Change management at enterprise scale
A go-live can pass every technical test and still fail because the people weren't ready. Why change management is where ERP value is won or lost, and how to run it at enterprise scale.
Executive summary
Find a sponsor who refuses exceptions in public, map whose day changes and by how much, train people in their own role on your data, and measure adoption in the system every week, by department, through go-live and past it.
- Finding 1
The checkbox version doesn't fail loudly. Adoption looks fine on paper while people route around the system, and you find out when the quarter closes wrong.
Figure 2 - Finding 2
Active sponsorship is the biggest predictor of adoption. One skeptical executive can sink a deployment the project team otherwise ran well. I've seen it happen.
Table 1 - Finding 3
Resistance is rarely where the org chart suggests. To find it, map who's affected and how much their actual day changes.
Figure 3 - Finding 4
Generic courseware teaches the product and leaves people to work out their own job. That gap is where the help desk volume comes from.
Table 2 - Finding 5
You can see adoption in the system. Watch it by department every week and act on the outliers while they're still outliers.
Figure 4
The report in four numbers
- 90%
- Training completion, marked green on the status report. It counts attendance.
- 1.5x
- The speed the go-live training video gets watched at, with the sound off.
- 1
- skeptical executive is enough to sink a deployment the project team otherwise ran well.
- 20%
- One department's transactions against expected volume. It's telling you something.
I've watched a go-live go off without a single technical defect and still fail. The integrations held, the data reconciled and the cutover ran on schedule. Two weeks later finance was closing the month in spreadsheets, the warehouse had gone back to its old workarounds and the help desk was drowning.
| Workstream | Measure | This period | Status |
|---|---|---|---|
| Integrations | Interface testing | Held | Green |
| Data conversion | Reconciliation | Reconciled | Green |
| Cutover | Schedule | On schedule | Green |
| Change management | Training completion | 90%a | Greenb |
| cHelp desk volume not on the report | |||
- aCounts attendance. It says nothing about whether anyone can do the work.
- bGreen because there's nothing to report. The workstream was bolted on at the end.
- cIf it's climbing at the same time as completion, the training didn't land.
The system worked and the people weren't ready. You can spend two years and tens of millions building exactly the right system, and if the people who have to use it on Monday don't trust it or don't know how, you've bought an expensive way to keep doing what you were already doing.
Most of what passes for change management wouldn't have saved that go-live: posters in the break room, a launch email from the CIO that nobody finishes reading, and a training video pushed out the week before go-live and watched at 1.5x with the sound off.
1The checkbox version fails quietly
When change management is a checkbox, adoption looks fine on paper while people route around the system in the background. You find out when the quarter closes wrong.
2Active sponsorship is the biggest predictor of adoption
Active means the sponsor shows up, uses the language, makes the hard calls and holds their own leadership accountable in their part of the business. A logo on a slide doesn't count.
Change has to cascade from the top. If a division head signals privately that the old way is fine for now, their whole organization hears it, and no amount of training reaches them. I've seen one skeptical executive sink a deployment the project team had otherwise run well.
Table 1When a sponsor says they're fully behind it, ask these
If the answers are vague, the sponsorship is decorative. Better to know that now than to find out during deployment.
3Segment people by how much their day changes
You can't prepare people you haven't bothered to understand. Real stakeholder analysis is unglamorous and it's most of the work: map who's affected, how much their actual day changes, and where the resistance sits, which is rarely where the org chart suggests.
High influence, small change
Brief them properly and recruit them. It's the cheapest advocacy you'll get.
High influence, big change
Decides the outcomeGive each group a named plan, keep the sponsor engaged, use role-based training and measure it weekly.
Low influence, small change
Inform them. This is where most programs over-invest.
Low influence, big change
Hands-on enablement and floor support. Quiet, numerous, and the first to go back to the workaround.
4Train people on their own job in your system
Generic vendor courseware teaches the product. It doesn't teach your job in your system, and the gap between those two is where the help desk volume comes from.
Table 2What to replace in the training plan
| Instead of | Do this |
|---|---|
| Role-based paths built on your configuration and your dataPeople learn the transaction they'll actually run, not the demo version of it. | |
| Layered delivery: awareness early, capability close in, reinforcement afterNothing survives a single exposure two weeks before a change of habit. | |
| A practical task completed in a sandbox, and scoredCompletion measures attendance. A completed transaction measures capability. | |
| A named super user per department, trained early and given standingPeople ask the person two desks away before they ever open a ticket. | |
| Job aids at the point of work, in the flow of the taskNobody reads a manual at 4pm on a Friday with a queue in front of them. |
5Measure adoption in the system, weekly, by department
The system shows you who's transacting, who isn't, and who does it once and then goes back. Watch it by department every week and act on the outliers while they're still outliers.
Transactions
Who is actually processing work in the system, by department, against expected volume.
Warning signA department at 20% is telling you something.
Workarounds
Shadow spreadsheets and old-system logins still in use.
Warning signLegacy read access still busy in month two means adoption isn't what the dashboard says.
Help desk mix
The share of tickets that are training rather than defect.
Warning signHigh and flat means the enablement didn't land.
In practice the technology is the bounded, testable part of the program: it passes the test or it doesn't. People aren't bounded that way, and they get exactly as much rigor as you decide to give them.
End of report
Put independent eyes on your program.
If you're betting tens of millions on an ERP program, a candid second opinion is the cheapest insurance you'll buy.