ERPNext gets recommended for the wrong reason more often than any other platform we implement. The reason given is that it is free. The reason it actually works, when it works, is different, and so are the situations where it is the wrong choice.
The licence is not the saving
There is no per-user fee, and for a company with sixty warehouse and site users against ten office users, that arithmetic is significant. But implementation, data migration, training and hosting are unchanged, and they are the larger part of any ERP budget. Choosing ERPNext to save money and then under-funding the implementation produces the most expensive outcome available: a system nobody uses.
Where it is genuinely strong
Manufacturing, particularly bills of materials, multi-level production and subcontracting. Inventory across multiple warehouses. Project and task tracking tied to costs. And the underlying Frappe framework, which makes building a custom doctype or report a normal piece of work rather than a special project, if you have the technical capacity to use it.
Where it struggles
Localisation and statutory reporting outside its strongest markets, which means more configuration work for regional tax and audit requirements. Out-of-the-box reporting is thinner than the commercial platforms, so expect to build the reports management wants. And the partner ecosystem is smaller, which matters on the day you need a second opinion or a replacement implementer.
The ownership question
ERPNext rewards a company that wants control: your data, your server if you want it, your ability to change the system without asking a vendor for permission. It punishes a company that wants a system somebody else worries about. Neither answer is wrong, but the two are frequently confused during selection, and the confusion surfaces about a year in.
How we usually start
One module group that hurts today, usually inventory and purchasing, on real data, with a named internal owner trained to configure it. Finance follows once the stock numbers are trusted, because an accounting module fed by unreliable stock data produces confident nonsense.