An RFP work breakdown divides the production of a tender response into bounded work packages with identifiable outputs. Each package states its source or internal release purpose, included and excluded work, required inputs, accountable owner, acceptance test and downstream uses. The accompanying dictionary explains those boundaries. A traceability map then connects the packages to the buyer’s asks and submission destinations. It can show one shared evidence product supporting several answers without pretending that those answers, their reviews and their approvals are the same work.

The question workbook has an owner in every row, so the bid appears assigned. Two writers then ask different teams for the same test report. Neither asks who may release the report or decide the fallback approach it supports. Both answers become complete drafts while the pricing assumption still describes another solution. The question list is covered with names, but the evidence, decision and reconciliation work has no defined output or completion test.

Define the work product before estimating its hours or placing it on a calendar. This playbook starts from an already controlled and interpreted question set; the subpart-coverage guide owns that earlier analysis. Here the result is a bid-production work breakdown, not the delivery WBS that a buyer may request inside the proposal. Effort estimation, detailed owner selection and critical-path scheduling use this result as an input. All organisations, question labels and work records in the examples are fictional. Sources were checked on 6 September 2026; their project-management concepts are adapted here, not presented as universal procurement requirements.

The product being planned is the response you must produce

Start with a dated scope statement for the bid-production effort. Identify the bidder entity, lots, response stage and submission objects. Attach the controlled question set and the instructions that govern it. If the buyer asks for an implementation work breakdown, producing that proposed breakdown is one bid deliverable. Installing the equipment after award belongs to the future contract plan. Combining both in one task list makes the bid workload impossible to interpret.

The NASA WBS Handbook, November 2021 revision, describes a product-oriented hierarchy and a companion dictionary defining work content and interfaces. Its scope is NASA project management. The adaptation here is to treat the bid response as the product and define its production work, without importing NASA’s mandatory project structure into an unrelated tender.

Use the whole controlled package, not only rows ending in a question mark. German VgV section 29(1) describes procurement documents as normally including invitations, procedural conditions and contract documents. In an applicable German procedure, that is a reason to inspect the complete issued set. It does not mean every contract duty must be performed before bidding.

Required certificates, samples, pricing files and internal release decisions may therefore enter the work breakdown from different sources. Give each a source reference or a stated internal purpose. An internal file check does not need an invented buyer question number. It does need a reason to exist, a defined output and a place in the controlled scope.

Write the completion test before assigning the task

Name a package by the output it must deliver: reviewed test extract, approved recovery choice, completed price-assumption row or checked submission file. Under that output, list the activities required to create it. Retrieve, compare, redact, review and approve are activities; an evidence packet cleared for a defined use is an output. A single label such as handle security hides too much to assign or accept.

The GAO Schedule Assessment Guide distinguishes the deliverables defined by a WBS from activities explaining how they will be produced. Its dictionary clarifies element boundaries and associated scope. Applied to a bid, this supports a separate description of what a package produces and what work is included, before the schedule assigns dates.

For each package, state what is included and what is explicitly outside it. An evidence package may include extracting and validating a historical result and obtaining permission for a specified excerpt. It need not include writing every answer that cites that result. A writer may adapt the released excerpt for the question, but cannot broaden its factual scope or disclosure permission. Those boundaries prevent two owners from each assuming the other performed the missing check.

Choose a level of detail at which an owner can deliver a result and a reviewer can accept it. Split a package when a separate authority, external handoff or independently usable output needs its own control. Keep small editing steps together when splitting would add administration without changing responsibility or evidence. There is no universal rule that every question needs five tasks, or that every paragraph deserves its own package.

A package dictionary that can support assignment and review
FieldRequired definitionFailure it exposes
Identity and parentStable package ID and one owning branch of the breakdownThe same work counted beneath several answers
Purpose and scopeSource asks or release purpose, included work and exclusionsAn unowned gap between two broad tasks
InputsRequired object, version, status and supplying package or partyA draft or missing input treated as approved
Output and acceptanceInspectable artifact or decision and the tests for accepting itCompletion based only on effort spent
ResponsibilityAccountable producer, required reviewers and decision authoritiesAssignment mistaken for permission to approve
Consumers and changeDownstream uses, applicability and reopening eventsA changed input leaving old answers marked current

One shared product can have several consumers without several owners of its production

The work hierarchy and the question-coverage map answer different questions. The hierarchy shows where production responsibility sits. The coverage map shows which asks and destinations use the resulting output. Put a shared evidence product in one branch, then link it to each consuming answer. Do not make a physical copy of the same production task under every question merely to make the hierarchy resemble the workbook.

