The cost of evidence should be tested against the order it serves

A hold is first an operational and cash event. Goods remain unfinished, undispatched or otherwise unavailable for sale while teams preserve records, investigate, retest, correct labels or protect affected people. The financial model should represent those observable consequences. It should not invent a probability of enforcement or a universal penalty and call the result expected compliance value.

The forced-labour product framework illustrates why this distinction matters. The Commission describes separate assessment, dialogue, investigation, decision and enforcement stages, while also stating that the regulation itself introduces no audit or reporting obligations.[S10] A supplier precautionary hold is not the same event as an authority decision. The internal model must name which state it represents and which actor authorised it.

A narrow working-capital calculation can still be useful. The planning formula is inventory value under hold, multiplied by the annual funding rate and hold days, divided by the declared day-count basis. It estimates the funding cost of value tied up for that period. It does not include every consequence and does not predict the hold duration. Inventory value, funding rate and days are user inputs, not facts supplied by the regulation.

Exhibit 58 applies that formula to an illustrative inventory value of 120,000 currency units. The low and high bounds use annual funding rates of 8 percent and 16 percent, with a midpoint of 12 percent. Hold durations of 15, 30 and 60 days are planning scenarios on a 365-day basis. The exhibit is deliberately labelled illustrative because none of these inputs is an observed supplier or enforcement figure.

The midpoint for an illustrative 30-day hold is 1,184 currency units after rounding: 120,000 multiplied by 12 percent, multiplied by 30, divided by 365. The corresponding low and high values are 789 and 1,578 currency units. This range describes sensitivity to the declared funding rate only. It is not a statistical confidence interval, and the midpoint is not a forecast mean.

A complete hold account needs those separate lines because they have different owners and cash timing. Storage may accrue daily. Testing may require an external quotation before work begins. Rework may consume materials and capacity. Lost sales depend on the contract and subsequent buyer decision. Worker remedy must be determined by the harm and the affected person's needs, not capped by the remaining order margin.

Legal and advisory costs also need careful treatment. A scope review may resolve which evidence is necessary and prevent irrelevant work. It should not be presented as a fine avoided. Likewise, a shipment hold can protect against a worse outcome without proving that the alternative would have been an enforcement action. The decision record should state what the hold achieved, what remains unresolved and who may release the goods.

A hold can also distort the unit-cost denominator. Goods under critical review are not eligible shipped units. Moving them out of the denominator increases the measured evidence cost per eligible unit, but that increase is not the main reason to act. The primary question is whether the product and people are protected. Cash exposure helps management resource the response; it does not decide the legal or ethical release condition.

Evidence work can be organised inside the supplier, bought from specialists or supported through a buyer-shared service. Each route changes more than the invoice. It changes who defines the record, who can retrieve it, who corrects it and what happens when the commercial relationship ends. The cheapest visible service fee may not be the lowest total cost once portability and continuity are included.

An internal evidence team keeps process knowledge close to production and worker-management systems. It can connect batch events, specification changes and corrective action without sending every question through an intermediary. Its costs include training, supervision, cover for absence and independent review where needed. Internal control does not remove the need for competent laboratories, legal advice or specialist safeguarding support.

External specialists can provide expertise that would be inefficient to maintain for occasional work. The buyer of those services still needs a clear scope, competence check, confidentiality terms and usable output. A report locked inside a consultant's format may answer one questionnaire while leaving the supplier unable to reconstruct the evidence later. Procurement should therefore test data export, source access and correction rights alongside price.

A buyer-supported common portal may reduce repeated formatting and make request purposes clearer. It can also concentrate dependency. If several customers use different schemas, the supplier still needs a canonical internal record and mapping layer. If the buyer changes portal, access terms or hosting arrangements, the evidence must remain retrievable. A shared platform does not replace product-specific testing, worker engagement or the supplier's own release decision.

Personal information adds a sharper boundary. Where the GDPR applies, its principles include purpose limitation and collecting only the personal data necessary for that purpose.[S23] The EDPB separately explains cumulative criteria for transfers to organisations outside the EEA and conditions that apply in addition to other GDPR rules.[S24] These questions depend on the actors and data direction. A common portal is not automatically unlawful, but convenience alone does not answer necessity, access or transfer review.

The commercial comparison should therefore include exit cost and service continuity. Can the supplier export every record with its version history and evidence links? Can a worker's restricted information be withheld from a broad product extract? Will the service remain available during an incident, and who pays for urgent retrieval? These are operating requirements, not technical accessories.

A mixed model is often the most testable starting point. The supplier retains the canonical evidence register and decision rights. Specialists provide bounded analysis. Buyer portals receive controlled extracts. The pilot can then price each interface and observe where duplication remains. No vendor should be described as regulator-approved unless that specific status and scope are independently evidenced.

