itself, and questions in plain English.