Reuse needs a scope test. Two answers can use the same historical report while requiring different explanations, excerpts or permissions. If they need the same released object, share that object and retain separate consumption checks. If one use requires a different population, period or recipient permission, define the additional work explicitly. Similar filenames are not enough to prove that the work is interchangeable.

Work also runs in the opposite direction: one ask can need several outputs. A question asking for an implementation approach with evidence may depend on a selected solution, a current resource assumption, an approved example and an answer assembled within the buyer’s limit. Link each contribution to the ask it enables. The subpart-coverage record still controls whether the eventual answer performs every requested instruction.

Keep hierarchy and dependency links distinct. A package can belong under technical response while depending on a decision produced in commercial preparation. That dependency does not change its parent or transfer its accountability. Shared labels such as related to are too vague for this purpose; identify whether a link means contains, produces, uses, answers or requires approval.

Elmshore needs seven packages behind two answers and a price assumption

Fictional Elmshore Records has a controlled fragment of a tender response: Q4 requests a migration method with dated rehearsal evidence and a fallback explanation; Q9 requests a continuity answer using the same relevant rehearsal evidence; K2 is the required price-assumption field for the selected fallback arrangement. Assume the question analysis, source completeness and permitted disclosure routes are already established. This fragment is not presented as the entire tender workload.

The work definition contains seven leaf packages. E1 produces the reviewed and disclosure-cleared rehearsal extract. D1 produces the authorized fallback choice, using E1 and separately identified feasibility inputs. A4 and A9 produce the two reviewed answers. C2 produces the reconciled price-assumption entry with its required commercial review. V1 checks consistency across those three response outputs. P1 produces the checked submission objects for this fragment after V1. Each package has its own acceptance condition.

For management, E1 and D1 sit under evidence and decisions; A4, A9 and C2 sit under response outputs; V1 and P1 sit under release preparation. That is two plus three plus two, or seven leaf packages. The three parent headings add no extra production effort. E1 has two direct answer consumers, and D1 has three direct response consumers, but each is still produced once. The dependency links do not increase the leaf count.

The initial board says that A4 and A9 are drafted. D1, however, is waiting for the competent service owner’s decision. Both drafts must retain that limitation; neither answer can pass its acceptance test, and C2 cannot be accepted against an unresolved choice. Mark the specific input as missing. Do not report the fragment as nearly complete merely because the two longest pieces of prose exist.

E1 is evidence of the disclosed rehearsal under its recorded conditions. It is not proof that the future offer will necessarily achieve the same result. D1 needs the appropriate feasibility and commitment decision. If those inputs do not exist, add or identify the work that produces them before calling the breakdown complete. The seven-package example assumes those external inputs are already identified; it does not make them disappear.

Elmshore’s seven leaf packages, with review inside each defined scope
Package and outputInputs required for acceptanceCompletion proof
E1: released rehearsal extractIdentified source, factual scope and disclosure permissionReviewed version with permitted uses recorded
D1: authorized fallback choiceE1 and the identified feasibility and authority inputsDecision on exact approach, conditions and responsibility
A4: reviewed Q4 answerE1 and D1All Q4 asks answered within the prescribed destination
A9: reviewed Q9 answerE1 and D1Continuity explanation checked for its own question and use
C2: reconciled K2 entryD1 and current approved pricing basisPrice assumption matches the selected approach
V1: cross-output consistency recordA4, A9 and C2 at their reviewed versionsNo unresolved material conflict across the three outputs
P1: checked production setA4, A9, C2 and accepted V1Required files and fields match the approved fragment

The work outside the answer box still needs a place

Canada’s current guidance on preparing a bid asks suppliers to inspect pre-closing deliverables and provide the supporting information required by the solicitation. Its examples include samples, certificates and past-project descriptions. Use the actual tender requirements to identify such work. A completed question workbook cannot establish that a separately required physical or supporting item exists.

French Service Public Entreprendre guidance explains that buyers can prescribe a technical-response framework, page limits, tables or restrictions on annexes. Those constraints can create production and verification work beyond drafting the content. Import the applicable instructions; do not invent an extra annex to compensate for a work package that has no permitted submission destination.

