---
title: "How much effort should each RFP weighting receive?"
description: "Convert published RFP weightings into a bounded response-effort budget, then document justified overrides for thresholds, evidence gaps and dependencies."
canonical: "https://zephior.com/insights/understand-rfp-evaluation-weightings"
last-updated: 2026-09-03
---

# How much effort should each RFP weighting receive?

> Convert published RFP weightings into a bounded response-effort budget, then document justified overrides for thresholds, evidence gaps and dependencies.

By [Tony Kim](https://zephior.com/authors/tony-kim). Published 2026-09-03; updated 2026-09-03. 24 minute read.

## Definition

A tender_evaluation_effort_allocation is a versioned working record that distributes an already-approved response budget across confirmed award criteria, subcriteria and related response work. Each row preserves the buyer weight or rank, method, threshold, source, response surface, baseline effort share, minimum control, evidence condition, bounded risk override, final budget, owner and review trigger. The record plans bidder work. It does not change the buyer method, predict a score, estimate the total bid, choose the offer or decide whether to bid.

## Problem

A percentage looks like a ready-made staffing rule. It is not. Giving a 30% criterion exactly 30% of the hours can leave a low-weight threshold under-evidenced, waste scarce specialists on a mature answer, or ignore the price workbook because its narrative is short. The opposite habit is just as weak: teams call every topic critical, route experts to whoever asks first and lose any connection to the published evaluation. A useful plan needs a stable baseline from the buyer method and a disciplined way to move effort only when evidence shows why.

## Point of view

Use published relative importance as the starting distribution for discretionary response work, never as a promise that hours equal points. Protect fixed compliance, production and submission work before dividing the remaining budget. Adjust that baseline for explicit thresholds, current evidence gaps, shared dependencies, work already complete and scarce review paths. Every transfer must name where the effort came from, what condition justifies it and when the decision expires. If the source method is missing or contradictory, stop the allocation at the affected level rather than inventing a private weighting.

## Preserve the buyer’s method before assigning one hour

Start with the procurement that will actually be evaluated. Record the procedure, stage, lot, current document set and any formal clarification that changes assessment. Then capture each award criterion and material subcriterion with its exact label, published relative importance, score scale, threshold, formula, response instruction and source locator. A criterion in a notice may be expanded in an invitation or schedule. The allocation cannot be safer than the source chain beneath it.

Keep unlike things apart. A condition of participation tests whether a supplier may take part. A mandatory requirement may determine whether a tender proceeds. An award criterion compares the merits of tenders. Price, a demonstration and a written quality response may also follow different calculation paths. UK Cabinet Office guidance under the Procurement Act 2023 expressly distinguishes conditions of participation from award criteria and says assessment follows the published methodology and relative importance. That is a rule for the buyer’s assessment, not a bidder timesheet formula.

Relative importance does not always arrive as percentages. The UK framework permits weighting, ranking or another description. Directive 2014/24/EU Article 67 permits a range and, where weighting is not objectively possible, decreasing order of importance. FAR 15.304 requires applicable factors, significant subfactors and their relative importance to be stated clearly, while saying the rating method need not be disclosed. TED eForms can represent a weight, fixed value, threshold, order of importance or a more complicated expression. Preserve the form that was issued. Converting “technical is significantly more important than price” into 70:30 would create evidence the buyer never published.

Give every extracted value a type. Useful values include exact_weight, weight_range, ordinal_rank, narrative_relative_importance, fixed_value, threshold, complicated_expression and unresolved. Preserve the issued unit and denominator too. Twenty points out of the technical score is not necessarily 20% of the final award. A 60% technical envelope with a 20% technical subcriterion gives that subcriterion 12% of the combined result only if the documents actually use nested multiplication in that way.

When figures conflict, do not average them or select the most convenient schedule. Mark source_conflict and route it to the controlled source and clarification process. When only a rank is known, keep the rank. When a weighting range is published for a later stage, record the range and the event that will select the final value. Allocation can continue for unaffected rows, but the uncertain slice remains a reserve or bounded scenario rather than a fabricated point estimate.

