Manufacturing as a Service: Why Capacity Is Not a Delivery Promise

A marketplace listing is not a production commitment

Manufacturing as a Service is attractive because it promises access to distributed production resources without requiring every buyer to build a private supplier network from the beginning. A company can describe a requirement, discover relevant capabilities, and request production from an appropriate partner. The difficulty is that a listed capability and an executable order are separated by engineering, capacity, quality, and commercial decisions. A useful network has to support that journey rather than treating a successful search result as evidence that a part can be delivered on the required terms.

Fraunhofer IOSB describes Smart Factory Web as a network intended to improve order fulfillment through flexible production capabilities and shared resources. Smart Factory Web project. An April 2026 research preprint presents a Factory-X architecture and prototype for automating interactions between manufacturers and service platforms. A prototype evaluation is evidence about the studied implementation, not proof that all marketplaces achieve the same results. Factory-X MaaS preprint. This article develops an original operational scenario to explain what a manufacturer should test before relying on a distributed capacity offer.

Describe the requirement before searching the network

Imagine a buyer needing a small batch of machined enclosures. The requirement includes material, geometry, tolerance, surface finish, inspection evidence, quantity, delivery location, and drawing revision. Some of those fields can be expressed in a structured format; others may require an engineering discussion. A search that uses only the process name will produce a broad list, while a search demanding every conceivable detail may exclude useful candidates prematurely. The design should support progressive disclosure: enough information to find plausible resources, then sufficient detail for qualification and quotation.

Distinguish mandatory conditions from preferences. A required material grade should not be traded against a slightly shorter lead time unless an authorized engineer approves the substitution. A preferred shipping method may be more flexible. If a ranking combines all criteria into a single score without exposing those distinctions, the result can conceal an unacceptable compromise. A useful shortlist should explain why each candidate appears, which requirements are supported, and which need confirmation. That explanation is operational evidence, not merely a way to make a recommendation look more persuasive.

Interpret capacity in its production context

Capacity is not a universal number of machine hours. It depends on setup, tooling, staffing, maintenance, inspection, material availability, and the current order mix. A supplier may have theoretical time available but lack the fixture needed for a particular part. Another may have the equipment and fixture but no qualified operator during the requested window. A network should express capacity at a level appropriate to discovery while preserving the distinction between an indicative opportunity and a firm commitment.

Use a quotation conversation to resolve the dependencies. Ask what assumptions support the proposed delivery date and which inputs the buyer must provide by a particular time. If the buyer changes a drawing, the earlier capacity statement may no longer apply. Record that relationship. A capacity record that remains unchanged after a material or inspection requirement changes can mislead downstream planning. The practical objective is not perfect real time visibility into every supplier's schedule. It is a clear path from a discovery signal to a confirmed commitment whose assumptions both parties understand.

Count the supplier's quotation effort

A marketplace can create many inquiries without creating proportionate business value for suppliers. Each request may require engineering review, costing, clarification, and a decision about whether to quote. If most requests are unsuitable or unlikely to convert, the platform can consume scarce expertise. Measure the effort per qualified request and per accepted order, not only the number of inquiries generated. A network that makes discovery easy for buyers should also make it practical for suppliers to decline unsuitable work early and explain what information is missing.

Consider a hypothetical supplier receiving forty requests in a month. If each requires thirty minutes of initial review, the first screening stage consumes twenty hours. If only four become orders, that is five hours of initial screening per order before detailed quotation work. These are illustrative assumptions, not industry averages. They show why better requirement matching can be as important as more traffic. An onboarding and search process that reduces irrelevant inquiries may improve supplier participation more than a dashboard celebrating a rising number of request messages.

Define the handoff from quotation to order

A quotation is prepared under assumptions about revision, quantity, delivery conditions, and validity. An order should explicitly reference the accepted version and any negotiated changes. Otherwise the platform may retain several plausible records without establishing which one governs production. The handoff should identify the responsible parties, the approved technical package, the quality requirements, and the route for resolving changes. A simple accepted status is insufficient when an attachment was revised during negotiation or a supplier proposed a condition in a separate message.