Conditional work needs a visible branch and an applicability decision. A partner declaration may be required only for a particular delivery model. Keep the model decision separate from the work needed under each outcome. Until the authorized configuration is established, the inactive-looking branch is unresolved, not safely omitted. Once resolved, retain the basis for excluding the unused branch and reopen it when the configuration changes.

Include necessary coordination and release work without inventing elaborate ceremonies. A defined reconciliation record or a verified file set is inspectable. Attend weekly meeting is an activity that may support that work, not proof the output is complete. If ongoing coordination is needed, state its scope and records and keep it separate from discrete deliverables when later estimating effort or reporting progress.

Name the input that makes the next package usable

A dependency should identify the object and status needed: approved service choice D1 version two, not simply wait for operations. A writer can sometimes outline an answer before that decision, but cannot finalize its commitment. Describe that permitted partial work without marking the missing input complete. This gives the scheduler useful information while preserving the difference between starting a draft and releasing an answer.

Check circular dependencies during work definition. If legal needs a price position and pricing needs legal’s settled term, identify the provisional options and the decision that resolves them. A cycle of final approvals is not a workable handoff. It may need an intermediate comparison package with explicit assumptions. Do not break the cycle by renaming an unapproved version final.

W3C PROV-DM separates entities, activities and responsible agents, and describes provenance relationships for what occurred. This is useful when distinguishing a file from work performed on it and the responsible party. Planned work should remain a plan; a dependency entry is not evidence that an output was generated, used or approved.

Ownership at this stage makes a package assignable; it does not settle every authorization. Record the accountable producer and the review or decision roles needed. The evidence-ownership guide examines those authorities in detail. Before starting controlled work, confirm the assignment and access required. An agent may propose a role based on an authorized record, but should not infer a person’s decision power from a department name.

Check for missing work in both directions

Read forward from each applicable ask and each submission instruction to its output-producing packages. A link to a writer is not enough if an evidence or decision prerequisite has no producer. Then read backward from each package to a buyer requirement or an authorized internal release purpose. Unexplained packages may be optional work, duplication or a missed source reference. Resolve their status instead of quietly increasing the scope.

Check the parent boundaries as well. Children should collectively cover the parent’s defined work without counting the same production twice. Distinguish a work item from its output count: one package can produce six identified samples, while one document can require several independently controlled decisions. If a later estimate sums parent totals and child effort together, the hierarchy will overstate the work even though each row looks reasonable.

When a source changes, preserve the old package definition and record the affected source-to-package links. For Elmshore, a changed fallback requirement can reopen D1 and the dependent answers, pricing entry and release checks. E1’s historical report does not become a different historical event; its adequacy for the changed use must be reassessed. Reuse is a conclusion to prove again where affected, not a reason to delete the change.

Give every material change an impact result: unchanged with reason, revised, added, retired with basis or awaiting decision. Do not count a superseded output as current completion. Conversely, do not order every team to restart work whose inputs and acceptance tests remain valid. The controlled relationship map should identify the affected work and the evidence supporting that boundary.

Changes to distinguish in the work definition
Observed changePackage questionRecorded treatment
Buyer adds another required destinationCan an existing output satisfy it, and what adaptation or review is needed?New consumption check or additional production work
Approved decision changesWhich answers, price entries and checks use that decision?Reopen the identified consumers against the new version
Evidence permission narrowsWhich uses are no longer within the released scope?Hold those uses and define replacement or permission work
Conditional delivery model changesWhich branch becomes applicable or ceases to apply?Preserve the old basis and activate the correct package set

A defined package is ready to plan, not automatically ready to submit

The handoff contains the hierarchy, package dictionary, requirement links, typed dependencies, owner acknowledgements and unresolved conditions. Dates and effort may be added by the relevant planning process, but they should not be invented merely to fill cells. This work definition tells the estimator what to estimate and the scheduler what must precede what. It does not establish capacity, duration or a feasible submission date on its own.

Report definition readiness separately from execution readiness. A well-specified evidence package can still be unstarted; an available draft can still be unreviewed. Use explicit states for defined, input missing, in production, awaiting review, accepted for stated use and superseded. If a package supports several uses, retain any use-specific holds instead of hiding them in one green status.

