Modern Operations on Top of a Legacy ERP, Without Rip-and-Replace

Orders, customers, items, and addresses all live inside the legacy ERP, which we can get to over ODBC. The data is authoritative, but reading it straight into modern interfaces every time someone clicks something is slow and chatty.

Get a clean, always-current copy of the orders somewhere the rest of the app can actually work with them.

A background worker pulls new and changed orders every thirty seconds, writes them as Order and Shipment records into a modern SQL Server schema, and keeps customer and address data in sync alongside them.

The legacy ERP is still the system of record. The ops app reads from a modern layer that's never more than a few seconds behind it.

Every retailer has its own chargeback rulebook. Freight terms have to match carrier types. Hazmat can't go by air. PO Boxes need a specific carrier. Account numbers have format rules. Some customers want their items pre-configured for label-pack quantities. The list is long and it looks a little different for every retailer.

Catch the violations before the shipment leaves the building, not weeks later when a debit lands.

A shipment validator runs more than thirty checks against every shipment. Things like address completeness, ship and tender date math, customer ship-window rules, freight terms vs. carrier compatibility, hazmat pairing, account-number format, and customer item setup. Every error has a stable code, and the validator re-runs whenever anything on a shipment changes.

The exception queue isn't a guessing game anymore. Anything the rules can catch shows up with a clear code, a message, and the shipment it belongs to.

"When does this shipment have to leave?" depends on who's getting it. Some retailers measure the window backwards from a fixed delivery date. Some let the math run through weekends. Some have their own named tenders, like Target's two-day, Walmart's one-day, and Walgreens' rules. And every window still has to respect UPS transit times and the holiday calendar.

Figure out first ship, last ship, first tender, and last tender dates for every shipment, using whichever customer case applies and the right calendar.

A ship-window calculator encodes all of those customer cases and the per-customer carve-outs. It takes the required date, applies whichever math fits, counts UPS business days, skips holidays and weekends, and hands back four dates: first ship, last ship, first tender, last tender.

The ship windows in the ops app line up with the ones each retailer actually enforces. So when someone asks "is this at risk?" the app gives the same answer the retailer would.

Multiple operators work the same queue at the same time, and a list that's even a refresh out of date is a list where work gets done twice or not at all.

Make sure every operator's looking at the same view at the same time.

The shipment grid is built on ag-Grid with custom validation badges and filtering. A SignalR hub pushes every change (new shipment, validation update, status change) out to every connected client as it happens.

Everyone's looking at the same thing. The race conditions and double-work that come with legacy ERP screens mostly go away.

Not every rule applies the same way to every customer. Some retailers explicitly allow things that the generic rule would flag.

Quiet down the noise the team's already decided is fine, without softening the rules for everyone else.

A per-customer "acceptable errors" config lets the team mark specific error codes as fine for specific customers. The validator still runs every check; the filtered output just leaves out the ones the team has already accepted.

The exception queue stays focused, and operators aren't dismissing the same false positive a hundred times a day.