Many brands underestimate the effort involved in getting a proper ERP setup into the business, often because the sales process made it sound simpler than it is. The ERP goes live, and everyone is exhausted. The retro gets booked for a fortnight later and tends to turn into a victory lap or a blame session, sometimes both in the same hour.
Neither tells you much about the thing that matters now, which is how much work the business is doing to keep the new setup running smoothly. If that work isn’t written down and called out, nobody takes over improving it, and over a few months it becomes the normal way of doing things.
This one’s for whoever owns the business side of an ERP programme in the weeks after go-live. That’s usually a finance or operations lead who has been asked to run the retro, and who can see the gap between the project’s status report and what their team is doing every day to close the month.
Two questions in one
A retro built around delivery milestones can tell you whether you hit the date and how many tickets got closed. It can’t tell you whether finance trusts the numbers without heroics, or whether the team could repeat last week without phoning the implementation partner.
Most retros squash two separate calls into a single thumbs up or down. One is whether the project delivered what it said it would on time and in budget. The other is whether the business can now run accurately without the project team stitching it together each week. Treat them as one question, and you’ll either write off a decent phase one, or sign off an operating model that isn’t ready because the cutover happened on time.
A phase one can easily be written off as a failure when it did deliver its scope, because the business’s idea of what it needed moved on while it was being built. The retro has to separate those cases, or the size of the problem ends up depending on who is shouting loudest.
Mapping what works today
Start with what each department and channel can do end to end today. Use the core flows rather than system modules, so order to cash and period close come before anything labelled “retail live”.
Then sort every gap into the right category. Some are defects against committed scope, and some are temporary processes the team chose deliberately. The rest are either approved deferrals or new requirements that were never in the deal. Getting that split right stops the project being rewritten as a failure after the event, and it stops the business accepting a half-built process as finished.
Reconciliation and real cycles
A lot of retro time goes on integration dashboards. A green tick tells you a message moved between systems, and you still have to check the outcome happened somewhere else. A return can be receipted in the warehouse and still not refunded, and the connector screen will show nothing wrong.
Ask whether you can show that nothing is missing or wrong across the chain that turns trading activity into numbers, starting with orders and payments. That needs a reconciliation and control log the business owns, listing unmatched items with the money at stake and how long each has been open. Each one needs a route to resolution as well. If producing it means the project team pulling data at midnight, you’re not stable yet.
Review the first real operating cycles as well as the demo paths. The first supplier payment run and the first proper month-end close will tell you more than any test script.
A process that passed testing hasn’t been proven in live operation yet, and a quiet first month mostly means you haven’t hit the awkward corner.
The manual work
If a marketplace or wholesale channel went live before its integration did, someone is moving orders across by hand and reconciling whatever doesn’t match afterwards. You can cut scope at launch, and the accounting obligation still has to be met by somebody.
Write each workaround down as an entry in a control log you’ll keep using. Record what happens, how often and how many hours it takes, who it depends on and what the control is. Then give it a test for when it can stop.
A controlled manual step can be the right decision. The problem comes when nobody admits it exists or owns it, and over time it moves from systems into finance, or from operations into merchandising.
Check the data being created now
Migration data quality usually gets reviewed, but the retro also needs to check whether the data being created now is already going wrong again. If product data still starts life in a supplier spreadsheet and it’s wrong there, next month’s ERP reporting will be wrong too.
So find where the defects are created and who owns each field, then agree the validation and the point in the critical path where the data has to be stable.
Who can run it
Training plans tend to cover which buttons to press. Knowing company policy, and having the authority to make a call, sit with the business, so every core process needs an owner who can answer those questions once the project team steps back.
Check whether the permanent team can run the process and resolve problems without relying on project relationships. That needs named process owners and an escalation route people use. It also needs documentation someone can follow and cover for when people are off. Agree what hypercare exit looks like, and who picks things up in BAU after it.
The date and the money
Write down what the go-live date was tied to and what was traded off to hit it, including the risks taken knowingly, while people still remember the reasoning.
The cost side goes wider than the software and the statement of work. Add the headcount you brought in, and the internal hours going into reconciliation and workarounds that never appear on an invoice.
If there’s no credible baseline, say the benefits can’t be measured yet and agree how you’ll measure them from here. Alongside that, the business needs a realistic view of what’s left to do, and of which benefits only arrive once the manual work stops.
Every open item should leave the retro with a decision. Some get fixed, and some need time planned in before they can be. The rest are either closed with evidence or carried as a risk that someone has signed for.
Run the retro twice
It’s worth splitting this across two sessions. Run a review soon after cutover to capture what happened, including the controls that saved you and the things that got through on luck. Run a second after the first real operating cycles, once you have reconciliations and workload data, and use it to decide whether the project can step down and who owns the operating model from there.
If you’re the finance or operations lead running this retro, find the spreadsheet that makes the numbers work, and the person who knows how the CSV needs massaging before it loads. Both are normal at this stage. The retro is where someone decides who plans to take that work away.






"A green tick tells you a message moved" is the line I'll be stealing. Shipping has the same problem: tracking says delivered, but the freight invoice, the duty bill and the stock receipt still have to match before anyone knows what that container really cost. Do you see brands put landed cost inside the ERP scope, or does it usually stay a spreadsheet workaround?