ERP projects fail publicly and expensively, and the failure is usually attributed to the software. In our experience the sequencing decision matters more than the product choice.
The appeal of the big bang
Switching everything at once is superficially attractive. There is one cutover, one training effort, no period of running two systems, and a clean date to point at. It is also a decision to concentrate every risk in the project into a single weekend, with no room to learn between the first mistake and the rest.
What phasing actually costs
Phased rollout is not free. You run two systems in parallel for a period, which means interfaces between them, reconciliation work, and staff who must know both. This is genuine cost and it is the reason people avoid phasing.
It is almost always cheaper than the alternative, because the alternative is not a smooth big-bang, it is a big-bang plus a recovery period.
Sequencing by pain and by dependency
Start with the module causing the most operational pain, provided its dependencies allow. This is usually inventory or finance. An early win in a visibly painful area buys the political capital the rest of the programme will need.
Then follow the dependency graph rather than the org chart. Procurement needs inventory. Production needs both. Payroll can usually go whenever it likes, since it touches little else.
One site before all sites
Within each module, go live at one location first. Run it in parallel for two weeks. The problems found there are found once and fixed everywhere. The alternative is finding the same problem simultaneously at eleven locations with eleven separate escalations.
Master data is the real project
On every ERP engagement we have run, master data has taken longer than expected and mattered more than expected. Duplicate suppliers, inconsistent units, product records that disagree between systems. This work cannot be automated fully and cannot be delegated entirely to the implementation partner, because the decisions are business decisions.
Start it in week one. It will not be finished when you want it to be.
The measure of success
A good ERP go-live is uneventful. Nobody outside the project team remembers the date. That is not a lack of ambition, it is the whole objective: the business keeps running while its foundations are replaced underneath it.