From Rising RFQs to Order Realisation: An Observation of Purchase Intent and Conversion Differences in Manufacturing

Report ID: TAIWHA-RFQ-OBS-20260813　|　Evidence cutoff: 12 August 2026

Principal conclusion

An increase in requests for quotation without a corresponding increase in orders does not, by itself, demonstrate a broad decline in underlying purchase intent across manufacturing.

Public-company cases show that quotation activity can diverge from order timing or supplier allocation. Some projects may still proceed to purchase, but take longer to move from quotation to order; others may proceed while the order shifts to another supplier.

LSI Industries also provides a counterexample: project quotation activity rose alongside its book-to-bill ratio and backlog. Xometry, meanwhile, shows that revenue, active buyers and high-spending accounts on a digital manufacturing platform can grow at the same time, although it does not disclose changes in RFQ or search activity.

A further explanation remains to be tested: the same project may invite more suppliers, generate more quotation versions, or appear repeatedly across channels. The resulting expansion in RFQ counts is arithmetically possible, but its prevalence in practice has not been measured. Local cases therefore show that RFQs and orders can diverge in timing or supplier allocation. Representative data on mature projects are still lacking, however, and do not establish whether manufacturing as a whole is experiencing a sustained structural deterioration.

Artificial intelligence may reduce buyers’ costs of searching, generating RFQs and comparing suppliers more quickly, while complex manufacturing quotations still require engineering, costing, quality and capacity judgments. This asymmetry is plausible as a mechanism, but matched buyer-supplier data on working time and mature-project outcomes are not yet available. It cannot be presented as an established industry fact.

Three distinctions underpin the analysis:

The number of RFQs is not the same as the number of distinct procurement projects.
A customer completing a purchase does not mean that any particular supplier necessarily wins the order.
An order not appearing in the current system does not mean that no real-world outcome has occurred.

Research scope and use of sources: This report draws on public company filings, official rules, policy research and academic materials.

Company cases are used to identify possible mechanisms. Public-procurement rules clarify the scope of particular processes. Vendor materials do not support conclusions about market effects. The analysis is limited to observations, mechanisms and criteria for reassessment.

RFQs, Procurement and Outcome Records

At first glance, more requests for quotation (RFQs) without a corresponding rise in purchase orders (POs) appears to be a conversion-rate problem. In practice, it combines at least three questions. Did the buyer’s project ultimately result in a purchase? If the project continued, did the supplier under observation win? If an outcome occurred, was it recorded by the current customer relationship management system, platform, legal entity or sales channel? This report refers to the individual supplier under observation as the “focal supplier” and collectively calls these record domains the “current system”.

The orders visible to a supplier can therefore be decomposed using a diagnostic expression:

Observable orders ≈ number of distinct projects N
                  × probability that a project ultimately results in a purchase p
                  × focal supplier's conditional win rate a
                  × probability that the outcome is captured by the current system o

This expression is not a forecasting model and assigns no industry parameters. It simply shows that, after RFQs increase, a change in project purchase timing, supplier allocation or outcome visibility can prevent the POs visible to the focal supplier from rising in parallel.

An RFQ is not inherently a unit of demand. One distinct project may generate multiple RFQ events, invite multiple suppliers and produce multiple quotation versions. Award, purchase order, subsequent release, shipment, invoicing and payment are also separate objects. Microsoft’s overview of the RFQ workflow records RFQ cases, vendor invitations, replies, acceptances, purchase agreements and purchase orders separately.

Magna’s 2026 Annual Information Form illustrates an additional constraint in automotive supply chains. A PO may extend across multiple years without committing the customer to a minimum purchase volume. Actual supply quantities depend on subsequent production releases.

Public materials therefore establish that these process objects are distinct and provide cases in which quotation activity diverged from order timing or supplier allocation. Outcome visibility remains an unquantified measurement risk. Existing evidence also does not establish that local instances of divergence have developed into a structural deterioration across manufacturing as a whole.

Four States Behind Rising RFQs

The same pattern—more RFQs without a matching rise in orders—can reflect at least four materially different conditions.

Longer Procurement Cycles and Deferred Orders

Kadant reported that quotation activity for large capital projects remained active. Tariffs and economic uncertainty lengthened the time from quotation to order, and some capital orders were deferred into 2026; see Kadant FY2025 Form 10-K. This case is more consistent with a longer procurement cycle than with the disappearance of demand. It concerns one issuer and cannot represent manufacturing as a whole.