**Evaluation facts that must survive into the effort plan**

| Field | Example value | Planning rule |
| --- | --- | --- |
| Criterion identity | Q2 implementation, lot 3 | Never merge by similar title |
| Importance type | exact_weight: 25% | Keep the buyer’s form |
| Denominator | 25 of 70 quality points | Do not assume 25 of the final 100 |
| Threshold | At least 3 of 5 | Create a floor and route the gate |
| Response surface | Schedule 4, section 2.1 | Budget the required object, not just prose |
| Evidence state | Current design, proof missing | Use only as an override input |
| Source state | Current, amended, unresolved | Block only the affected calculation |

## Separate fixed work, discretionary work and contingency

AN-076 starts after leadership has approved the total response envelope. If the bid has 260 staff-hours, this dossier does not decide whether 260 is enough or whether the pursuit should continue. Those questions belong to effort estimation, capacity and bid governance. Here, the task is to keep all 260 hours visible and assign the portion that can sensibly follow evaluation influence.

Take fixed work out first. File production, final compliance review, pricing validation, signatures, accessibility checks, portal rehearsal and submission evidence often need effort regardless of their apparent score weight. The same is true for minimum qualification work. If these tasks compete anonymously inside a criterion budget, they are usually funded too late. Give each a named package, owner and closure proof. AN-077 owns the complete pass/fail gate register; this allocation only carries the hours needed to reach an approved gate state.

Next define discretionary criterion work. It includes research, solution decisions, evidence retrieval, drafting, design, specialist review and red-team work that can be linked to an assessed response. Do not equate it with page count. A two-cell price model can require more commercial control than a six-page method statement. A mature ten-page case study may need only a scope check. Effort follows the work required to make the response assessable and supported.

Keep a small contingency only if the approved plan contains one. The reserve is not free labor and should not become a place to hide unassigned work. State its permitted triggers, such as an expected clarification, pending buyer template or scheduled offer decision. If no contingency was approved, transfers must net to zero immediately. A request for more total effort returns to the authority that owns the response envelope.

The resulting equation is simple: total approved effort equals fixed control work plus discretionary criterion work plus named contingency. Every row uses the same unit, normally staff-hours or half-days. Mixing hours, story points and vague labels such as “high attention” makes reconciliation impossible. Calendar elapsed time and waiting time stay in the schedule; they are not labor allocations.

**Response effort envelopes**

| Envelope | What belongs here | What does not |
| --- | --- | --- |
| Fixed controls | Required review, production and submission work | Optional polishing |
| Discretionary criteria | Work linked to assessed response quality or price | Buyer-only evaluation activity |
| Named contingency | Approved response to defined unresolved events | Unowned scope or routine overruns |
| Outside AN-076 | None | Total bid estimate, no-bid decision or portfolio capacity |

## Use weighting as a starting share, not a staffing command

For an exact percentage model, calculate a baseline only across the rows that share a meaningful denominator. If 224 discretionary hours cover price and all scored quality criteria, a criterion carrying 25 of the final 100 points starts with 56 hours. The calculation is 224 multiplied by 0.25. Keep the unrounded result for reconciliation, but plan in the work unit the team can use. Rounding differences go to an explicit reconciliation row, not whichever criterion happens to be last.

A nested model needs two steps. Suppose quality is 70% of the award and implementation is 25% of the quality score. If the documents say subcriterion weights are proportions within quality, implementation carries 17.5% of the combined result. The baseline can use 0.70 multiplied by 0.25. If “25 points” already refers to the combined total, multiplying again is wrong. Store buyer_weight, weight_level, denominator and normalized_weight so another reviewer can reproduce the result.

