The most common advice on choosing an ERP is "define your requirements first" — true, and not much use: if defining requirements were easy the problem wouldn't exist. The harder truth is that most of what separates one system from another shows up neither in the feature list nor in the demo, but in details you discover in month three.
Start from your hardest day, not from a feature list
Feature lists look alike because they're written to look alike. Instead, write down the three hardest processes you actually run — month-end close, the stock count, payroll week — step by step as they happen today, including the steps that happen outside any software.
Then ask every vendor to run those processes in front of you, in the system. Not a prepared demo — your process, with your data. The difference between vendors surfaces exactly there, and usually on the question "and what if we need to undo that step?"
The questions that reveal the most
- How do I correct an invoice that posted wrong last month? The answer reveals whether the system respects the principle that entries aren't deleted, or lets you edit history.
- What happens to stock and cost if I enter a back-dated purchase invoice? This is the question that separates a real accounting engine from an attractive interface.
- How do I get all my data out if I decide to leave? A confident vendor answers plainly. Hesitation here is itself the answer.
- Who can change prices and discounts, and is the change logged? Permissions look like an administrative detail right up until you need them.
- What happens when the internet drops at the point of sale? "It usually doesn't" is not an answer.
- Who trains my team, how long does it take, and who do I call in three months? The real cost of support appears after the launch enthusiasm wears off.
Warning signs worth stopping at
A price that only exists in a call
A price that appears only after you hand over your details usually means the price is set by what you look able to pay. Ask for the full first-year cost including setup, training and add-on modules.
"We'll build anything you need"
Unlimited customisation sounds like flexibility and is usually debt: every change built for you becomes something that must be maintained at every upgrade, and you're the only one paying to maintain it.
An accreditation that is never named
"Approved by the tax authority" is a claim that must be followed by a document name and a number. If it isn't, treat it as absent and ask directly.
No real trial
A demo with no trial account means you'll discover the system after paying. Any confident vendor lets you put your own data in yourself.
How to run a trial that tells you something
Most trials are wasted on browsing. A useful trial is narrow and unkind: pick one week, load real data — twenty real items, five real customers, real opening balances — then run that week's actual transactions through the system in parallel with your current method.
At the end of the week compare two numbers only: stock on hand and customer balances. If they differ from your books, the real question isn't "which is right" but "can the system explain the difference to me?" A system that can show you the trail is trustworthy. One that produces a number with no audit trail behind it is not.
Finally: involve the people who'll actually use it, not only the managers. A system the manager picks and the storekeeper hates will be defeated by a spreadsheet within two months.