Re-sourcing and Supplier Migration

Superior Industries reported unusually active RFQs as tariff-driven localisation accelerated. During the same period, some unfinished business moved to other suppliers, placing pressure on the company’s volume and profitability; see Superior Industries Q1 2025 Results. The rise in RFQs here is more consistent with re-sourcing, dual sourcing or the search for a challenger supplier. End demand need not rise, and the focal supplier may lose orders even as RFQs increase. Company distress, tariffs and management attribution all affect this case, so it cannot be extrapolated into an industry-wide proportion.

RFQ Activity and Operating Outcomes Moving in the Same Direction

LSI Industries reported that project quotation activity in its lighting segment was above the prior-year level. Over the same period, its book-to-bill ratio was 1.1 and backlog also increased year on year; see LSI Industries Fiscal Q2 2025 Results. Xometry’s digital manufacturing platform also recorded growth in revenue, active buyers and high-value accounts; see Xometry FY2025 Form 10-K.

LSI does not disclose a consistently defined cohort of mature projects linking RFQs to POs, so the case cannot establish that matching quality improved. It nevertheless refutes the strong proposition that rising RFQ activity necessarily dilutes purchase intent. Xometry does not report changes in RFQ or search activity and therefore cannot directly serve as that counterevidence. The relatively high contribution from existing accounts indicates only that new and established customers are separate analytical groups.

Preliminary Indications of Procurement Outcomes Moving Across Channels

The absence of a PO online may mean that no purchase occurred. It may also mean that the transaction moved to a human-assisted channel, intermediary, framework agreement, another legal entity or an off-platform channel.

A working paper by Zhang and co-authors on AI-enabled B2B procurement offers preliminary indications; its English title and abstract are available on SSRN. The abstract states that AI increased early browsing and search activity. Online purchasing did not increase directly, while reliance on human-assisted offline channels rose. The full paper has not been obtained in a form that can be checked, and complete effect sizes are unavailable, so this report does not rely on it for a central conclusion.

Taken together, the public-company cases provide relatively direct support for two forms of divergence: longer order timing and supplier migration. They also provide a counterexample in which quotation activity and operating outcomes rose together. The possibility that observation channels create apparent divergence is currently supported only by weak preliminary evidence. A sustained deterioration in manufacturing RFQ quality across the industry remains unestablished.

RFQ Counts and Procurement Projects

Differences Between RFQ Counts and Distinct Project Counts

The first step in assessing “divergence” is not to compare total RFQ counts, but to define the object being counted. Microsoft’s RFQ workflow records an RFQ case, supplier invitations, replies and amendments separately. Based on these process distinctions, this report decomposes RFQ activity into five counting levels.

Total RFQ activity
≈ number of distinct projects
× average number of RFQ events per project
× average number of suppliers invited per event
× average number of quotation versions per invitation
× average cross-channel duplication factor

This is a descriptive decomposition for use when activity units and counting definitions are fixed. It is not an identity, estimation model or set of industry parameters. Projects differ in invitations, versions and channel structures; formal analysis uses project-level aggregation while retaining those differences.

Suppose a project certain to proceed originally invited 3 suppliers and later invited 10, while still awarding the work to 1. Measured by project, conversion remains 100%. Measured by supplier invitation, it falls from approximately 33% to 10%. This is an arithmetic illustration, not an empirical market parameter. It nevertheless shows that, unless projects and invitations are separated, an apparent conversion decline may reflect only a broader counting base.

Public workflow documentation establishes only that an RFQ can be amended, reissued, extended to additional suppliers and preserved in multiple versions. Engineering changes, drawing or bill-of-material revisions, raw materials, exchange rates, tariffs, validity periods and lead times are candidate triggers to examine in company records. Whether they generate repeated RFQs still requires project-level verification. Under a counting-based view, an additional version for the same project raises the RFQ count without increasing the number of distinct projects.

