Short answer: Score a shortlist against your own processes before you see a single demo. Weight the criteria that are expensive to get wrong — entity structure, manufacturing depth, statutory compliance, and who will maintain customization — and treat licence price as one line among many rather than the headline. Most ERP failures trace back to a decision made on a demo, not to the product that was chosen.
Every demo works
That is what a demo is: a scripted path through the happy case, driven by someone who has run it four hundred times, on data shaped to fit. You will not learn anything from it that distinguishes one product from another, because they all clear that bar.
What separates them is the awkward part of your business. The pricing rule with three exceptions. The approval that has to route differently for one legal entity. The stock reservation policy your operations lead has been maintaining in a spreadsheet for six years because no system has ever quite handled it.
Score it before you look at it
Write the criteria down first, with weights, before any vendor is in the room. Ours usually looks something like this:
- Entity and currency structure. How many legal entities, how much consolidation, which currencies. This is where NetSuite earns its premium and where a single-entity distributor should refuse to pay it.
- Operational depth. Manufacturing, batch and lot traceability, variant matrices, project accounting. Products differ enormously here and it is the hardest thing to bolt on later.
- Statutory compliance. In India, GST with IRN and QR e-invoicing, e-way bills, TDS and GSTR. Some platforms have native localization; others need a certified connector and a partner who has actually shipped one.
- Customization model and who maintains it. Can your team read it? Will it survive the vendor's next release? An extension model that forces a rewrite every upgrade is a recurring cost disguised as a one-time build.
- Total cost over five years. Licences, hosting, implementation, the partner retainer and the internal headcount. Not the per-user monthly figure.
- People availability. Can you hire or contract someone who knows it, in your market, in a reasonable timeframe.
Then rate each product against each weighted criterion and look at the total. The exercise is worth doing even when you already know the answer, because it tells you what you are trading away.
The questions that actually separate products
Take three or four of your genuinely awkward processes and ask each vendor to demonstrate them, on your data, not theirs. Ask specifically:
- Show me this working, with these exceptions, without custom code.
- If it needs custom code, what does that code look like and who can maintain it?
- What breaks when you release the next version?
- Who at your partner has implemented this exact process before, and can I speak to that customer?
The answers to those four are worth more than an entire evaluation matrix built from feature checklists, because feature checklists are written by the people selling to you.
Readiness matters more than product
The most common finding when we run a selection is that the shortlist barely matters, because the organisation is not ready to implement anything yet. Master data is a mess, nobody owns the process definitions, and there is no internal person with the authority to decide when two departments disagree.
An ERP implementation makes those problems visible; it does not fix them. If that is where you are, the honest first project is a data cleanup and a process mapping exercise, and it will cost a fraction of what a failed implementation costs.
Why we run this as a paid engagement
We implement Odoo, NetSuite, Dynamics 365 Business Central, SAP Business One and ERPNext. That means we have no quota to fill on any of them, and the selection can genuinely go against the platform that would have been the largest implementation for us.
It is a paid engagement, deliberately. Free ERP selection advice is a sales process wearing a lab coat, and you can usually tell what it will conclude before it starts.