The pattern is remarkably consistent
The scope grew between the business case and the build, because every department saw the project as its chance to fix something. The vendor configured what it was told, which was a description of how the business works today rather than how it should work. The data was migrated late, dirty, and defended by nobody. Testing proved the software worked, not that the business could operate it. Training was a week of sessions in month eleven.
Go-live then arrives with the organisation unprepared, the workarounds begin, and the benefits case — which assumed adoption — never lands. Two years later the business is running a system it does not trust and has started buying bolt-ons to compensate.
None of that is a technology failure. Every item on the list is a decision that was not taken, or was taken by the wrong person.
Decision one and two: ownership
The business owns the process; IT owns the platform. If the process design is led by the technology function, the result is a system that works and a business that does not. Every process being configured needs a named business owner who can decide how it will run after go-live, and whose authority outranks departmental preference.
The second decision is what stays standard. Every system has core processes that should not be customised, and the value of the investment is largely in adopting them. A rule that says 'we will change the system only where the change is worth more than the cost of the change, including the cost of every future upgrade' is worth agreeing before configuration begins, in writing.
Decision three, four and five
Data: who owns each data set, who signs off that the migrated version is correct, and what happens to the records that are incomplete. Data migration is where projects discover that the business has been running on three versions of the customer list. Start it early and give it an owner with authority.
Definition of done: what has to be true for the project to be complete. Not 'system live' but 'invoices issued through the system, month-end closed in the system, and the legacy system switched off'. Without that definition, go-live becomes the end of the project and the beginning of an indefinite support phase.
Post-go-live accountability: who is responsible for adoption in the first ninety days, and what the measures are. Tickets raised, workarounds in use, proportion of transactions through the system, days to close the month. Somebody has to be watching those numbers, and it is not the vendor.
The commercial protections
Contract for phases, not for a monolith: a discovery and design phase, a build phase, a migration and test phase, and a go-live and stabilisation phase, each with deliverables and acceptance criteria you can actually assess.
Pay against milestones, not against time. Insist on written acceptance criteria before each phase begins, and keep a retention until stabilisation is complete. Make sure the contract says what happens if the system does not meet the specification — remedy, replacement, or exit — because at that point the relationship has already failed and the clause is all that is left.
Finally, plan the internal cost properly. The largest under-budgeted line in most implementations is the client's own people: the process owners, the testers, the trainers and the temporary cover while everyone is on the project. If that is not resourced at the start, the project will take the resource from operations, and both will suffer.