Criteria for Identifying Structural Divergence

“Candidate structural divergence” is a measurement hypothesis used in this report. It would have to pass all nine of the following tests:

Distinct projects have been deduplicated, and line-item aggregation, splitting and transaction allocation have been corrected.
Projects have reached a preregistered maturity window, with projects still in progress treated as incomplete observations.
RFQ events per project, suppliers invited per event, quotation versions and cross-channel duplicates have been controlled for.
Data have been segmented by industry, product model, degree of standardisation, customer size, new or established relationship, domestic or export market, geography and channel.
Seasonality, prices, tariffs and policy shocks have been addressed.
Buyer-project outcomes, focal-supplier outcomes and outcome visibility can be observed separately.
Final-outcome capture, status verification for mature projects and cross-system reconciliation have been estimated separately.
Each segment has preregistered its definition of the real outcome, maturity window and rules for judging persistence and materiality; after these steps, the cumulative incidence or cause-specific hazard of a real purchase from any supplier still declines.
Order deferral, supplier migration, changes in business mix, observation gaps and counterexamples cannot explain the decline.

These nine tests constitute the measurement framework used in this report; they are not an industry law validated by external research. No public, representative dataset currently links RFQs, invitations, quotations, awards, POs, subsequent releases and shipments for the same cohort of projects. The nine tests therefore have not all been satisfied.

Segmentation by Business Setting and Maturity Window

The following table is a framework for setting different observation definitions and maturity windows. It does not imply that public data have established the typical proportion or direction of each setting.

Business setting	Typical timing and procurement output	Common error when directly comparing RFQs with POs
Standard parts, catalogue items and maintenance, repair and operations (MRO) supplies	Shorter cycle; price, inventory and lead time are often compared	Low conversion for one supplier can coexist with strong project purchase intent
Make to order	Materials, minimum order quantities, configuration, lead time and capacity require multiple rounds of clarification	Quotation versions are miscounted as new projects
Engineer to order, tooling and capital goods	Budget quotation, design-for-manufacturability review, specification freeze, approval and samples create a long cycle	Deferral or lack of maturity is recorded as low intent
Automotive and quality-critical products	Production part approval, first-article inspection, audits, approved supplier lists and dual sourcing; releases still follow a PO	A PO is misread as a fixed volume, or completed qualification as a failed purchase
Procurement dependent on a downstream project	The buyer first bids for an automaker, original equipment manufacturer or end-customer project	An upstream RFQ is misread as end-customer-authorised demand
Large customers	Formal multiple quotations, framework agreements, split awards, backup suppliers and approvals are more common	Invitation-level conversion is diluted even though project authorisation may be more credible
Small customers	The process may be more direct, but can also reflect inadequate budgets or platform exploration	Direction cannot be assumed from customer size alone
New customers	Supplier discovery, qualification and switching frictions coexist	A challenger’s low win rate is misread as low project intent
Established customers	Repeat purchases may proceed through long-term or blanket orders; an occasional RFQ may serve only as a price benchmark	No RFQ does not mean no order, while an RFQ does not necessarily represent new demand
Domestic and export markets	Exporting adds landed cost, certification, exchange rates, Incoterms, logistics, payment and intermediary chains	Channel duplicates, timing differences and missing final outcomes are conflated with low conversion

Different settings also require different maturity windows. Standard parts or replenishment purchases may produce an outcome within weeks. Capital goods, tooling, qualification and downstream-project dependencies may span multiple quarters. A single 30-, 60- or 90-day window cannot be applied to all of them.

China’s Purchasing Managers’ Index for June 2026, published by the National Bureau of Statistics, shows divergence across industries and firm sizes. This supports only the proposition that the macroeconomic environment and business mix may influence aggregate RFQ-to-PO performance. PMI is not direct evidence of RFQ counts, values or conversion rates.

Strength of Public Evidence

Existing Evidence and Its Strength

The value of the available material is not that it provides an industry-average conversion rate. It distinguishes mechanisms that have case-level support from hypotheses that still require testing.

The table uses qualitative evidence ratings of “moderate”, “low to moderate” and “low”. These ratings combine source primacy, fit between evidence and proposition, retrievability and the permissible scope of inference. They are not statistical confidence levels.

