Series: AI × Supply Chain build breakdown
Stack: Claude Code · Shopify · inventory database · Google Sheets
The problem
A month-end count sounds trivial until you do it for real: every SKU at landed cost, packaging and merch in, free stock split out, month-on-month so the accountant can see what moved. By hand it’s half a day, and it breaks easily.
What I built
A skill that pulls on-hand from the database, values everything at landed cost (production + shipping + packaging, to four decimals), splits sellable from non-sellable, lists the free stock from order data, and writes two clean tabs into the accountant’s sheet with the assumptions shown so she can check my working.
Where it failed
Every trap was a double-count. The inventory app keeps three rows per SKU (main, grab-and-go, hold/damaged); sum them and the whole count’s inflated. And one unit pool feeds the case packs, so count both singles and cases and you’ve booked the same stock twice. It only got trustworthy once I designed those out.
One rule I won’t bend: unknown costs stay blank, never €0. A €0 says I know the cost is nothing. A blank says go find it. Don’t lie to your accountant.
Leverage rating: 5 / 10
Kills a monthly half-day and the valuation errors that rode along with it. It feeds the accounts, so it earns the 5.
Related: this is fast and trustworthy because of the master-data layer the Shopify Ops Starter installs. Email me for the method.