Baseline is deliberately mechanical. It stops the loudest stakeholder from claiming the whole budget and exposes the cost of each change. It is not the final answer because points and labor describe different things. Buyer weighting describes evaluation influence. Bidder effort describes the work needed from the current response state. A criterion that is already supported can donate hours; a lower-weight criterion with an unsolved design and a published threshold may need them.

Price belongs in the model when price influences the award, even if most pricing work occurs in a workbook. Record modelling, validation, approval and reconciliation rather than inventing prose. Where the commercial work budget is fixed by another approved plan, link that package and remove it from the discretionary denominator. State which treatment you used. Otherwise the allocation can count price twice or pretend it costs nothing.

Ordinal and narrative models need a different baseline. Create importance bands that preserve every published ordering constraint. The top-ranked criterion must not receive less baseline attention than a lower-ranked one without an approved override. Do not attach invented percentages. Allocate whole work units through a documented comparison, keep a range where several distributions satisfy the source, and show the unresolved reserve. AN-076 can plan under uncertainty; it cannot erase uncertainty.

**Minimum reproducible allocation fields**

| Field | Meaning | Example |
| --- | --- | --- |
| buyer_weight | Issued importance value and unit | 25% within quality |
| normalized_weight | Comparable share proved from the method | 17.5% of final award |
| baseline_effort | Mechanical starting work budget | 39.2 hours |
| minimum_control | State required before release | Reviewed threshold evidence |
| risk_override | Bounded transfer with cause | +8 hours from price validation |
| final_budget | Approved usable work allocation | 47 hours |
| stop_condition | Evidence that ends more work | Design approved and red-team test passed |

## Move effort for evidence, thresholds and dependencies

An override is a transfer, not an adjective. Record the recipient, donor, number of work units, reason, evidence, approver and expiry trigger. “Security is important” proves nothing because importance is already represented by the buyer model. “Move eight hours from the accepted price workbook to security architecture because the required data-flow evidence is missing and the criterion has a 3-of-5 minimum” is a decision another person can test.

Use a short set of evidence classes. A published threshold creates a minimum-control need. An evidence gap exists when the offered capability is decided but the required proof is unavailable or stale. A decision dependency exists when drafting cannot become accurate until product, legal, delivery or commercial owners choose something. A shared dependency deserves one work package linked to all affected rows. A scarce specialist path may justify earlier review effort because waiting until the end would strand the response.

Credit work already done. Approved reusable evidence, a completed buyer template or an answer that has passed the defined review can reduce remaining effort. Maturity must be proved against the same entity, offer, criterion, version and disclosure permission. Similar-looking text is not completion. The donated hours remain visible, including the acceptance evidence and the destination that receives them.

Do not use win probability, imagined competitor quality or an unsupported target score as an override. Those numbers can make any transfer look rational. Strategy decisions may change the work required, but AN-076 receives the approved offer choice and its consequences rather than choosing it. Likewise, page limits shape production burden but do not reveal evaluation value. A short response can still need substantial analysis and proof.

Cap each override at the smallest amount needed to reach the next named state. Review again when that state is reached. This prevents an evidenced risk from becoming an unlimited claim on resources. A security architect may need six hours to approve the data flow and two more for response review. Funding an open-ended “security workstream” gives no stopping point and no way to return unused time.

**Permitted override reasons**

| Reason | Required evidence | Bounded destination |
| --- | --- | --- |
| Published threshold | Source, value and stated consequence | Work needed to reach reviewed floor |
| Evidence gap | Required proof and current absence | Retrieve, validate and approve evidence |
| Decision dependency | Named unresolved decision and owner | Decision preparation and propagation |
| Shared work | Dependency links to several rows | One package with allocated consumers |
| Scarce reviewer | Availability and required review scope | Booked review plus correction allowance |
| Accepted maturity | Approval tied to current scope | Negative transfer to named recipient |

## Fund the floor, then stop when the next hour lacks a job

A low weight can still carry a hard boundary. The buyer may require a minimum score on one criterion, a combined quality threshold, a complete schedule or a mandatory demonstration. Record the exact source and stated consequence. Give the affected response enough effort to reach a reviewed minimum state before optional polish elsewhere. The pass/fail dossier determines whether the gate is satisfied; this plan only reserves and releases the work.

