Every ERP selection I have been part of starts the same way: a shortlist of three or four vendors, a round of polished demonstrations, and a room full of managers who leave each session slightly more impressed than the last. The demo is a performance, rehearsed hundreds of times, and it is designed to make the product look inevitable. The problem is that nothing in a demo tells you how the system will behave with your chart of accounts, your subcontractor payment cycles, or your month-end close.
Over twenty years and eight enterprise implementations, the selections that held up were the ones where the decision was made on paper before it was made in the room. Here is the framework I use.
Start from scenarios, not feature lists
The standard RFP approach — a spreadsheet with four hundred features and a Yes/No column — is close to useless. Every mature ERP will answer “yes” to ninety percent of it, and the answers tell you nothing about how painful each “yes” is in practice.
Instead, write ten to fifteen business scenarios drawn from your actual operations, in enough detail that they cannot be faked. In construction, mine usually include: a subcontractor progress payment with retention and advance-payment recovery; an inter-company material transfer between two projects; a variation order that changes both budget and schedule; and a month-end close with WIP recognition. Ask each vendor to run your scenarios, with your sample data, in a working environment — not slides.
Build the scorecard before the first meeting
The scorecard has to exist — with weights agreed by the steering committee — before any vendor walks in. If you build it afterwards, it becomes a justification document for the vendor people already liked. My typical weighting for a mid-size contractor:
- Functional fit to scenarios (30%) — scored per scenario, by the process owner who will live with the result.
- Local implementation capability (20%) — the partner's bench in your country, references you can call, and the CVs of the actual consultants (not the pre-sales team).
- Integration and openness (15%) — documented APIs, middleware options, and evidence of working integrations with the systems you already run.
- Total cost over five years (15%) — licenses, implementation, infrastructure, and the support contract. One-year pricing comparisons are how vendors hide cost.
- Localization and compliance (10%) — VAT, e-invoicing (ZATCA in the Saudi context), Arabic support, and payroll/GOSI where relevant.
- Vendor viability and roadmap (10%) — will this product still be invested in five years from now?
Score the partner harder than the product
Most ERP failures I have witnessed were not product failures; they were implementation-partner failures. Two questions expose more than any reference letter. First: name the consultants who will be on our project, and let us interview them. If the vendor resists, the A-team you met in pre-sales will not be the team that shows up. Second: give us two references where the project went badly, and tell us what you did. Every experienced partner has them; only honest ones will discuss them.
Run the money conversation in parallel, not last
Commercial negotiation left to the end compresses into two weeks of discounting theater. Instead, request full five-year cost models at the shortlist stage and normalize them yourself: same user counts, same modules, same support tier. The differences that emerge — one vendor licensing by named user, another by concurrent session; one including test environments, another charging for them — routinely swing total cost by 25–40% and are invisible in a headline license price.
The decision meeting
When the scoring is done, present the committee with the weighted result, the raw scores per scenario, and a one-page risk register per vendor. Then let the score carry the decision. The moment leadership overrides a completed scorecard for “strategic” reasons, everyone in the organization learns that the next selection will be political — and process owners stop investing effort in evaluations.
A note on timeboxing the selection
Selections that run past four or five months do not get better — they get political. Fatigue sets in, champions change jobs, and vendors refresh their pricing upward. A realistic calendar for a mid-size organization: two weeks to write scenarios and agree the weighted scorecard, three weeks for scenario-based sessions with the shortlist, two weeks for reference calls and consultant interviews run in parallel with commercial normalization, and one decision meeting. Ten weeks, end to end. Publish that calendar to the vendors on day one; serious partners respect a disciplined buyer, and the ones who try to stretch the process are telling you something about how they will run the implementation.
One final habit: keep the losing vendors' scorecards on file. Implementations occasionally fail, partners occasionally exit markets, and the organization that can reopen a documented evaluation is a year ahead of the one starting from a blank page.
Selection gut-check
- Did every vendor run our scenarios with our data in a working system?
- Have we interviewed the delivery consultants, not the pre-sales team?
- Is the cost comparison normalized over five years?
- Was the scorecard weighted and signed off before the demos began?
A disciplined selection does not guarantee a successful implementation — sponsorship, scope and change readiness decide that. But an undisciplined selection almost guarantees a difficult one, because you begin the hardest project in enterprise IT already married to the wrong partner.