The systems integrator pitches you a migration. NetSuite to SAP, or SAP to Dynamics, or NetSuite to a newer NetSuite tenant with the modules you should have bought the first time. The slide deck says 12 months. The project plan says 18. The actual go-live happens at month 22, with a launch team that has been on the project so long they have stopped reading the original requirements. By month 28, the ROI committee is reviewing why most of the promised benefits did not show up and the consulting bill is into the second million.
The reason migrations underperform is not that ERPs are bad. It is that the surface area of an ERP is enormous, and migrations require getting all of it right at once. Every customization, every report, every integration, every report sub-template, every workflow, every approval matrix has to be rebuilt or remapped, with stakeholders who only have time to validate at the end. The risk and the cost compound until the math no longer pencils.
There is a third option between "stay frozen on the old version" and "migrate to a new platform." Wrap and extend.