Proposition under observation	Current evidence strength	Principal basis	Scope and interpretive limits
The quotation-to-order cycle for complex capital projects may lengthen	Moderate	Kadant FY2025 Form 10-K, Item 7, “Overview”	One issuer; not representative of the full industry
Re-sourcing may raise RFQ activity while orders move between suppliers	Low to moderate	Superior Industries Q1 2025 Results, CEO comments and subsequent events	Confounded by tariffs, company distress and management attribution
Quotation activity can rise alongside the book-to-bill ratio and backlog	Moderate	LSI Industries, page 2 of the news release	A counterexample from one issuer; it does not directly identify whether matching quality improved
Revenue, active buyers and high-spending accounts on a digital manufacturing platform can grow together	Low to moderate	Xometry, Form 10-K, “Operational and Business Metrics”, pages 49–51	Does not disclose changes in RFQ or search activity, or RFQ-to-PO conversion
RFQs can serve purposes beyond immediate purchasing	Moderate	Research on procurement strategies for multi-item RFQs; FAR Part 10	An academic survey and public-procurement rules cannot estimate prevalence across private manufacturing
AI may lower the cost of search, comparison and alternative-supplier discovery	Low	OECD research on competitive dynamics; OECD progress report on the EU Coordinated Plan on AI, Volume Two	Supports only a capability mechanism; it provides no causal estimate for manufacturing RFQs
Manufacturing RFQ quality is declining broadly and structurally	Not supported	No representative linked cohort of mature projects	Not established by current evidence

Counterevidence and Institutional Reference Points

Counterevidence limits the strength of the conclusion. LSI shows that greater quotation activity can coexist with a higher book-to-bill ratio and backlog. Xometry shows that revenue, active buyers and high-spending accounts on a digital manufacturing platform can rise together, but it does not report changes in RFQ or search activity. Historical material from the European Commission provides another counterexample: digitisation increased the number of bids in its public-procurement sample and reduced some process costs. The sample is limited and some outcomes are self-reported. It is not direct evidence about generative AI or private manufacturing; it shows only that lower participation friction need not produce worse procurement outcomes.

Neutral institutional evidence shows that a quotation is not inherently a commitment. Federal Acquisition Regulation 13.004 distinguishes a quotation from a binding offer; FAR Part 10 permits market research before a formal solicitation; and the Government Procurement Law of the People’s Republic of China illustrates a formal public-procurement process involving multiple suppliers before a successful supplier is selected. These rules do not represent all private manufacturing, but they show that conversion measured by supplier invitation can be mechanically diluted by the procurement mechanism.

Procurement, Suppliers and Outcome Records

Suppliers often assess RFQ quality by asking whether an RFQ led to a PO for them. This metric has operational value, but cannot by itself determine whether underlying purchase intent has declined. It combines the buyer’s project, the competitive outcome and the observation system into one measure.

Three Observation Levels for Procurement Outcomes

Object observed	Question answered	Outcome to record	Common misreading
Buyer project	Did the buyer make a real purchase from any supplier?	Real purchase, award, subsequent release and shipment	Inferring that the customer did not buy because the focal supplier lost
Focal supplier	If procurement continued, did this supplier win?	Sole win, split win, competitor awarded, incumbent retained	Substituting invitation-to-order conversion for market demand
Outcome visibility	Was the real outcome captured by the current system?	Final-outcome capture, status verification and cross-system reconciliation	Coding unknown, offline or cross-entity outcomes as losses

The focal supplier’s outcome and the buyer’s procurement outcome are separate analytical objects. Formal multiple-quote processes, the use of challenger quotations to negotiate with an incumbent, split awards, dual sourcing and localisation can all change the focal supplier’s sole-win probability or order allocation. The Superior Industries case shows that RFQs can increase while orders move to other suppliers.

Outcome visibility is a separate measurement problem. A transaction may occur through another legal entity, intermediary, human-assisted offline channel, framework agreement or subsequent release. A CRM may also lack the reason for a loss or the final status.

Outcome visibility is therefore represented by three measures:

the share of independently evidenced final outcomes captured by the current system;
the share of mature projects whose current status has been verified;
the share of the systems in scope that have been reconciled.

These measures are not interchangeable.

Differentiated Evidence from Public Cases

Case	Principal outcome layer affected	What it supports	What it cannot establish
Kadant	Order timing	Active quotation activity can coexist with order deferral	The industry-wide manufacturing cycle
Superior Industries	Supplier allocation	Rising RFQs can accompany orders moving to other suppliers	The respective contributions of tariffs, distress and re-sourcing
LSI Industries	Orders and backlog	Quotation activity and operating outcomes can move in the same direction	An improvement in consistently defined RFQ conversion
Xometry	Buyers, accounts and revenue	Buyers, accounts and revenue on a digital manufacturing platform can grow together	Without changes in RFQ or search activity, platform characteristics cannot be extrapolated into an industry-wide rule

These cases indicate that multiple competing explanations exist, but do not identify their respective causal contributions or prevalence. Explaining the divergence requires returning to the different tasks that buyers perform through RFQs.

