We deploy the same ERP codebase both ways, so we have no commercial stake in the answer. Here is how we actually help clients decide.
Where on-premise still wins
Shop-floor latency. If operators scan barcodes at a weighbridge or a packing line, every interaction is a round trip. On a marginal 4G link, 300ms becomes 3 seconds under load, and operators start batching entries on paper — which destroys your data quality.
Connectivity reality. Many industrial estates still run on a single ISP with no meaningful redundancy. If production halts when the line drops, that is not a software decision, it is a business continuity one.
Machine integration. PLCs, weighbridges and analysers usually speak to something on the same LAN. Bridging that to the cloud adds a gateway you now have to operate.
Where cloud clearly wins
Multi-site consolidation. Three plants and a head office reconciling nightly exports is a solved problem, and the solution is a single shared database.
Disaster recovery. A tested cloud DR plan costs a fraction of a second on-premise site, and far more businesses discover their backup was broken than discover their cloud region failed.
Predictable operations. If you do not have someone whose job includes patching servers, you do not have on-premise — you have an unpatched server.
The hybrid most manufacturers actually need
Transactional operations run locally at the plant; consolidated reporting, BI and the customer-facing portal run in the cloud, fed by change-data capture. Plant operations survive a link failure, and the board still gets one set of numbers.
The question that settles it
Ask: "If the internet drops for four hours, what stops?" If the answer is "dispatch and production", start on-premise for those modules. If it is "someone can't run a report", go cloud and move on.