Guide · Manufacturing
This question usually arrives after a bad week. A report was wrong, or an order was missed, and somebody said out loud that the system is the problem. Sometimes it is. Far more often the system is fine and three things around it are broken, and replacing the system would cost twenty times what fixing those three things would.
The licence is the smallest part and the part everyone focuses on. The real cost is in four places, and none of them appear on the vendor's quote.
Your data. Years of parts, customers, pricing and history have to move, and the old system's version of a customer is rarely the new system's version. Somebody has to decide what happens to every record that does not map cleanly. That somebody works for you, not for the vendor.
Your people. Everybody who touches the system relearns their job. For a few weeks they are slower and more error-prone, during which you are still shipping.
The undocumented rules. Every ERP that has been in place ten years has accumulated behaviour nobody wrote down — the code that means something specific, the field being used for a purpose it was never meant for, the report somebody built in 2016 that finance still relies on. These surface during cutover, one at a time.
The risk of a bad cutover. This is the one that should keep you up. A failed ERP migration at a small manufacturer is not an inconvenience, it is an existential event. You cannot ship what you cannot see.
It is not never. There are four situations where building around the existing system is the wrong advice and we would say so.
The vendor is gone, or support has ended. If nobody can patch it and it runs on an operating system that is no longer getting security updates, you are running on borrowed time and no amount of building around fixes that.
It cannot do something structural. If the system fundamentally cannot handle a second location, or lot tracking you now need for compliance, and it was never built to, that is architecture rather than configuration.
You have outgrown it by an order of magnitude. A system sized for five users and a thousand parts behaves differently at fifty users and eighty thousand.
Nobody can get data out at all. Rare, but real. If the data cannot be reached by any supported means, building around it is not possible, only reimplementing it.
Which, for most small manufacturers most of the time, it is. The signs are consistent.
The system holds the data correctly, and the complaints are all about getting things out of it, or about work happening in spreadsheets alongside it. Your people know it and are fast on it. It does the core job — orders, inventory, purchasing — adequately. And the specific failures you can name are three or four processes rather than the system itself.
If that describes you, the honest advice is to leave the ERP alone and fix the three processes. It will cost a fraction of a replacement, it can be done in weeks rather than a year, and if it turns out badly you have lost a small amount of money rather than a year of your business.
Write down the last ten things that went wrong and made somebody say the system is the problem.
Then mark each one: could this have been prevented by a different ERP, or was it a process that runs outside the ERP, or a report that took too long, or something one person knew and nobody else did?
Most lists come back with seven or eight in the second group. Those do not get fixed by replacement — they follow you into the new system, because they were never about the system. The one or two that genuinely are architectural tell you whether you have a real case.
If the list comes back mostly in the first group, you probably do need to replace it, and you should talk to someone who does implementations rather than to us.
In practice the common outcome is neither. You leave the ERP as the system of record, build the two or three things it does badly as small applications alongside it, and revisit the replacement question in three years when you have better information and less pain.
That is not a compromise or a delaying tactic. Keeping the system of record boring and stable while changing the things around it is how most well-run small manufacturers actually operate.
Do the ten-things test first; it costs nothing and it is clarifying. If you would rather have someone walk it with you, the audit is thirty minutes and produces a written, ranked list within five business days — including, where it applies, the honest answer that a particular thing is not worth fixing.
More guides: QuickBooks Desktop integration · ERP reporting · Quoting off spreadsheets
Write to us. Two or three sentences is plenty: what the task is, who does it, and how often.
We read and answer every message ourselves.
chris@edenmillsoftware.comBoth of these reach Chris directly. No form, no queue.
Book a free 30-minute Process Audit Email chris@edenmillsoftware.com Call 410-375-4340Two or three sentences is plenty: what the task is, who does it, and how often.
