Why selections drift, and what discipline buys
Most vendor selections produce a defensible-looking scorecard and a decision that was actually made somewhere else: in a conference demo six months earlier, in a colleague's recommendation, in a sponsor's preference nobody wanted to contest. The purpose of a disciplined selection is not paperwork. It is making the evaluation hard to steer, by fixing requirements, weights and scoring anchors before any vendor gets a hearing. This playbook covers the generic process for selecting significant vendors of software, equipment or services; the Before Selecting a Vendor guide covers the pre-commitment risks and pairs with it. Two principles run through every step. Comparability first: vendors answer the same questions, in the same format, scored by the same people against the same anchors. Evidence over assertion: a claim only scores when it has been demonstrated, referenced or contracted. Selections run this way take slightly longer to set up and considerably less time to defend, in front of the board and, occasionally, in front of a losing bidder's lawyers.
Step 1: Fix requirements while the field is still empty
Gather requirements from the people who will live with the vendor: operators, not only managers. Separate them ruthlessly into three classes: gates (fail one and the vendor is out, regardless of price), weighted requirements (the basis of scoring), and preferences (recorded, never scored). Keep gates few and genuinely fatal, such as regulatory compliance, data residency or a non-negotiable integration; every fake gate narrows the field for no benefit. Write each weighted requirement so that evidence can settle it. "Strong reporting" cannot be scored; "a named business user builds a defined report unassisted in the sandbox" can. Version-control the requirement set and freeze it when the RFP is issued, with any later change requiring steering-group sign-off and simultaneous communication to all bidders. This step is where selections are won or corrupted, because a requirement written after market contact is, more often than anyone admits, a requirement written by a vendor.
Step 2: Use an RFI to map the field, not to pick favourites
When the market is unfamiliar, run a short request for information before the RFP: ten questions, a two-week turnaround, no commitments. Use it to learn the market's shape: how vendors segment by size and sector, which delivery and pricing models exist, which capabilities are standard and which are roadmap slides. The RFI answers should improve your requirements, and it is far cheaper to discover a naive requirement now than during contract negotiation. Resist scoring RFI responses into a ranking. The document exists to educate the buyer, and treating it as round one pushes vendors into sales mode early. Cut the field to four to six credible RFP recipients using gates only, and record why each excluded vendor was excluded, one line each, because someone senior will ask later, usually after meeting one of them at an industry dinner. Keep the RFI answers on file: they are statements vendors made in writing before they knew the scoring model, and they resurface usefully during negotiation.
Step 3: Lock the evaluation model before proposals arrive
Build the scoring model while no vendor can benefit from its design. Set criterion weights with the Weighted Scoring Model in a workshop that includes the budget holder, the operational owner and procurement, and record the rationale for each weight. Have every evaluator declare vendor relationships before the model is agreed: prior employment, ongoing commercial contact, conference hospitality. A declared interest can be managed with panel design; an undeclared one surfaces later, usually from a losing bidder, and taints scores that were honestly given. Then define anchored scores, so that a mark means the same thing across evaluators.
| Score |
Anchor |
Evidence required |
| 0 |
Not available or not demonstrated |
No credible evidence offered |
| 1 |
Claimed, on the roadmap, or via a third party |
Vendor assertion only |
| 3 |
Demonstrated in a generic environment |
Standard demo or reference description |
| 5 |
Demonstrated against our scenario |
Scripted demo, sandbox task or contracted commitment |
Step 4: Run the RFP for comparability
Issue the RFP with the frozen requirements, the published model, a mandatory response format and a mandatory pricing schedule that forces like-for-like totals: licence or unit costs, implementation, support tiers, assumed volumes, indexation, and every charge that appears after year one. Publishing the weights and anchors to bidders does not weaken the buyer; it makes vendors spend their effort on the criteria you actually value instead of on relationship-building. Route all questions through a single channel with answers shared to every bidder. Score responses on the RFP Scoring Sheet, independently first, before any panel discussion, with each score initialled: the spread between evaluators is information, and a criterion where scores diverge widely has either a weak anchor or thin evidence. Convene the panel only after individual scoring to argue the divergences, and change scores only with a recorded reason. The output is a ranked shortlist of two, occasionally three, plus a written note of the missing evidence for each, which becomes the agenda for Step 5 rather than a loose end.
