Your ERP did not fail. Your rollout did.
Most failed enterprise systems work correctly and are quietly abandoned. The failure is almost always adoption, and it is predictable months in advance.
- Published
- 28 May 2026
- Reading time
- 6 minutes
- Category
- Delivery
- Topics
- ERP, Change management, Adoption
A system nobody uses is indistinguishable from a system that does not exist. In my experience the technical build is rarely why an implementation fails — the software works, and the staff quietly keep using the register beside it.
The system was designed for the org chart, not the job
Software designed from a management description of the process encodes the process as management believes it works. The people doing it daily have twenty workarounds that exist for good reasons. If the system makes their job slower, they will keep the workaround and enter data twice — until they stop entering it at all.
This is why discovery means sitting with the person who does the work, not only the person who commissioned the software.
It demanded a habit change with no payoff for the user
Every system moves effort from one group to another. Data entry usually lands on the people with least incentive to care about the reports it feeds. If the supervisor entering attendance gets nothing back, entry quality drops within weeks — and the reports built on it become confidently wrong.
Whoever enters the data must get something from the system. Give them something worth opening it for, or the data quality decays and takes the reports with it.
Go-live was treated as the finish line
The first four weeks after launch produce more real requirements than the whole design phase, because now people are using it against real work. Projects that treat that period as scope creep end up with a system frozen at the moment it was least correct.
Nobody removed the old system
If the register still exists, it will be used. Migration means the old process actually stops — with a date, a plan and someone accountable. Running both is not caution; it is guaranteeing that neither has trustworthy data.