Test a realistic exception. The buyer accepts a quotation and then changes a critical tolerance before material is cut. Can the supplier pause the order, request review, and issue a revised commitment? Can the buyer distinguish the original agreement from the amended one? These questions matter more than whether the platform offers a visually smooth checkout. Manufacturing orders can involve technical uncertainty that a generic retail transaction model does not represent. A useful service workflow makes those differences manageable rather than forcing them into free text notes that are easy to overlook.

Carry quality evidence through delivery

The delivered part needs a relationship to the requirement and the evidence agreed for it. That may include inspection results, material documentation, and traceability information appropriate to the order. The platform should not imply that one generic supplier rating replaces those order specific requirements. A supplier can perform well overall while a particular batch needs investigation. Preserve the identifiers connecting the order, produced items, relevant evidence, and any deviation accepted by the buyer. Those links support both routine receipt and later problem resolution.

If a nonconformity occurs, the network needs a clear process for reporting, containment, investigation, and disposition. Determine which information is shared with the platform and which remains between the commercial parties. Public ratings are a poor substitute for a technical corrective action process, especially when a dispute concerns confidential drawings or measurement methods. The operating model should support evidence based resolution while avoiding unnecessary disclosure. A marketplace can facilitate the interaction, but responsibility for product acceptance and technical decisions should remain explicit rather than disappearing behind a platform's general terms.

Avoid booking the same capacity twice

A supplier participating in several marketplaces may publish overlapping availability. Without a reconciliation process, several buyers can interpret the same indicative window as available to them. A network should distinguish discovery, tentative reservation, and confirmed allocation. Each state needs a validity period and a rule for release. A reservation that never expires can make useful capacity appear unavailable; an untracked verbal promise can create conflicting expectations that the planning system does not recognize.

Run a trial with simultaneous requests through two channels. Inspect how the supplier updates availability and how buyers learn that a window has changed. This is not an argument that every network must centralize the complete production schedule. It is a recommendation to test the business semantics of the advertised capacity. The supplier needs a manageable way to maintain commitments across its own sales team and external platforms. If that work depends on repeated manual copying, include the effort and error risk when evaluating whether the service model is economically useful.

Evaluate the network with a balanced scorecard

Buyer metrics can include time to a qualified shortlist, time to an executable quotation, and delivery performance for accepted orders. Supplier metrics can include screening effort, quotation conversion, clarification cycles, and the cost of maintaining capability records. Network metrics can include stale offers, unresolved disputes, and the reliability of identity and access processes. These measures should be interpreted together. Improving one side's convenience by increasing the other side's unpaid work may weaken the network even if the software's transaction count rises.

Use a bounded pilot with an understandable product family and a manageable number of participants. Preserve the existing sourcing baseline and record the support provided during the trial. A heavily assisted pilot can demonstrate a promising workflow, but its support cost should not disappear from the evaluation. Ask what would happen when the number of participants grows or the project team stops answering every question personally. Expansion is justified when the operating model can maintain useful matches and credible commitments, not simply when the architecture can accept more accounts.

Make the decision to expand evidence based

This article uses public Fraunhofer material and a 2026 research preprint for context. Its operational examples and arithmetic are editorial illustrations, not observations from participating factories. The preprint is identified as research rather than a universal commercial benchmark. Actual outcomes depend on the product, supplier capabilities, buyer requirements, commercial arrangements, and quality system. No general claim of cost reduction or delivery improvement follows merely from connecting an additional marketplace or exposing a capability through an API.

The useful next step is a complete rehearsal from requirement to accepted delivery for one realistic order. Include at least one clarification, one revision, and one potential capacity conflict. Ask both buyer and supplier to record the work required. If the network helps them reach a defensible commitment with less friction, the trial has produced meaningful evidence. Manufacturing as a Service becomes credible when discovery, qualification, commitment, and quality evidence form one understandable process. A larger catalog is valuable only when those relationships remain reliable as more companies participate.