High weight does not remove diminishing returns. Once the offer is approved, every subpart is answered, each material claim has applicable proof and the rendered response has passed its criterion review, another rewrite may add less value than resolving a medium-weight evidence gap. Define a stop condition in observable terms. “Good enough” is not observable. “All five requested elements mapped, architecture approved, two claims cited, cross-document figures reconciled and red-team findings closed” is.

Use a marginal work queue after each floor is secure. Every proposed task names the defect it will remove, the criterion it affects, the owner and the completion test. Rank tasks by published influence and the severity of the unresolved response condition. This is not a score simulation. It is a comparison of visible work: a missing answer element normally precedes stylistic tightening; an unsupported claim precedes a new graphic; an unresolved commercial assumption precedes another executive edit.

Protect the final review path. A response is not finished when drafting hours are consumed. Hold the review allowance until the required owner has inspected the final rendered object, including tables, attachments and cross-references. If an upstream decision arrives late, reopen only the rows and shared work it affects. Do not finance the correction by silently taking time from production or submission control.

Unused effort returns to the response envelope. It is not a reward for filling the page. The bid manager may transfer it to the next supported defect, restore contingency or return it to portfolio capacity under the approved governance. Record the disposition so actual-versus-plan data remains useful for the next tender.

## A 260-hour university platform response does not mirror 70:30

Consider a fictional university procurement for a research-data platform. The tender assigns 70% to quality and 30% to price. Quality is divided into implementation at 25 points, security and data governance at 20, service management at 15, and accessibility plus sustainability at 10. The documents state that security must score at least 3 of 5. The team has approved 260 staff-hours for the response. Thirty-six hours are protected for compliance, production, accessibility of the submitted files, approval and portal rehearsal, leaving 224 discretionary hours.

A pure weighting baseline gives 67.2 hours to price, 56 to implementation, 44.8 to security, 33.6 to service management and 22.4 to accessibility plus sustainability. Those decimals are calculation evidence, not a sensible work schedule. The team rounds only after examining the current response. The price model is already built and needs validation rather than redesign. Implementation has an unresolved migration sequence. Security lacks an approved data-flow diagram and has the published threshold. Service management can reuse accepted service-level evidence. The final criterion needs fresh accessibility testing but has a strict four-page limit.

The approved plan assigns 40 hours to price, 66 to implementation, 56 to security, 35 to service management and 27 to accessibility plus sustainability. The total remains 224. Each change has a donor and recipient. Price donates 27 hours after the finance owner confirms the existing model, assumptions and review path. Implementation gains 10 hours for the migration decision and propagation. Security gains 11 hours for architecture evidence and threshold review. Service gains one hour after rounding and dependency reconciliation. The last criterion gains five hours for testing and evidence selection.

These allocations do not claim that the university will award a particular score. They describe the current work required to produce an assessable response under the published method. If the price assumptions change, the 40-hour commercial budget reopens. If the buyer removes the security threshold, the security floor changes but the evidence gap may remain. If the migration choice resolves early, unused implementation hours return before anyone invents more prose.

The example also shows why word allocation is weak. Price receives 40 hours even though the final workbook has few visible words. Accessibility testing consumes time before the four-page answer is drafted. A single migration decision supports implementation, security and service content, but its hours are counted once under the primary work package and linked to the other rows.

