One profit per SKU across three marketplaces, reconciled to 0.00%

The task

A home-goods manufacturer sells through five seller accounts on three marketplaces and had no reliable profit per SKU: statements differ by platform, some operations settle in another month, ad spend and account fees never reached products. Excel reconciliation took days and still drifted. They needed one source of truth for profit, a demand forecast for purchasing, and a check on marketplace overcharges.

What we built

• One pipeline for all five accounts: 12 months of history, daily refresh, restart-safe jobs. Marketplace access is read-only — nothing in the seller accounts is ever changed.
• We reverse-engineered the platform's monthly statement logic and verified it against the client's own numbers, day by day and SKU by SKU. COGS comes in as a file from the client's ERP, as agreed.
• 7–30 day demand forecast: several models compete on a held-out period and the winner ships. Daily retraining, degradation monitoring, the acceptance metric computed automatically.
• Stock-out signals and a trend detector, shipped with backtests on history. Commission, logistics and storage are checked against the platform's tariff — a ready basis for claims.
• Role-based dashboard, an AI assistant over account data, alerts to Telegram and email, a weekly P&L. Acceptance on measurable KPIs, weekly demos, automated tests in CI.

Outcome

0.00%

variance vs the client's manual benchmark across units, revenue, COGS, profit and margin for a control SKU over a month; threshold was 1%

2.7×

more accurate than naive: weekly WAPE on SKUs driving 80% of revenue, 29.6% vs 81.4%. The 25% goal is open: history covers 66.9% of SKUs

1.18M rows

of sales across 5 accounts and 12 months processed; 10.7% were ordered and settled in different months

~7.5% of spend

in one account's monthly statement never reached the SKU report (no SKU attached). Found and included in total spend; reconciled down to one control line

More cases