The questions that produce different answers from different vendors — and the ones that do not.
An ERP request for proposal should establish scope, ask scenario-based functional questions, and require specifics on implementation approach, team, data migration, integration, support and commercials. The useful questions are the ones a vendor cannot answer yes to without committing to something checkable.
Most RFPs are long and produce near-identical responses, because they ask closed questions that invite a yes. A question like "does your system support batch traceability" discriminates between nobody. "Describe how a user traces every customer who received a specific raw material lot, and how long that takes" produces answers you can compare.
Your business, volumes, sites, users and the processes that are genuinely distinctive. Vendors cannot price what they cannot see.
Ten to fifteen real cases, each asking how the product handles it and whether in standard configuration.
Phases, duration, what they need from you, and what happens when something slips.
Who specifically will run it, their experience, and whether they are committed or indicative.
Price with its assumptions stated, change-request rates, and what the support agreement covers after go-live.
An RFP lists 612 requirements, each to be answered standard, configurable, customisation or not supported. All three vendors return over 95 percent in the first two categories.
The responses are worthless for discrimination, and predictably so. 'Configurable' covers everything from a checkbox to three weeks of work. Requirements phrased as features invite feature answers. And no vendor responding competitively will mark anything 'not supported' if any reading of the question allows otherwise.
A better instrument asks fewer questions that are harder to answer evasively: weighted requirements with the genuinely essential ones marked; scripted scenarios the vendor must demonstrate with your data; a request for the named implementation team with their relevant experience; references of your choosing rather than theirs; and total cost over five years on a template you supply so the figures are comparable. Thirty well-chosen questions separate vendors. Six hundred do not.
| Term | What it means |
|---|---|
| RFI | Request for information — an early, lighter enquiry to build a longlist. |
| RFP | Request for proposal — the detailed requirement and commercial request. |
| Scoring matrix | Weighted criteria applied consistently across responses. |
| Scripted scenario | A business case the vendor must demonstrate with your data. |
| Evaluation panel | The group scoring responses, which should include the people who will use the system. |
Every vendor answers yes. The matrix produces no discrimination and consumes weeks.
A capability delivered by development costs differently from one delivered by configuration. Ask explicitly.
The people in the proposal should be the people on the project, or the proposal is not about them.
A price without stated assumptions cannot be defended against a change request later.
| Metric | What it tells you | What distorts it |
|---|---|---|
| Requirements marked essential | Should be a minority. If most are essential, prioritisation has not happened. | |
| Scoring variance across the panel | Wide variance means the criteria are ambiguous, not that the panel disagrees. | |
| Scenario completion in demonstration | The hardest evidence available in a selection, and the least gameable. | |
| Cost comparability | Whether responses can actually be compared, which requires supplying the template. |
We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.