Who should use this guide?
This guide is written for business owners, technology leaders, procurement teams, architects, security reviewers, finance partners, implementation leads, and operational administrators who need to turn an uncertain buying decision into a defensible recommendation. Smaller teams may combine several of these roles. The responsibilities do not disappear when the job titles do.
Use the guide before a shortlist becomes politically fixed. It also works as a review checklist when a project is already underway. In that case, record which decisions are complete, which are based on assumptions, and which need evidence before the next approval gate.
Define the decision before collecting information
A good evaluation starts with a written decision statement. Name the business problem, the affected teams, the expected outcome, the deadline or constraint that matters, and the person accountable for the final recommendation. Without that statement, research expands endlessly and stakeholders score different problems.
Then define the decision boundary. Clarify the processes, entities, regions, users, data, integrations, controls, and services included in scope. Record important exclusions as well. An excluded requirement can be just as significant as an included one when vendors estimate products and services.
- Business scope: processes, operating units, regions, products, and outcomes.
- Technology scope: applications, environments, data, identity, integrations, analytics, and infrastructure.
- Delivery scope: design, configuration, migration, testing, training, cutover, and support.
- Commercial scope: products, usage metrics, services, assumptions, renewal, and exit.
Build an evidence model
Teams often use a scoring spreadsheet but never define what earns a score. A vendor statement, a presentation slide, a configured demonstration, and a tested result are not equivalent evidence. Set the evidence hierarchy before vendors respond.
For this decision, useful evidence can include requirements, demonstrations, architecture reviews, reference checks, delivery plans, and commercial proposals. Give higher confidence to current, observable, and contractually supported evidence. Give lower confidence to generic statements, unconfigured screenshots, future roadmap claims, and assumptions that have no named owner.
| Evidence level | Example | How to score it |
|---|
| Observed | The team sees the agreed scenario work with representative data | Score against the scenario and record any configuration or dependency |
| Documented | Current product, architecture, security, or service documentation supports the claim | Confirm version, scope, and contractual relevance |
| Referenced | A comparable customer explains how the capability operates | Check similarity, limitations, and implementation context |
| Asserted | A response says the requirement is supported | Treat as unverified until stronger evidence is supplied |
| Planned | The capability appears on a roadmap | Do not score as current capability unless the decision explicitly accepts the risk |
Ask questions that expose operating reality
Feature questions are easy to answer positively. Operating questions are harder, and more useful. Ask who configures the capability, which product or edition provides it, what happens when data is incomplete, how errors are detected, how access is reviewed, how releases are tested, and which team owns the service after implementation.
For ERP, SCM, use end-to-end scenarios that cross team and system boundaries. When reviewing Infor CloudSuite, Microsoft Dynamics 365, Oracle NetSuite, SAP S/4HANA, distinguish the vendor's product responsibility from implementation-partner work and customer-owned activities. A complete-looking solution may still depend on middleware, specialist products, manual controls, or internal administration.
Data, integration, and reporting
Data is not a late implementation task. Inventory authoritative records, identifiers, duplicates, history, retention, quality rules, ownership, and reporting definitions during evaluation. Ask how the proposed design handles corrections, late events, partial failures, and reconciliation.
Every important integration needs a business purpose. Document its trigger, direction, data, timing, volume, mapping, credentials, failure behavior, monitoring, recovery, and support owner. Avoid accepting a connector name as proof that the required workflow is covered.
Reporting needs the same discipline. Define metrics, calculation rules, refresh expectations, security, drill paths, export needs, and semantic ownership. A polished dashboard built on inconsistent definitions does not improve decision quality.
Security, privacy, resilience, and compliance
Security evaluation should reflect the actual configuration and operating model. Review identity, privileged access, segregation of duties, encryption, logging, monitoring, vulnerability handling, incident responsibilities, backup, recovery, data location, subprocessors, retention, and deletion. Current certifications may support due diligence, but they do not prove that your implementation is compliant.
Translate regulatory and policy obligations into controls the project can design and test. Name the owner who accepts any remaining risk. If evidence is unavailable, record the state as not verified rather than filling the gap with an assumption.
Implementation and organizational capacity
Compare the proposed solution with the organization's ability to deliver and operate it. Count the business decisions, data work, integrations, custom development, testing, training, and process change—not only the implementation duration in a proposal.
A credible plan identifies named roles, dependencies, customer responsibilities, acceptance criteria, environments, test cycles, cutover activities, operational readiness, and post-launch support. It also explains how scope changes will be governed. The relevant risk for this guide is choosing a capable product that the organization cannot implement or govern.
Commercial and total-cost review
Normalize the proposals before comparing totals. Use the same users, roles, entities, environments, transactions, storage, support level, implementation scope, integrations, data assumptions, and growth scenarios. Separate recurring cost, one-time delivery, optional work, contingency, and internal labor.
Review renewal mechanics, price changes, minimum commitments, audit rights, service credits, data rights, subcontractors, transition assistance, and exit provisions. A commercial agreement should reflect the solution that was evaluated, including its dependencies and service responsibilities.
Common mistakes
- Starting with a preferred vendor and writing requirements around it.
- Giving vendors different scenarios, data, or commercial assumptions.
- Scoring roadmap statements as available capability.
- Ignoring administration, release management, and internal support effort.
- Leaving data, integration, security, or adoption work until implementation.
- Comparing headline prices instead of normalized total cost.
- Treating stakeholder consensus as a substitute for evidence.
Run the decision workshop
Bring the accountable business owner, process leads, architecture, data, security, delivery, procurement, finance, and operations into one working session before the recommendation is finalized. Send the evidence pack in advance. During the session, review the decision statement, mandatory requirements, major scoring differences, unresolved assumptions, delivery dependencies, commercial scenarios, and the risks that need explicit acceptance.
Do not use the workshop to replay every demonstration. Focus on disagreements that could change the outcome. When two options score closely, test the assumptions behind the scores and identify the evidence that would resolve the difference. Record decisions in plain language. Each action needs an owner, due date, and approval consequence. If missing evidence cannot be obtained in time, show how that uncertainty affects confidence rather than quietly treating the requirement as satisfied.
The workshop should finish with one of three outcomes: a supported recommendation, a short evidence-gathering step with a fixed deadline, or a decision to stop because the business case is not strong enough. Continuing by default is not a neutral choice; it spends time and weakens negotiating leverage.
Final decision record
The recommendation should explain why the selected option fits the decision statement, which evidence supports it, which compromises were accepted, what remains unverified, and which conditions must be satisfied before contract or go-live. Include dissent when it identifies a material risk. That record protects continuity when project members change and gives implementation teams the context behind important choices.
Revisit that record whenever scope, evidence, cost, ownership, or delivery assumptions materially change.
Finish with a short action list: named owner, due date, required evidence, approval gate, and escalation path. A guide creates value only when it changes the next decision.