Step 5: Make finalists demonstrate, not present
Finalists get a demonstration stage built from your scenarios, not their slideware. Issue the same scripted cases to each, drawn from real operations and including the awkward ones: the month-end exception, the compound discount, the credit note that spans two systems. Require the people who would run your account to attend, not the sales organisation. Take references beyond the curated list: ask each finalist for a customer of similar size that left within the last two years, and treat the handling of that request as evidence in itself. Speak to reference customers without the vendor on the call, and ask operational staff rather than the executive sponsor what broke in the first year and how the vendor behaved once the invoice was no longer at risk. Everything from this stage scores into the same anchored model as the RFP, so demonstration polish cannot displace documented evidence. Where the purchase includes implementation, evaluate the delivery partner and its named team with the same anchors, because contracts are signed with companies and experienced through the individuals who turn up.
Step 6: Price the risk before you price the deal
Before negotiation, complete a Vendor Selection Risk Matrix for each finalist covering delivery, financial, dependency and exit risk, and score the shortlist with the Procurement Risk Calculator before commercial negotiation starts. The intent is to convert vague unease into priced items. A finalist whose product is excellent but whose company is small enough that your contract would be a third of its revenue is a concentration risk with a price: escrow arrangements, step-in rights, prepayment limits. A finalist whose migration tooling is weak carries a transition risk with a price: fixed-fee migration, acceptance-based payment. Feed each priced risk into the negotiation agenda as a term to secure rather than a worry to hold. The ordering matters, because risk raised after prices are agreed reads as re-trading, whereas risk raised before pricing simply is the agenda.
Step 7: Negotiate with two finalists live
Keep two finalists in genuine contention until terms are agreed, and let both know it. The cost of running two negotiations is a few weeks of effort; the cost of negotiating with one is every term you will never know you could have had. Negotiate the risk items from Step 6 first, then commercial structure (volume bands, renewal caps, service credits with teeth, exit and data-return obligations), and unit price last, because unit price is the term vendors concede most readily and the one that matters least over the life of the contract. Append what was promised during the process to the contract as representations: proposal claims, demo commitments, question-round answers. The scored evidence file makes this stage faster, because a vendor shown exactly where it trails the alternative on a criterion tends to close the gap in terms rather than argue the score.
Step 8: Record the decision while it is still fresh
Close the selection with a decision record, not just a contract. Complete the Decision Log Template with the recommendation, the final scores, dissenting views recorded in their owners' words, the risks accepted knowingly, and the conditions attached to the choice, such as the escrow or migration terms from Step 6. Circulate it to the steering group and file it where the renewal team will find it in three years. The record protects the organisation in both directions. It stops post-hoc rewriting when the vendor disappoints ("we knew about that risk; we priced it"), and it protects the evaluators when a losing bidder escalates. It also makes an honest post-decision review possible, because without a record of what was expected, nobody can say whether the selection delivered it. Put that review in the diary now, twelve months out, while the implementation memory is fresh enough to be honest.
When to bring in perspectives with no stake in the outcome
A disciplined process removes most steering, but it cannot remove the buyer's own blind spots: requirements that mirror the incumbent, weightings that flatter the sponsoring function, risk scores softened because the team is tired and wants to finish. Independent review pays at two points. Before the RFP is issued, an outside reading of the requirements and weights catches the corruption that is cheapest to fix, since after issue every correction is visible to bidders. And between scoring and signature, when the ranking is close or the panel has argued to a stalemate, selected senior operators from the Global Board who have bought and sold in the relevant market can pressure-test the shortlist against what these contracts look like in year three rather than at signature. The report is confidential and arrives in days, which fits inside even a tight procurement timetable.
