Profit Intelligence

Agricultural advisors had no systematic way to compare a dairy farm's financial performance against its peer group. Monthly data lived in Excel files — no benchmark, no anomaly detection, no narrative. Advisors spent hours manually interpreting raw numbers and still missed early warning signs in farms trending toward credit risk.

Built a platform that ingests monthly farm financial and production data, computes 45+ dairy-specific metrics across every cost category, and benchmarks each farm against a configurable peer group. Advisors get interactive dashboards across rolling 3-, 6-, 9-, and 12-month windows. A KPI highlights widget surfaces the four most decision-relevant numbers on every dashboard load. An AI layer lets consultants auto-generate written summaries grounded in actual farm data — with a prompt-injection classifier gate to keep the tool focused in a regulated advisory context.

• 01

  Storing metric formulas as versioned Python closures in the database lets domain formulas evolve independently of code deploys — the metric catalog becomes a first-class data artifact, not hard-coded logic.

• 02

  An audit gate before data promotion is more valuable than post-hoc anomaly alerts — catching a data entry error before it corrupts 12 months of benchmarks saves far more downstream trust.

• 03

  A prompt-injection classifier before every LLM call matters in regulated contexts — it kept the AI feature from being repurposed as a general chatbot by users who found the interface.

• 04

  Rolling window chart variants (R3, R6, R9, R12) are more useful than fixed-period views — advisors reason in trailing averages, not point-in-time snapshots.