Solution Story · Process Automation

The end of double data entry

Almost every window is custom-sized — which meant every window order was keyed twice: once into the internal system, once into the manufacturer's. At 500–1,000 orders a day, that's not a workflow. That's a typing pool. We replaced it with an API and a set of exception queues.

The problem

Custom products, duplicated keystrokes, invisible errors

Windows can't be picked off a shelf — nearly every one is made to measure. The branch keyed the order into the internal system, then keyed the same order into the manufacturer's system. Two entries meant twice the work and twice the typos — and errors surfaced in the worst possible way: when the wrong window physically arrived at a branch, weeks later, with no way to trace what went wrong.

Before: if an order was wrong, they found out when the wrong window showed up.
Error handling by delivery truck is the most expensive kind
What we built

Key the intent once — integrate everything else

The branch creates only the order header: which customer, which job. The platform talks to the manufacturer by API, waits for the order file to come back, and populates the lines — the hundred windows, the pricing, the margins — automatically.

Branch creates order header customer · job · PO Window interface API order to manufacturer receive order file back populate lines · pricing · margins Manufacturer builds the windows returns order file Complete order lines · pricing · margins no second keying Exception queues broken files · format changes · pending POs · size mismatches one-click fix → API pushes it back
Happy path fully automated; everything else lands in a named queue where a single expert team resolves it — without swivel-chairing between systems.

Order pending

Waiting on the manufacturer's file — visible, aging-tracked, never forgotten in an inbox.

File exceptions

The vendor's file is broken or the format changed — caught at ingestion, not at delivery.

Business rules

Validations stop bad orders early — wrong window sizes, missing approvals, orders over value thresholds needing sign-off.

The action column

Every error comes with its fix: update the PO, correct the line, push back via API. One click, both systems updated.

112 window guys became one expert team

Errors used to be solved one-off at whichever branch the wrong window arrived at. Pulling all exceptions into central queues let a small expert team handle every branch's problems — faster, consistently, and with the tooling to fix things in one click.

Errors caught before manufacturing

Validation moved from "when the truck arrives" to "when the order is placed." Wrong sizes and rule violations stop at the queue, not at the job site.

Branch-level catalogs

Each branch operates almost as its own company — so branches only see the window manufacturers they actually carry. Location-specific control runs through the whole platform.

Results

A thousand orders a day, one calm queue

771,438
manufacturer documents integrated
500–1,000
orders processed daily
keying — the header only
1
expert team instead of 112 improvisers
The happy path is easy to automate. The value is in what you do with everything that falls off it.
Why the exception queues, not the API, are the real product