A cost model describes intended work. A tabletop drill tests whether people can perform that work when several problems arrive together. The exercise should use a fictional order, controlled records and a fixed decision table. It is not a simulation of authority enforcement and should not produce a readiness score. Its purpose is to expose missing evidence, unclear authority and actions that could harm a worker or release an unreviewed product.

The proposed scenario combines a changed dye recipe, missing upstream records, a worker allegation and an excessive reporting request. Each event has a different decision path. REACH restrictions can apply to imports, so a changed material or finish should trigger product-specific scope and evidence review rather than an administrative assumption that the old test still covers the new state.[S16] The reviewer must determine the actual substance question, sample and batch relationship from the operative requirements.

Missing upstream records require containment and reconstruction. The team should identify affected input lots, processing events and finished batches, preserve the available record and name what cannot be established. A missing record is an evidence gap, not proof of abuse. It is also not a zero-risk answer. The forced-labour framework's distinct assessment and investigation stages reinforce the need to separate a concern, an internal review and an authority finding.[S10]

The worker allegation changes the priority. Protection, confidential intake and escalation outside the production deadline must come before evidence packaging. The ILO and IOE handbook uses forced-labour indicators to help companies identify and assess risks in practice.[S09] An indicator starts enquiry; it does not license a public accusation. The drill should record who can take safeguarding action, how the affected person can confirm remedy and which information must remain restricted.

The reporting request tests a separate control. The team should ask its purpose, check what can be reused and distinguish product evidence from corporate reporting. The adopted amending text gives protected undertakings a right, subject to its conditions, to decline information beyond the voluntary standards for specified sustainability-reporting requests.[S35] That mechanism should not be stretched into a refusal to supply necessary safety or investigation evidence. National implementation and the applicable standard version remain review dependencies.

Exhibit 59 sets out the rehearsal register. Its result cells deliberately state that the drill has not been run. Retrieval times, decision quality and remediation effectiveness must be recorded only during an actual exercise. A completed template is not a completed drill.

A facilitator should introduce the events in sequence and preserve the decisions made at each point. The changed recipe may reduce the eligible batch before the worker allegation arrives. The reporting request may distract staff while the upstream record remains missing. This interaction is the useful stress: the drill shows whether urgent protection and product control retain owners when administrative work competes for the same people.

The observation sheet should record elapsed retrieval time, active work, unresolved evidence, decision-maker and protective action. It should also note false starts, inaccessible records and ambiguous instructions. Those observations become pilot costs and design changes. They must not be backfilled with target values after the exercise. If no retrieval occurred because a record was absent, the honest entry is unavailable, not an unusually fast response.

Closure requires more than a meeting. Each gap needs an owner, resources and a review condition. A corrected template addresses an administrative defect only if the revised route works on the next exercise. A worker-related action needs appropriate confirmation and recurrence review. A product hold needs an authorised release decision based on the actual evidence. The drill budget should include these follow-up actions, otherwise the exercise becomes a performance without operational change.

The final decision should compare accepting, conditionally accepting, repricing or declining the order. Acceptance requires both economic headroom and a clear release path. Conditional acceptance may be appropriate when evidence work, buyer clarification or product review has a funded owner and a feasible completion point. Repricing may address identifiable incremental work. Declining may be necessary when the route cannot meet critical conditions or the commercial terms cannot fund the required response.

A positive contribution never overrides a critical failure. Product safety, applicable chemical restrictions and worker protection are not optional enhancements to be traded against price. The order model should therefore present two outputs side by side: an economic calculation and a gate register. A high margin beside an unresolved critical gate means do not release. A low administrative cost beside complete evidence does not, by itself, prove that the order is attractive.

The illustrative case shows why negotiation should focus on the work driver. Removing duplicated administration improves the scenario without claiming to reduce necessary testing or remedy. Repricing can be evaluated against the same transparent lines. Yet if the changed dye recipe still lacks the required release evidence, neither version authorises shipment. The financially preferable route remains conditional on the gate.

The investment memorandum should attach actual buyer terms, the verified scope decision, external quotations, internal-hour observations and the drill register. Assumptions should remain in separate sensitivity cells. A premium can be included only as a buyer-supported commercial input, not as a reward automatically generated by compliance. Financing approval, EU market access and buyer acceptance are separate decisions that this model cannot promise.

The decision record should close with the unresolved questions as well as the recommendation. Who confirmed the legal scope? Which evidence remains missing? Which costs are quotations and which are estimates? What did the rehearsal fail to retrieve? What must change before the next order? Those answers make the economics answerable to the order rather than making the order serve a predetermined compliance story.