They paid for the old system, hired consultants for six months, and still everyone used spreadsheets on the side. That is the story a 50-person machinery parts importer told when they came to ERPNext Consulting. They had run the old “go-live” but the data was full of mistakes, staff hated it, and the finance team reconciled everything in Excel anyway. They were 11 months in, and the system had become an expensive way to store bad information.
How to know your ERP is actually failing
Nobody will announce it. You know it when you see people exporting data from the system just to re-enter it into their own spreadsheets. This importer had exactly that: eleven departments, eleven private spreadsheets, and zero trust in the shared records.
The ERPNext Consulting team’s first move was not to promise a better system. It was to sit in the importing office and watch how orders actually moved from supplier to customer.
- Ask your warehouse team what number they trust more: the system’s stock figure or the notebook on the desk. That tells you everything.
- Check how long month-end closing takes. If it is more than a week, the system is not your source of truth.
- Look at the last 5 invoices: how many times did someone manually edit a price or a date because the system got in the way?
The restart that worked
ERPNext Consulting cut the scope down to three things: purchase orders, stock, and sales invoices — nothing else for the first two months. They made the warehouse staff responsible for the stock counts and gave them the authority to correct errors on the spot. Within four weeks, the private notebooks started disappearing. Six months later, month-end closing went from nine days to three, and the managing director could finally see the gross margin on every single order.
What saved this project was humility
The biggest mistake on the first attempt was trying to automate everything at once. This time, the team took eight weeks just to fix one process the way everyone actually worked, and only then added the next.
If your company is in a similar mess, do not pay for another big launch. Pick the single pain point that hurts the most, run it properly for a month, and make that your proof of concept. That is how a failed project became a working one.