A cloud purchase order system changes where the software runs, not what it does. Judge it on behaviour and on how your data comes back out. The useful question is not where the software runs but what it does when nobody is watching it, and how easily your own records come back out of it. Hosting is a delivery detail; those two are the product.
Cloud mostly buys you access, not capability
A hosted purchase order system means approvers can act from anywhere and the record is not on one person's laptop. Both are real and neither is a capability the software gained. It is worth being clear about that, because cloud is often sold as though it changed what the product does rather than where it runs.
Ask where the data goes and how it comes back
Whose infrastructure holds your orders, in which country, and how you export them. For a purchase order record the export question is the sharper one: these documents are your evidence of what you agreed, and a record you cannot take with you is a record you do not really own.
Availability matters less here than in daily tools
A purchase order record consulted a few times a week can tolerate an outage far better than a till or a booking system. That is worth knowing because it means uptime guarantees should not dominate the decision, and paying a premium for them is usually paying for the wrong thing.
Questions people ask about cloud purchase order system
Is a cloud purchase order system better?
It is more accessible, which matters most for approvers who are rarely at a desk. It is not more capable by virtue of being hosted, and the decision should turn on behaviour and export terms.
What should we ask about data?
Where it is held, how you export requisitions and orders, on which plan, and what happens to it after you cancel. The export answer matters more than the hosting one.
How much uptime do we need?
Less than for a till or a booking system. A record consulted a few times a week tolerates an outage, so uptime guarantees should not dominate the choice.