An agent can read authorized inputs, suggest candidate packages, identify uncovered asks, compare versions and check references or counts. It must preserve uncertainty about scope and applicability and distinguish planned from observed work. Access to private systems, task assignments that commit other people, external evidence requests, buyer contact, spending, approval and submission require their own authority. The resulting work breakdown makes those boundaries inspectable before execution begins.

Useful outcomes from RFP question work breakdown

  • Every applicable ask has a route to the work needed to satisfy it.
  • Shared evidence and decisions are produced once while their separate uses remain visible.
  • Package boundaries prevent both missing work and duplicate assignments.
  • Completion means an accepted output, not a message saying someone has finished.
  • Conditional and externally dependent work remains visible without being treated as already performed.
  • Estimators and schedulers receive stable identifiers, inputs and handoff conditions.

How to run the work

  1. 01

    Fix the production scope

    Identify the bid, entity, lots, response stage and current question and instruction versions. Include required attachments and release work outside numbered questions.

  2. 02

    Name the required outputs

    For each interpreted ask, identify the answer, evidence, decision or physical product needed. Distinguish producing the bid from performing the future contract.

  3. 03

    Define bounded packages

    Group work around outputs that can be owned and accepted. Describe scope, exclusions, inputs and completion proof in a package dictionary.

  4. 04

    Map shared uses and dependencies

    Give shared work one home and link all consumers. State the specific input that each dependent package needs and whether a conditional branch applies.

  5. 05

    Reconcile the whole scope

    Check from every ask to its work and from every package back to a buyer requirement or justified release purpose. Resolve gaps, duplicates and circular handoffs.

  6. 06

    Release the work definition

    Have responsible owners confirm the package boundaries and acceptance tests. Pass the controlled set to estimation and scheduling, retaining change history and unresolved decisions.

Questions that change the decision

  • What concrete output makes this part of the response possible?
  • Is this a separate package or an activity within an existing package?
  • Which consumers can use the same approved evidence version?
  • Does a dependency require raw material, reviewed material or an authorized decision?
  • What establishes that a conditional package is applicable?
  • Can another person verify completion without asking the producer what done means?

Where teams lose control

01

One question is treated as one writing task despite hidden evidence and decision work.

02

A shared input is duplicated under every answer and counted several times.

03

A common source is assumed to remove the need for destination-specific review.

04

A post-award delivery activity is confused with preparing the bid’s delivery plan.

05

Unresolved applicability is recorded as not applicable.

06

A completed draft is reported as an accepted and releasable answer.

Measure the finished job

Measure the completed workflow, including review effort and exceptions. Output volume on its own is not evidence of a better process.

  • Applicable asks without a defined work-output route
  • Packages without scope boundaries or acceptance tests
  • Duplicated production assignments for the same shared output
  • Dependent packages waiting for an identified input or decision
  • Conditional branches with an unresolved applicability basis
  • Changed sources with unreviewed downstream package impacts

Common questions

Should every RFP question become one task?

Not necessarily. One question can need evidence, a decision and several response products; several questions can consume one shared input. Keep the question-to-work map separate from the work hierarchy, and define packages at a level where an output can be owned and accepted.

Is the work breakdown the same as the question-subpart tree?

No. The subpart tree preserves what the buyer asks and how the answer must cover it. The work breakdown specifies what the bidder must produce to satisfy those asks and prepare the submission. It takes the controlled subpart analysis as an input and links back to it.

How do we avoid counting shared evidence twice?

Give its production one package and link the released output to every authorized consumer. Keep any adaptation, factual-scope test and use-specific approval in the consuming work. Sharing a report does not mean every answer can use it unchanged or under the same permission.

When is a work package small enough?

When its scope, responsible producer, required inputs and completion evidence are clear enough to manage. Split at meaningful handoffs, different authorities or independently usable outputs. Do not split every editing action merely to increase the task count, or combine unrelated decisions merely to keep the board short.

Can a complete breakdown prove we will meet the deadline?

No. It provides the work definition for estimation, capacity checks and scheduling. A feasible date also needs effort and duration evidence, availability, dependency timing and appropriate contingency. Keep unknown timing visible rather than claiming deadline readiness from a complete list.

Primary references

Tony Kim

Tony Kim

Founder and CEO

Tony writes about applied AI, dependable product engineering and the systems that turn complex response work into controlled delivery.

Proposal software for source-grounded RFP, RFI, DDQ and questionnaire response work.

Bid, proposal, presales, security and compliance teams. Start with the workflow, constraints and evidence you already have.