Two offers indicate the same chamber capacity and similar productivity. The first includes only standard tests; the second encompasses load development, tools, validation support, and local assistance. The initial price does not allow for a comparison: first, boundaries, performance, responsibilities, and exclusions must be made equivalent. The selection must demonstrate which solution satisfies the site's need throughout its lifecycle.
1. Define the need before the request for proposal
The RFP must describe the intended use, materials, load families, required capacity, layout constraints, and markets served. Distinguish between useful capacity, productive capacity, and expected availability. A larger volume does not guarantee more production if loading, cycling, cooling, or handling become the bottleneck.
For a tunnel, specify formats, container status, interfaces with the washer and filler, line speed, and conditions. For an autoclave, describe loads, packaging, handling, and side separation. For an oven, clarify whether the objective is sterilization, depyrogenation, or both, with the relative required evidence.
[GUIDEGXP RECOMMENDATION] Prepare a common basis for comparison with verified data, hypotheses, and information yet to be confirmed. Ask suppliers to respond explicitly to every requirement and to declare their assumptions. This prevents an apparently advantageous price from depending on conditions that the site cannot guarantee.
2. Separate indispensable requirements and comparative criteria
Indispensable requirements determine the eligibility of the solution. They may concern load compatibility, demonstrable performance, critical interfaces, and data integrity. Comparative criteria distinguish between offers that are already technically acceptable: ease of maintenance, consumption, flexibility, or quality of support.
Do not compensate for a critical gap with a good commercial score. A weighted matrix can assist the decision only if preceded by a verification of non-negotiable requirements. The score must be supported by evidence, avoiding evaluations based on impressions or commercial presentations.
[QRM] Identify the risks that the site can manage and those that require a modification of the offer. For each significant difference, indicate the impact, necessary action, cost, and responsible party. Risk assessment must not become an automatic justification for accepting unproven performance.
3. Evaluate process competence
Ask how the supplier addresses cycle development, load families, worst-case scenarios, and site transfer. Verify the ability to explain the link between design and performance. A useful answer distinguishes what is already proven from what will require studies on the specific material.
References must be relevant to the technology, scale, and application, while respecting confidentiality. The number of installed machines does not replace process understanding. Evaluate examples of problems faced and investigation methods, without requesting confidential data from other clients.
For depyrogenation, examine experience in thermal profiles, tunnel balancing, and endotoxin challenge. For moist heat, consider air removal, penetration, liquids, or porous loads according to the intended use. Competence must correspond to the difficulty of the project, not just the machine's commercial category.
4. Compare design and integration
Evaluate usable dimensions, accessibility, materials, drainage, racks, doors, and maintenance. Verify the space required to replace components and calibrate instruments. A layout that allows for installation but hinders routine intervention can increase downtime and costs throughout the plant's life.
Interfaces must be explicit: utilities, drains, HVAC, network, line signals, and supply boundaries. Ask for minimum and maximum conditions required and the consequences of variations. Clean steam quality is distinct from cycle performance, but the supplier must clarify which input conditions are necessary for the intended operation.
For pass-through systems, evaluate door management, material status, and protection toward the receiving side. For tunnels, examine accumulation, stops, and restarts. Performance must be evaluated in the context of the whole line, avoiding single-machine optimizations that shift the problem upstream or downstream.
5. Verify automation and data availability
Request a description of the architecture: PLC, HMI, recording, users, audit trail, backup, and interfaces. Clarify which functions are included in the standard configuration and which require licenses or additional components. The term "GMP compliant" must be translated into verifiable requirements and tests.
Evaluate recipe management, versions, parameter limits, and behavior during failures. Ask for examples of records and methods for retrieving original data. An attractive report is not enough if events, metadata, or information necessary for review are missing.
Define access to programs and configurations according to ownership and licensing agreements, without assuming rights not included. Clarify how the site will maintain the system in case of obsolescence or termination of support. Remote access must respect the authorizations, controls, and logging provided by the site.
6. Make the CQV package comparable
Separate standard documents, project adaptation, test execution, and support for deviation resolution. Specify who prepares and approves protocols, who provides tools and loads, who performs analysis, and who writes the reports. The item "validation package" can encompass very different activities.
[REGULATORY REQUIREMENT] Annex 15 allows the use of supplier evidence when appropriate and justified, while maintaining site control over its suitability. The presence of a FAT does not automatically eliminate post-installation checks. The RFP must therefore describe the level of documentation necessary to support this assessment.
Agree on the format and content of data: identification of instruments, calibrations, maps, original data, calculations, and deviations. Ask for anonymized examples or templates when available. Evaluate the ability to deliver reviewable evidence, not just the number of pages in the file.
7. Define FAT, SAT, and contractual criteria
The acceptance plan must indicate test conditions, criteria, responsibilities, and management of exclusions. Link declared performance to specific loads, formats, and utilities. A productivity guarantee without conditions can become difficult to verify at the time of delivery.
Establish how open points and retests will be handled. Distinguish between what prevents shipping, installation, or productive use. Clarify who bears the costs of retesting when an agreed performance is not reached, within the limits of approved commercial agreements.
The technical review must support the contract without replacing the project's legal and commercial expertise. Clauses must reflect concrete decisions: resource availability, timelines, documents, and acceptance criteria. Avoid generic wording that leaves the meaning of "accepted machine" unresolved.
