Operations demo
Tracks what's on the clinic shelf against a par range, turns usage rates and lead times into a reorder list, and keeps a separate watch on the supplies that expire. Built by directing an AI assistant in plain English — one of a few small tools at nateeisenstein.com.
A short retrospective for stakeholders — spend, orders, and usage over the last 7 days. Updates as you order and use stock.
Starts from every item at its par min, filled to par max. Change the quantity or the pack size on any line and the totals update — set a weekly budget to see where you're over. Click an item name to open it.
| Item | Vendor | $ / unit | On hand | Par min–max | Days left | Lead | Status |
|---|
Every lot of a dated supply — vaccines, medications, test kits — soonest first. Anything within 30 days or already expired is flagged.
| Item | Qty | Lot expires | In | Status |
|---|
"Simulate a week" subtracts each item's weekly usage (perishables come off the nearest-to-expire lot first) so you can watch stock fall toward par. It's a stand-in for what a real system reads from the point of use.
Hand-counting a stockroom doesn't scale. A real version reads consumption from the point of use — barcode or RFID scans as items leave a PAR bin, the EHR logging supplies against a visit, or a smart cabinet for the controlled and high-value stock. Deliveries post from the distributor's advance ship notice, and a light cycle-count schedule corrects drift on the items that matter.
The trigger here is a fixed par min, which is the simple version. The more accurate one
is a reorder point that moves with demand: average daily use × lead time + safety
stock, where the safety stock is sized from how much demand and lead time actually
vary and the service level you want to hold. Order size can be a straight fill-to-max or a
proper economic order quantity that balances the cost of ordering against the cost of
holding.
Dated stock needs first-expired-first-out picking, alerts well before the expiry date so it can be used or returned, and a lot lookup for recalls. Forecasting has to be expiry-aware — ordering a six-month supply of something that expires in eight weeks just buys waste. The split in this demo (dated vs. not) is the start of that.
The recap here is built from an activity log — every order placed, shipment received, and unit used. In the real thing that log is the system of record, not a side effect: it feeds the stakeholder summary, reconciles against the distributor's invoices and the general ledger, and shows cost per visit or per provider once usage is tied to the EHR. A scheduled job would send it every Monday.
Par levels are per location and per room. A network view rolls up what's short across all of them, suggests transfers before purchases when one clinic is long and another is short, and separates centrally-held stock from what sits in each clinic.
Single file, nothing to run, no login. The par logic, the reorder math, and the expiry checks are what the real service would run — only the data source changes.