Choosing an ERP partner is harder than choosing software, because the software is documented and the partner is not. These ten questions cost you nothing and they separate the engineers from the resellers quickly. Ask them in the first conversation, before anybody sends a proposal.
1. Who exactly will build this?
Not the company, the person. Ask for the name and the role, and ask whether that person is in the meeting. If the people who sell and the people who build never meet, your requirements will lose something in the handover.
2. What will you tell me not to do?
A partner who agrees with everything has not understood your business or does not intend to argue with you. You want somebody who says that a requirement is expensive and worth little, and explains why.
3. How do you deliver: in one go or in slices?
Big bang go lives fail loudly. Ask how the project is cut into pieces, which piece goes live first and what value it has on its own. If there is no answer, the risk sits entirely with you.
4. Where does the code live and who owns it?
The answer should be: in a repository you have access to, and you own it. If the customisations live only on the partner's server, changing partner later means paying twice.
5. What happens to my data if we stop working together?
Ask for the exit path in concrete terms: database dump, documentation, handover to another team. A good partner has thought about this and is not offended by the question.
6. Which version do you put me on, and why?
Version choice has consequences for migration, for available modules and for how long you can stay before an upgrade. A shrug here is a bad sign.
7. How do you handle updates and migrations?
Custom code plus new versions equals work. Ask how they keep customisations upgradeable, how often they upgrade, and what an upgrade typically costs. Anyone who says it is free has not done one.
8. What does testing look like on your side?
Ask whether there are automated tests for the custom parts and whether there is a test environment separate from production. Manual clicking is not testing, it is hoping.
9. Who do I call when the warehouse stops at seven in the morning?
Ask for the reachable channel, the hours and the realistic response time. A single engineer with a clear rule beats a hotline that opens a ticket you never hear about again.
10. Show me something you built that failed and what you changed.
The most revealing question of the ten. Everyone has a project that went sideways. A partner who cannot name one is either very young or not honest, and both should worry you.
Two more that only owners ask
- What is the running cost after go live, in numbers, for the next three years?
- Which of my processes would you leave outside the system, and why?
You do not need perfect answers to all twelve. You need answers that are specific, that admit limits and that survive a follow up question. Vagueness in a first call becomes a change request in month five.