8. Technical evaluation matrix
| Dimension | Evidence to request | Critical signal |
|---|---|---|
| Process | Approach to loads, development, and limit conditions | Universal recipe proposed without data |
| Design | Drawings, accessibility, and defined interfaces | Site constraints not considered |
| Automation | Functions, records, versions, and failure tests | Compliance declared without evidence |
| CQV | Responsibilities, protocols, and deliverables | Package not described |
| Service | Coverage, times, and available expertise | Promises without operational conditions |
| Life Cycle Cost | Consumption, maintenance, licenses, and scenarios | Comparison limited to machine price |
Apply weights consistent with the project and document the justifications. A solution may be preferable for a site with strong internal expertise and less suitable for a site that depends on external support. The matrix must make these assumptions visible, maintaining a brand-neutral evaluation.
9. Evaluate maintenance, spare parts, and assistance
Ask for a maintenance schedule, critical components, replacement times, and spare parts availability. Distinguish between consumables and long-lead-time parts. Evaluate which stocks are justified by the risk of downtime and which can be managed through availability agreements.
For assistance, specify geographical coverage, hours, skills, and response times. Distinguish between phone support, remote diagnosis, on-site arrival, and restoration: these are different services. A short response time does not guarantee the immediate availability of the necessary component.
Consider internal staff training and knowledge transfer. Clear manuals, updated schematics, and diagnostic tools can reduce reliance on the supplier. Activities requiring authorized personnel or specialized skills must be identified and planned.
10. Build a transparent TCO
Total cost includes purchase, installation, site adjustments, qualification, development, energy, utilities, maintenance, consumables, licenses, and end-of-life. Include the cost of downtime according to declared assumptions. Avoid apparent precision when reliability or consumption data are not yet available.
Compare offers over the same timeframe and for the same production demand. Separate one-time and recurring costs, declaring any discounting and economic assumptions used by the project. Do not introduce universal rates or prices: use site data and agreed scenarios.
For consumption, ask for measurement conditions and boundaries: startup, standby, cycle, cooling, and auxiliary utility usage. A reduction per cycle may be less significant if the useful capacity changes or downtime increases. The comparison must refer to a relevant productive unit.
11. Example: lower price, different scope
In an illustrative comparison, offer A has a lower machine price but excludes recipe adaptation, validation tools, and post-SAT support. Offer B includes these activities but requires an annual license for certain recording functions. Neither is automatically better.
The team normalizes the scopes, estimates excluded activities with site data, and evaluates coordination risk. It then compares production and maintenance scenarios. If the most uncertain assumptions change the ranking, it requests clarifications or tests before the decision, instead of hiding the uncertainty in the final score.
The recommendation documents performance, costs, residual risks, and conditions to be closed in the order. This allows decision-makers to understand why a solution is preferred and what commitments support the result. The choice remains vendor-neutral and project-specific.
12. Manage obsolescence and end-of-life
Ask for the support policy for hardware and software components, version availability, and obsolescence notices. Evaluate how access to data, recipes, and documentation will be maintained. The risk is not only physical failure: an unsupported component can make it more difficult to maintain safety and validated status.
Define retrofit possibilities and impacts on interfaces. Future migration must preserve records and the ability to demonstrate the process. Agreements should clarify what information and services will be available, without assuming that an upgrade is always simple or included.
Consider dismantling, disposal, and data management at decommissioning. Plan for necessary archiving before losing access to the system. These costs may be far in the future, but they must be visible when differences between solutions make them materially relevant.
13. Checklist before the order
Verify that indispensable requirements are met, exceptions closed or accepted, and responsibilities defined. Confirm drawings, utilities, performance, documents, tests, training, spare parts, and support. Check for consistency between the technical and commercial offer: an activity described in the former might be excluded in the economic terms.
The final decision must indicate open assumptions and related deadlines. Establish how changes after the order will be handled and which approvals they require. A clear contractual and technical baseline reduces ambiguity during construction, installation, and validation.
Red flags are capacity without reference load, validation promises without responsibility, inaccessible data, and undefined service. The most robust selection connects the site's need to demonstrable performance, comparable costs, and sustainable support over time.
14. Verify capacity with the production calendar
For a batch machine, build a sequence that includes preparation, loading, cycle, unloading, and inter-batch activities. Consider which operations can overlap and which require the same resource. Exposure time alone does not represent the duration of the service required from the machine.
For a continuous line, connect validated speed, formats, changeovers, startups, and stops. Verify that the washer and filler can operate under the considered conditions. A high nominal tunnel capacity does not increase production if the downstream interface requires frequent stops or if the critical format imposes a more restrictive window.
Compare an ordinary scenario and scenarios with higher demand, maintenance, or reduced availability. Use declared assumptions and ranges when data is uncertain. This analysis allows for discussing redundancy, sizing, and capacity reserves on a concrete basis, avoiding buying unusable space or underestimating necessary resources.
Finally, verify the sensitivity of the choice to the most significant assumptions: load mix, changeover times, and support during a failure. If a small variation changes the preferred solution, that variable deserves confirmation before the order. The robustness of the decision depends on the quality of the assumptions as much as on the detail of the financial spreadsheet.
References and pathways
Sources verified on September 23, 2026: EU GMP Annex 15 and Annex 11; ASTM E2500-25, official catalog; ICH Q9(R1), 2023. The commercial matrix and comparison model are original GuideGxP recommendations.
Consult the Sterilization & Depyrogenation Systems hub, the URS requirements, the qualification section, and Automation & Digital Systems.