Mechanisms Behind Repeated RFQs

A Classification of Procurement Tasks Served by RFQ Activity

RFQs do more than support immediate orders. The eight uses below form an observational framework for decomposing the issue. They are neither the only categories established by external research nor equivalent to probabilities of purchase.

Principal use	Outcome sought by the buyer	Typical timing	Common misreading
Direct purchase formation	Award, PO or split award	Short to medium	Competition among multiple suppliers still lowers the win rate for any one supplier
Budget formation and approval	Approval, adjustment of scope, suspension or no budget	Longer and conditional	A genuine need exists, but purchase authority has not yet been granted
Price discovery and benchmarking	Distribution of price, lead time, minimum order quantity and landed cost	May produce no PO	An information output is misread as a valueless RFQ
Supplier screening and qualification	Approved list, shortlist, trial, sample or audit outcome	A PO may be distant	Completed qualification is misread as a failed purchase
Greater bargaining leverage	Concession from the incumbent or improved terms	Challenger frequently receives no PO	The project continues, but the focal supplier does not win
Backup and resilient supply	Framework agreement, dormant qualification or a dual-source option	Timing uncertain	A long period without routine POs is misread as fictitious demand
Engineering validation or unresolved downstream project	Specification freeze, design-for-manufacturability review, engineering change, or purchase after the buyer wins downstream work	Long cycle, multiple versions and two timelines	A lost downstream bid is misread as failed supplier follow-up
Low-intent exploration	Expiry, silence or an explicit statement that there is no current purchase	No near-term outcome, or one far in the future	This is a candidate for low intent, but excludes technical duplicate records

Schoenherr and Mabert’s research on multi-item RFQs shows that industrial buyers consider price, supply security, procurement efficiency and portfolio formation at the same time. The study is based on an older US survey, and both its response rate and subjective measures limit extrapolation. It supports the proposition that procurement objectives are multiple; it does not establish the prevalence of the eight uses in this report.

Separating Procurement Commitment from Information Quality

To avoid conflating “a real project with incomplete information” and “complete information without purchase authority”, this report uses a P/Q framework with two separate axes, explicitly limited to an observational model:

Procurement commitment axis P: P4 is an authorised purchase; P3 is a named conditional project; P2 is a validated need without verified authorisation; P1 is early or weak project evidence; P0 is an explicit absence of a current purchase; and PX is unknown.
Information quality axis Q: Q4 is traceable; Q3 is sufficient to support a quotation appropriate to the setting; Q2 is sufficient only for a budget quotation; Q1 is incomplete; Q0 contains internal contradictions; and QX is unknown.

Q3 for a catalogue item and Q3 for an engineer-to-order project cannot require the same fields. Even if two projects are both classified P3/Q3, one may require only a catalogue lookup while the other involves tooling, samples, audits and engineering quotation. Their supplier response costs and timing structures remain entirely different. P/Q grades also cannot be converted automatically into a probability of purchase without calibration using company data.

Five further dimensions are recorded alongside P/Q as measurement hypotheses, preventing different questions from being compressed into a single score:

Dimension	How this report observes it	Relationship to P/Q
Procedural context	Formal multiple quotations, cross-border procurement, split awards, incumbent relationship, audit and compliance requirements	Explains why invitations are numerous and cycles are long; does not directly change procurement commitment
Response burden	Working hours required for sales, engineering, quality, costing and capacity judgments	Affects the economic constraint on supplier participation; cannot be replaced by information completeness
Supplier fit	Fit in technology, capacity, geography, certification, lead time and commercial terms	Affects conditional win rate; is not equivalent to whether the buyer’s project is real
Outcome visibility	Whether the final outcome appears in the current legal entity, channel, platform or system	Affects the observable outcome; is not the same as the real-world outcome
Record completeness	Duplicate records, cross-channel copies, superseded quotation versions and cross-system links	Affects counting quality; does not directly reduce procurement commitment or information quality

Duplicate records, cross-channel copies and superseded quotation versions first enter a record-completeness review. Their presence alone is not sufficient reason to reduce P or Q.

Buyer Information Gains and Supplier Response Costs

This report treats the information and cost asymmetry between the two sides as a mechanism to be tested. Within one procurement process, the buyer may know the budget, approval status, downstream-project dependency, preference for an incumbent, true number of suppliers invited and internal evaluation weights. The supplier has better knowledge of its own cost, capacity, engineering burden, quality fit and alternative opportunities.