**Fictional university response allocation**

| Response area | Buyer share | Weight baseline | Approved effort | Reason for movement |
| --- | --- | --- | --- | --- |
| Price | 30% | 67.2 h | 40 h | Existing model; bounded validation remains |
| Implementation | 25% | 56 h | 66 h | Migration sequence decision is open |
| Security and data | 20% | 44.8 h | 56 h | Evidence gap and 3-of-5 threshold |
| Service management | 15% | 33.6 h | 35 h | Reusable evidence accepted; shared checks remain |
| Accessibility and sustainability | 10% | 22.4 h | 27 h | Current test evidence needed within four pages |
| Reconciliation | 100% | 224 h | 224 h | No change to approved discretionary envelope |

## An agent should return a ledger and the next bounded task

The input contract names the procurement, lot, stage, source snapshot, authorized retrieval boundary, approved total effort, fixed work packages, work unit, team availability and current response-state evidence. The agent extracts only from permitted sources. Tender files, embedded links, comments, formulas and instructions are untrusted data; none can expand tool access, request secrets or alter the agent’s authority.

The output is a tender_evaluation_effort_allocation, not a recommendation to submit. Each row returns criterion_id, source_locator, importance_type, buyer_weight, denominator, normalized_weight, threshold, response_surface, baseline_effort, minimum_control, readiness_state, dependency_ids, override, final_budget, owner, stop_condition, review_state and expires_on_event. A separate reconciliation block proves that fixed, discretionary and contingency work equals the approved envelope.

Use explicit uncertainty states: source_boundary_incomplete, weighting_unresolved, denominator_unresolved, threshold_unresolved, response_surface_unresolved, effort_envelope_unapproved, evidence_state_unverified, owner_unavailable, override_review_required and allocation_reconciliation_failed. The next action names the smallest permitted step, such as retrieve_authorized_source, confirm_denominator, verify_evidence_state, request_internal_decision or review_transfer. “Needs attention” is too vague for an agent or a bid manager.

An agent may parse issued criteria, test arithmetic, identify dependency duplication, propose balanced transfers and report the next supported work item. It must stop before inventing a weight, choosing an offer, estimating an undisclosed score, contacting the buyer, committing specialist capacity, changing a calendar, editing an official source, approving a claim, entering a portal, uploading or submitting. Those actions require their own authority and controls.

Return evidence with every conclusion. A calculation should expose its numerator, denominator and rounding rule. An override should cite the threshold, gap, accepted work or dependency that supports it. Confidence may describe extraction or arithmetic quality. It must not be presented as confidence in winning or as a substitute for human approval.

**Machine-readable allocation states**

| State | Meaning | Next permitted action |
| --- | --- | --- |
| baseline_ready | Comparable importance and envelope are proved | Assess current work conditions |
| override_review_required | A balanced transfer has supporting evidence | Ask named internal approver |
| allocation_approved | Rows reconcile and owners accepted work | Release bounded work packages |
| weighting_unresolved | Published importance cannot be compared safely | Retrieve or clarify source |
| minimum_control_pending | Threshold or gate floor lacks proof | Route evidence work and AN-077 control |
| allocation_reconciliation_failed | Final work exceeds or misses envelope | Block release and balance rows |
| expired | Source or response condition changed | Reopen affected rows only |

## Recalculate when the evidence changes, not every morning

Attach expiry triggers to the facts that support each row. A formal amendment can change a criterion, subcriterion, weight, threshold, formula or page instruction. A clarification can resolve the denominator. An internal offer decision can remove a dependency. A failed evidence check can reopen work thought complete. A reviewer can expose a missing subpart. Each event affects a defined slice of the ledger; none justifies discarding the entire plan automatically.

Keep plan and actual effort separately. Actual hours reveal where work was underestimated, but they do not retroactively justify the allocation. Record work class and result: research completed, evidence rejected, decision approved, answer reviewed, defect reopened. A team that only records total hours learns that the bid was expensive. A team that records the work state can improve the next approved estimate and recognize reusable evidence.

AN-071 owns answer and format limits, AN-072 submission objects, AN-075 the bidder-action timetable and AN-077 pass/fail gates. The broader effort-estimation guide owns the total response forecast. The evaluation-strategy dossier owns offer choices, intended strengths and proof strategy. The ambiguity dossier owns clarification of missing or conflicting scoring information. AN-076 owns only the balanced allocation of an approved response budget against the confirmed evaluation model and visible work condition.

