Executive Operations Cockpit

Operating data was scattered across NAS folders and exported sheets. To understand yesterday's result or today's red flags, the owner had to ask repeatedly and reconcile multiple tables. Livestream GMV could be double-counted with paid orders, while after-sales, inventory, and private-domain teams used different definitions, making apparently precise numbers unreliable for decisions.

The product goal is not more metrics, but three decisions within five minutes: confirm yesterday's settlement, identify today's risks, and enter the issues that need action. Supporting materials, contracts, and assets stay in a resource center rather than competing with operating KPIs.

The home view has three layers. Settlement uses net public/private receipts, operating contribution profit, livestream performance, fulfillment rate, and refund rate. Alerts surface today's livestream issues, pending after-sales, inventory risks, and missing sources. Drill-down connects each issue to its responsible module, owner, deadline, and closure state.

Secondary views cover livestreams, order fulfillment and after-sales, WeChat private-domain operations, products and inventory, and supporting resources. A red flag can be traced to a session, order, SKU, or source status instead of ending at an unexplained overview chart.

Finance close data enters net private-domain receipts and contribution profit only after confirmation. Inventory without a seven-day sales base is not treated as infinite stock. Livestream sessions without ad spend are excluded from the ROI denominator. Every metric has a source, formula, and validity condition.

When orders, finance, or inventory are stale, the interface shows pending, awaiting sync, or awaiting confirmation, while retaining the last successful snapshot and failure reason. The principle is simple: admit uncertainty rather than let a false zero enter executive judgment.

On refresh, a local service reads only fixed NAS directories, cleans, aggregates, and deduplicates the data, then creates a desensitized local snapshot. Source files are never modified, renamed, or overwritten; if the NAS is unavailable or a format fails, the last successful result remains visible with the reason.

The dashboard is not exposed publicly, does not upload source spreadsheets, and never displays phone numbers, addresses, passwords, contract text, or attachment contents. In an operating system for a business, privacy and recovery are part of product trust, not back-office details.

I led product and experience: mapping the recurring questions of owners, operations, finance, supply chain, and private-domain teams; defining the home and drill-down architecture; and specifying net receipts, profit, ROI, refunds, fulfillment, inventory, data states, alert thresholds, and NAS mappings.

AI supported engineering implementation, documentation, rule checks, and iteration. I remained accountable for operating definitions, product priority, privacy boundaries, recovery behavior, and final acceptance.