Purchase order applications are usually evaluated by finance people sitting at desks, which tests the least important surface. The step that determines whether the whole cycle works is performed at a loading bay by somebody with no finance training and no patience, and that is where an evaluation should concentrate.
The test that matters
Hand the application to somebody who does not work in finance and ask them to record a receipt for four of ten items on one line of a long order. Count the taps and the seconds. If it takes more than a few of each, it will be batched, done from memory later, or not done at all, and the match will fail downstream.
Offline, because signal is not guaranteed
Loading bays, basements and rural sites are common dead spots. An application that requires connectivity at the moment of receipt will simply not be used there. Ask whether capture works offline and syncs afterwards, and test it rather than accepting the answer, because this is a frequently claimed and inconsistently implemented capability.
What else to capture at the same moment
A photograph of the signed delivery note, particularly where short deliveries and substitutions are noted by hand. That annotation is evidence that otherwise exists only on a piece of paper somebody will lose, and it is what makes a later credit note conversation short rather than speculative.
Questions people ask about purchase order application
Do delivery staff need training?
They should not. If receipting needs training, it will not happen reliably, which is a selection criterion rather than a training problem.
Should they have full accounts?
A limited role that can only receipt against orders. Simplicity here directly improves compliance at the step you most need.
What about barcode scanning?
Useful where items are labelled, and it removes the order-finding step, which is often the slowest part.