I came to ERP through accounting, not through code.
I read a trial balance before I read a database schema, and I think that is why the systems I build get used. A commerce degree, then Oracle E-Business Suite, then a year auditing hotel revenue line by line in Micros and Fidelio — which is where I learned that most reporting problems are process problems wearing a costume.
From there: hospital information systems, including Egypt’s national Universal Health Insurance programme, and Odoo from version 8 onward. Twenty-seven implementations later, the pattern has not changed. The software is rarely the constraint. What the business has never written down is.
Four steps, in this order, every time.
Listen before proposing
The first sessions are questions, not slides. Who touches this process, where does it stall, what gets re-keyed into a spreadsheet, and what does the month-end close actually look like.
Map what is, then what should be
As-is and to-be flows on paper, agreed with the people who run them. Most of the value shows up here — a process that nobody had written down end to end usually has two or three steps that exist for reasons nobody remembers.
Specify so there is no argument later
A BRD and a functional specification the business signs, with acceptance criteria written before build starts. Every gap is classified as configuration, customisation or process change, so the cost of each is a decision rather than a surprise.
Stay through go-live and after
Data migration, UAT, cut-over, key-user and end-user training, then hypercare. An ERP that goes live without the training is a licence renewal, not a system.