Chakshu

The problem

A consumer-goods account dispatched thousands of vehicles a week across seventy-odd plants. Field staff reported indents received, vehicles placed and why they fell short, in whatever format the day allowed. Detention charges depended on contract terms that lived in documents, not in the data. The weekly review deck was assembled by hand from all of it.

What I decided

Replace the dispatch spreadsheet with the smallest possible daily capture: nine fields per plant per day, from the person who actually knows, with the contract rules applied by the system.

The process I designed

1. Nine fields, no more. Indent received, vehicles placed, reasons when short, detentions. Everything else is derived or looked up.
2. Contracts as data. Per-customer detention terms are stored once; the charge is computed, not estimated.
3. A hierarchy that mirrors the field. Ground staff, coordinators, leadership, admin, mapped to the real reporting line from hiring manager to area to region.
4. Outputs people already use. Boards for leadership, a weekly deck as PNG, PDF or Excel, shared where the review already happens.
5. Built within free tiers. Session limits of 6,000 reads and 1,500 writes, offline persistence so a plant with bad signal still submits.

The product

Live. A role-based QA checklist across seven areas and a twelve-flow automated smoke test run before every release, which deploys on push through GitHub Actions. Multi-customer from day one; the first customer’s weekly decks are generated from its data.

Decisions I would defend

• Capture at the source. The hiring manager at the plant types the nine numbers. Nobody re-types them.
• Leadership reads one document. The day’s whole picture is one read, which keeps it free and fast.

What’s next

Inward materials capture (v1.1). A possible redesign of the earlier sales CRM so that Chakshu and the CRM live in one app.