How to Shortlist ERP Vendors: A Simple Scoring Model

From noise to a decision

After the RFP, you will have five or six vendors and a wall of features, each presented in the most flattering light. A weighted scoring model converts that noise into a choice your leadership can defend and your team can trust. The point is not false precision; it is discipline. When two products feel equally good, the model forces you to state which differences actually matter and by how much, which is exactly the conversation that prevents a decision made on who gave the best lunch or told the smoothest story in the room.

Pick 5–6 weighted criteria

Start with a small set of criteria so the weights stay meaningful. A common starting mix is functional fit at 30 percent, total cost of ownership at 25 percent, ease of use at 15 percent, implementation risk at 15 percent, and vendor viability at 15 percent. Adjust to your context — a fast-growing company may raise scalability, while a regulated one may raise audit and compliance. The key is that the weights reflect your strategy, not the vendor's strengths, so the result answers your question rather than theirs.

Define each criterion before scoring

Vague criteria produce vague scores. Write what good looks like for each: functional fit means must-haves met natively; ease of use means a new clerk is productive within a day; vendor viability means profitable and with a clear roadmap. Shared definitions keep every evaluator scoring the same thing, which is what makes the total credible to a CFO who will otherwise suspect the number was engineered to justify a decision already made in someone's head.

Score 1–5 against your must-haves

Score each vendor on each criterion using evidence from the scripted demo and reference calls, never the sales pitch. Multiply the score by the weight and sum across criteria. Set a hard rule that anything below three out of five on a must-have is eliminated regardless of price — a cheap system that cannot do the one thing you bought it for is the most expensive system you can choose. Document the evidence next to each score so the result is auditable later when someone asks why a familiar brand lost.

Keep a shortlist of 2–3

Never negotiate with a single vendor; without competition you lose all leverage on price, terms, and scope. Two is the minimum, three keeps the field honest and gives you a fallback if a finalist stumbles. Run finalist demos of your hardest process before issuing the contract, and use the score gap between first and second as your negotiating position rather than a gut feeling you cannot explain to the person approving the spend.

Watch for score inflation

Teams tend to rate familiar brands higher. Counter this by anchoring every score to a specific piece of evidence — a demo moment, a reference quote, a contract clause — rather than a reputation. If a score cannot cite evidence, it should default downward, because a number with no backing is just optimism wearing the costume of analysis and will not survive contact with a skeptical board member.

A worked example

Suppose Vendor A scores 4 on fit, 3 on cost, 4 on usability, 3 on risk, and 5 on viability; Vendor B scores 5, 4, 3, 4, and 4. With the weights above, A totals 3.75 and B totals 4.05 — B wins, but the closeness tells you cost and usability are where the negotiation should focus. That is the model earning its keep: it shows you not just who wins, but why, and where to push, so the conversation with the sales team is about the gap that matters rather than about features you already agreed were equal.

Common mistakes that break the model

Two errors quietly ruin scoring exercises. The first is too many criteria, which dilutes the weights until everything scores about the same and the model outputs a tie. The second is scoring after the contract is mentally signed, when every weakness gets a generous pass. Guard against both by keeping criteria few and by locking scores before the finalist negotiations begin, so the negotiation tunes the deal rather than rewrites the verdict after the fact.

When to revisit the model

A scorecard is not carved in stone. If a finalist's reference calls surface a serious delivery problem, it is correct to lower their implementation-risk score and let the ranking move — that is the model working, not failing. What you should not do is quietly swap weights after seeing results to protect a preferred vendor. Keep the original weights visible in the appendix so any change is a recorded decision with a stated reason, not a silent adjustment that undermines the whole exercise and the trust your team placed in it.

Make the shortlist defensible

Present the scorecard with its evidence to the steering committee before anyone signs. A model built on your own requirements and demonstrable proof survives scrutiny far better than a recommendation built on enthusiasm, and it gives the whole organization a shared language for the decision, which pays off later when the inevitable question arrives about whether the right system was chosen.