A broader RFQ can help the buyer understand price, lead time, capabilities and backup sources. Each supplier response may still require sales, costing, engineering, quality and capacity resources, as well as opportunity cost.

The economic constraint on supplier participation can be expressed with another diagnostic formula:

Response value
≈ probability that the project continues × focal supplier's conditional win rate × expected profit
  + relationship or learning value
  − response cost

This expression is likewise not an automated quoting rule. It shows only that, when more suppliers are invited to each project, the probability of one supplier winning will ordinarily decline while the buyer’s informational gain from a broader search may still rise. Suppliers facing high opportunity costs may decline to quote, while low-cost or automated responses may take a greater share of the visible quotation pool. Establishing whether this selection effect actually occurs requires simultaneous observation of no-quotes, response burden, supplier fit and the qualified-response rate.

Procurement Progress, Supplier Outcomes and Timing

The buyer’s project and each supplier have separate state paths. The report first distinguishes three branches—pre-authorisation budget or market inquiry, authorised sourcing, and qualification or supply-resilience inquiry—and then observes how they enter a shared procurement process:

Stable need or project identity
├─ Pre-authorisation budget or market inquiry
├─ Authorised sourcing
└─ Qualification or supply-resilience inquiry
    ↓
Supplier discovery and invitation
→ Response, no-quote or revision
→ Technical clarification, sample or qualification
→ Commercial assessment and negotiation
→ Award or agreement
→ PO
→ Subsequent release and shipment

Microsoft’s RFQ workflow confirms that inquiry, reply, acceptance, rejection and cancellation are distinct states. Building on those workflow distinctions, this report defines a set of project-outcome categories: procurement realised, split realisation, no budget, final no-decision, deferred or cancelled. Re-RFQ, qualification only, framework agreement without release, buyer loss of the downstream project, still in progress at the cutoff date and unknown are retained as separate states. This is an observational framework to be tested against company project data, not an industry distribution identified in public materials.

The focal supplier’s outcome may instead be a sole win, split win, competitor award, incumbent retention, no-quote, qualification failure, qualification only, not invited, still in progress or unknown. Once the two state paths are separated, “the project purchased, but not from this supplier” will no longer be misreported as “the customer had no demand”.

Timing structure. A single project requires at least five sets of timing records:

project emergence to final outcome;
initiation to close for each RFQ;
creation to supersession for each quotation version;
award to PO;
PO to subsequent release.

A quotation revision can restart the version clock but not the project clock. Projects still in progress at the cutoff date remain incomplete observations rather than default losses.

Tests Required for the AI Cost-Effect Hypothesis

OECD research shows that AI can reduce search and verification costs, expand the comparable choice set and support discovery of alternative or backup suppliers. It does not provide the adoption rate, bilateral working-time data or causal outcomes for private manufacturing RFQs. Protolabs FY2025 Form 10-K reports that its automated quotation and manufacturability-analysis technology reduces some skilled labour traditionally required for quotation and manufacturing. The company separately reports the use of AI in pricing and sourcing. This shows that supplier-side automation may also reduce cost, but does not establish the rate of decline, much less its rate relative to the buyer’s.

The first test concerns the cost differential: within project groups of comparable complexity, is the buyer’s working time per RFQ falling materially faster than the supplier’s time to complete each qualified quotation? The second concerns behavioural transmission: do project groups using AI invite more suppliers or create more quotation versions without improving specification completeness, the qualified-response rate, mature-project purchasing or time to award? If supplier-side automation lowers costs in parallel, or matching quality, completeness and mature conversion improve, the strong asymmetry proposition would be weakened.

Neither test is currently satisfied by public, representative data. “Buyer’s RFQ cost approaches zero while supplier response cost remains materially positive” can therefore be treated only as an unmeasured mechanism hypothesis.

Competing Explanations for RFQ Growth and Procurement Outcomes

Competing explanation	RFQ activity	Real purchase by mature projects	Principal identifying variables	Current status
Repeated RFQs for the same project or cross-channel proliferation	Rising	Stable or rising	Suppliers invited per event, quotation versions, channel copies and line-item allocation	Arithmetically possible; prevalence not measured
Time lag or supplier reallocation	Rising	Deferred or shifted to other suppliers	Cycle length, competitor award, incumbent retention and subsequent release	Supported by cases; not an aggregate estimate
Genuine quality deterioration	Rising	Continues to decline after maturity	No budget, cancellation, three measures of outcome visibility and segmented cumulative incidence	Not currently established

The three explanations can be distinguished only when applied to the same project, timing and outcome records. The next section therefore specifies the variables required for continuing reassessment.