Release the ledger as allocation_approved only when the source boundary is complete for the declared scope, all material rows have an owner and stop condition, fixed work is protected, transfers reconcile, thresholds have a minimum control, and the approver accepts the current evidence. Otherwise publish the exact bounded state and owner. A partial, inspectable allocation is safer than a complete table built from invented weights.

## Useful outcomes

- Every allocation row points to the exact criterion, method, version and response surface it covers.
- Published weights remain separate from bidder-created effort budgets and planning judgments.
- Fixed compliance, production and submission controls are funded before discretionary criterion work.
- Low-weight thresholds and high-consequence evidence gaps receive a visible minimum control.
- Hours move between criteria only through a recorded source, destination, reason and approver.
- Shared evidence and specialist work are counted once and linked to every dependent answer.
- Authors and reviewers know when further work has lower value than an unresolved upstream decision.
- An agent can explain the next permitted work item without pretending to know the evaluator’s score.

## Workflow

1. **Freeze the evaluation snapshot.** Record the procedure, lot, tender version, criteria, subcriteria, method, thresholds and response instructions before calculating any share.
2. **Protect non-discretionary work.** Reserve approved effort for compliance control, production, approvals and submission tasks that cannot sensibly be assigned by award weight.
3. **Calculate a traceable baseline.** Apply the published relative importance to the remaining response budget only at levels where the buyer method is known and comparable.
4. **Assess work conditions.** For each row, inspect thresholds, evidence readiness, response burden, dependencies, specialist scarcity and work already accepted.
5. **Approve balanced overrides.** Move bounded work units from named donor rows to named recipient rows, leaving the total response envelope unchanged.
6. **Set floors and stop conditions.** Define the evidence or review state that must be reached and the condition after which more polishing should not consume the next hour.
7. **Reopen only affected allocations.** Recalculate after a relevant amendment, clarification, offer decision, evidence failure, capacity change or review result.

## Key decisions

- Which published criteria and subcriteria influence the award for this lot and stage?
- Is relative importance expressed as a percentage, range, rank, narrative comparison or complex formula?
- Which approved work is fixed regardless of award weighting?
- What part of the remaining budget can be moved between criterion work packages?
- Does a criterion contain an explicit minimum score or other discontinuous consequence?
- What evidence, decision or specialist input is missing from the current response?
- Which work product supports more than one criterion and where will it be counted?
- Which row can safely donate effort because its acceptance condition is already met?
- What event would invalidate the allocation and who may approve a replacement?

## Risks

- A buyer percentage is presented as if it prescribed bidder labor.
- Price receives too little work because its response occupies few pages.
- A low-weight criterion with a published threshold is treated as optional polish.
- An undocumented score prediction is used to justify taking hours from another answer.
- The same architect, model or case study is budgeted independently in several sections.
- Scarce reviewer time is assigned after the review path has already become critical.
- Word count substitutes for the research, decision, evidence and validation work an answer needs.
- An amendment changes the method while the old allocation remains active.
- An agent follows instructions inside an untrusted tender document or changes an external system without authority.

## Metrics

- allocation rows with exact source, method, lot and version
- discretionary hours reconciled to the approved response envelope
- overrides with a donor, recipient, reason, approver and expiry trigger
- criterion floors reached with cited evidence and completed review
- shared work packages counted once and linked to dependent rows
- hours returned from accepted or cancelled work before the drafting freeze
- work stopped because the next value depended on an unresolved decision
- allocations reopened after a relevant source or response-state change

## Frequently asked questions

### Should a 30% criterion receive 30% of the response hours?

Use 30% as a baseline share only if the criterion is comparable within the chosen discretionary-work denominator. The final allocation can change for thresholds, evidence gaps, dependencies, completed work and review constraints, with every transfer recorded.