Indicators of RFQ Quality

The monitoring variable framework identifies what to monitor when reassessing the three explanations above. It comprises five groups of observations to be tested and does not impose a single threshold across industries.

Core Observation Measures and Risks of Misinterpretation

Object for reassessment	Core variables	Misreading to exclude
Project and record counts	Distinct project count, RFQs per project, suppliers invited per event, quotation versions, channel copies, and line-item aggregation, splitting and transaction allocation	Treating invitation proliferation, repeated RFQs, more versions or channel duplication as new demand
Outcomes and timing	Cumulative incidence of purchase from any supplier, focal supplier’s conditional win rate, time to award adjusted for incomplete projects, qualification completion, incumbent retention, no budget and final no-decision	Comparing only the median among awarded cases; sparse events call for restricted mean time or identifiable quantiles as additional measures
Observation completeness	Final-outcome capture rate, mature-project status-verification rate, cross-system reconciliation rate, and auditable coverage of P/Q and RFQ-purpose classifications	Coding unknowns as losses or treating unsupported classifications as fact
AI and bilateral burden	Buyer’s working time per RFQ, supplier’s working time per qualified quotation, specification completeness, qualified-response rate and share of AI-assisted work on each side	Treating “use of AI” itself as a causal explanation
Combined evidence	How project identifiers, specification freeze, budget approval, shortlists, technical discussions and commercial terms change together	Treating a single field as a calibrated intent score

Combined Indicators of Project Commitment

The following combinations are criteria for continued examination, not a calibrated intent score. A stable project or requisition identifier can add observable evidence of project commitment. A bill of materials, drawing version, quantity and required date that progressively stabilise, an interpretable budget or approval status, and a narrowing shortlist would indicate movement from exploration towards a concrete project.

A further observable signal is a change in the content of discussions: movement beyond a generic price inquiry towards manufacturability, quality planning, capacity and tooling ownership, with greater specificity around sample requirements, split deliveries, payment terms or Incoterms.

A different trajectory is more consistent with exploration, duplicate records or an unobserved outcome: the project cannot be deduplicated, and text remains highly similar across channels; critical specifications and timing remain missing for an extended period while the supplier pool continues to expand; deadlines and drawing versions are repeatedly reset without decision criteria or a final status.

No single field is sufficient to classify a project as low intent. Large, public or confidential procurements may limit feedback, while an RFQ for a standard part may naturally be brief.

Testing the Structural-Decline Thesis

Evidence Supporting a Structural-Decline Assessment

The current assessment is not static. The following evidence would strengthen the observation that RFQ quality is undergoing structural decline:

A representative linked cohort shows that, after projects are deduplicated and reach preregistered maturity windows, the cumulative incidence of a real purchase from any supplier continues to decline within each segment.
The share of no-budget, cancelled or governance-confirmed final no-decision outcomes rises persistently after those states have been predefined as terminal outcomes.
Final-outcome capture, mature-project status verification and cross-system reconciliation jointly exclude missing outcomes as the explanation.
Among AI-using project groups of comparable complexity, buyer time per RFQ falls materially faster than supplier time per qualified quotation, while invitations or versions proliferate and matching, completeness, mature-project purchasing and time to award do not improve.

Evidence Weakening a Structural-Decline Assessment

The following evidence would weaken that observation:

RFQ growth is explained mainly by more suppliers per project, more versions, re-sourcing, qualification or cross-channel copies.
Mature-project purchasing, subsequent releases, shipments and order values move in line with the number of distinct projects.
Low conversion mainly reflects awards to competitors or incumbent retention, while the buyer’s project still results in a purchase.
AI adoption raises specification completeness, qualified-response rates and mature-project conversion while shortening time to award.
Supplier-side automation causes the time required to complete a qualified quotation to fall at the same pace or faster.

Until such evidence emerges, this report retains two cautious assessments: a broad structural decline in manufacturing RFQ quality has not been established; and a persistent widening of the cost differential between buyers and suppliers due to AI remains an unmeasured mechanism hypothesis. The directions of any effects associated with new versus established customers, customer size, domestic versus overseas markets, and the existence of common thresholds across different product maturity cycles also remain unresolved.

Sources and Limitations

Scope of Sources

The evidence cutoff for this report is 12 August 2026. Fact checking draws on public company documents, official rules, policy research and academic materials; earlier reports and secondary reviews do not substitute for primary sources.

Principal Sources and Their Scope

Kadant Inc., FY2025 Form 10-K, Item 7, “Overview”: supports active quotation activity for large capital projects, a longer time from quotation to order, and some deferral; one issuer, not generalisable across the industry.
Superior Industries, Q1 2025 Results, CEO comments and subsequent events: supports the proposition that tariff- and localisation-driven re-sourcing can increase RFQs while orders move between suppliers; company distress and management attribution are confounding factors.
LSI Industries, Fiscal Q2 2025 Results, page 2 of the news release: supports the coexistence of greater quotation activity with growth in the book-to-bill ratio and backlog; the company does not disclose a consistently defined RFQ measure.
Xometry, FY2025 Form 10-K, “Operational and Business Metrics”, pages 49–51: supports simultaneous growth in revenue, active buyers and high-spending accounts on a digital manufacturing platform; it provides neither changes in RFQ or search activity nor an RFQ-to-PO conversion rate, and new and established customers need to be separated.
Magna, 2026 Annual Information Form, page 26, “Purchase Orders”: supports the distinction among POs, minimum purchase volumes, subsequent releases and actual shipments in automotive supply chains; it cannot be extrapolated directly to catalogue items.
Schoenherr & Mabert, An Exploratory Study of Procurement Strategies for Multi-Item RFQs in B2B Markets: Antecedents and Impact on Performance: supports the proposition that multi-item RFQs serve price discovery, supply security, procurement efficiency and portfolio formation at the same time; an older US survey and subjective measures limit extrapolation.
OECD, Artificial intelligence and competitive dynamics in downstream markets: supports the capability mechanism through which AI can reduce search and verification costs and expand the choice set; it provides no causal estimate for manufacturing RFQs.
OECD, Progress in Implementing the European Union Coordinated Plan on Artificial Intelligence (Volume 2): supports capabilities for procurement automation and discovery of alternative or backup suppliers; it is not evidence of adoption rates or outcomes.
European Commission, A strategy for e-procurement: provides historical counterevidence that digitisation can increase participation and reduce some process costs; the sample is limited and some results are self-reported, and the 2012 public-procurement setting is not evidence about generative AI or private manufacturing.
National Bureau of Statistics of China, Purchasing Managers’ Index for June 2026: supports the possibility that differences across industries and firm sizes affect aggregate performance; PMI does not measure RFQ counts, values or conversion rates.
FAR 13.004, Legal effect of quotations: supports the proposition that a quotation is not a binding offer in US federal simplified acquisition; it cannot be generalised to all private manufacturing relationships.
FAR Part 10, Market Research: supports the proposition that market research can precede formal procurement; it applies only to the US federal procurement context.
Government Procurement Law of the People’s Republic of China: illustrates the formal inquiry and supplier-selection process in public procurement; it cannot substitute for empirical evidence from private manufacturing.
Protolabs, FY2025 Form 10-K, Item 1, “Business”: supports the proposition that quotation for custom manufacturing involves manufacturability analysis and shows that automated quoting and manufacturability technology can reduce some traditional skilled labour; the filing separately reports the use of AI in pricing and sourcing, and the two cannot be combined into a general cost ratio.
Microsoft, Requests for quotation overview: supports the distinction among RFQ cases, supplier invitations, replies, amendments, acceptances, agreements and POs as workflow objects; software semantics are not evidence of market outcomes.

Limits on the Use of Supplementary Materials

Only the abstract of the working paper on AI-enabled B2B procurement referenced above has been checked; a full text suitable for verification has not been obtained. Commercial-platform promotional materials, vendor capability statements and pages whose full content could not be verified also do not support conclusions about market effects.

Critical Data Gaps

Public materials do not yet link projects, RFQs, supplier invitations, quotations, awards, POs, subsequent releases and shipments into one representative cohort of mature projects. Matched working-time measures for buyers and suppliers and cross-system reconciliation of final outcomes are also unavailable.

Until these gaps are closed, any assertion that RFQ quality has already declined broadly across the industry would go beyond the available evidence.

Version Updates and Correction Process

First published: 13 August 2026; version: 1.0. If material sources, measurement definitions or criteria for reassessment change, this report will be revised and the revision date recorded. This is an observational research report; its conclusions are limited to observations, mechanism explanations and criteria for reassessment supported by the current evidence.

Cookies and your privacy choices