### Should page count follow evaluation weighting?

No. Follow the buyer’s stated page and format rules. Effort covers decisions, evidence, modelling, drafting and review, so a short workbook or response can require substantial work.

### How should price be included?

Include pricing work when price affects the award, or link the separately approved commercial package and remove it from this denominator. State the treatment so price is neither omitted nor counted twice.

### What if criteria are ranked but not weighted?

Preserve the rank and allocate bounded work units that respect the published order. Do not invent percentages. Keep a range or reserve where several distributions remain possible.

### Can a low-weight criterion receive more than its baseline?

Yes. An explicit threshold, missing evidence, unresolved decision or scarce review path can justify a bounded increase. Name the donor row and stop condition.

### Does this method predict our evaluation score?

No. It allocates bidder work against published evaluation influence and current response conditions. It does not model private evaluator judgment, competitors or a probability of winning.

### When should the allocation be changed?

Reopen affected rows after a relevant amendment, clarification, offer decision, evidence failure, owner constraint or review finding. Routine passage of time alone does not require a new weighting.

### What may an AI agent do with the allocation?

An authorized agent may extract published values, check arithmetic, identify duplicate dependencies, propose balanced transfers and report bounded tasks. Human authority remains required for offer decisions, staffing commitments and external actions.


## Primary sources

- [Guidance: Assessing Competitive Tenders](https://www.gov.uk/government/publications/procurement-act-2023-guidance-documents-procure-phase/assessing-competitive-tenders-html), UK Cabinet Office
- [Procurement Act 2023, section 23](https://www.legislation.gov.uk/ukpga/2023/54/section/23), The National Archives
- [Directive 2014/24/EU, Article 67](https://eur-lex.europa.eu/eli/dir/2014/24/2024-01-01/eng/pdf), EUR-Lex
- [TED eForms business terms](https://docs.ted.europa.eu/eforms/latest/reference/business-terms/index.html), Publications Office of the European Union
- [TED eForms schema usage for award criteria](https://docs.ted.europa.eu/eforms/latest/schema/all-in-one.html), Publications Office of the European Union
- [FAR 15.304, Evaluation factors and significant subfactors](https://www.acquisition.gov/far/15.304), Acquisition.gov
- [FAR 15.305, Proposal evaluation](https://www.acquisition.gov/far/15.305), Acquisition.gov
- [German GWB section 127, award](https://www.gesetze-im-internet.de/gwb/__127.html), Federal Ministry of Justice and Federal Office of Justice
- [German VgV section 58, award criteria](https://www.gesetze-im-internet.de/vgv_2016/__58.html), Federal Ministry of Justice and Federal Office of Justice
- [French Public Procurement Code, Articles R2152-6 to R2152-12](https://www.legifrance.gouv.fr/codes/id/LEGISCTA000037730491), Légifrance
- [Evaluating Bids and Proposals with Rated Criteria, February 2025](https://thedocs.worldbank.org/en/doc/9dcb7971706bf29b2732779c39922b77-0290012025/original/Evaluating-Bids-and-Proposals-with-Rated-Criteria-Feb-4-2025.pdf), World Bank
- [Procurement Regulations for IPF Borrowers, seventh edition, September 2025](https://thedocs.worldbank.org/en/doc/c84273d1b230aeb2b0b8134de5dc8cd7-0290012025/original/Procurement-Regulations-7th-Edition-Sep-2025.pdf), World Bank


## Related articles

- [Turn tender evaluation criteria into a response strategy](https://zephior.com/insights/turn-evaluation-criteria-into-response-strategy)
- [Which bids should receive scarce proposal capacity?](https://zephior.com/insights/manage-capacity-across-a-bid-portfolio)
- [How to respond when price dominates tender evaluation](https://zephior.com/insights/respond-to-price-dominant-evaluation)
- [How should a tender scoring formula change the answer plan?](https://zephior.com/insights/translate-a-scoring-formula-into-an-answer-plan)