A bid investment decision tests whether an organization should spend more on one live tender at a named approval gate. It imports current eligibility, delivery, contract, response-cost and award-probability records rather than recreating them. The decision compares the incremental cost still avoidable with the contribution expected if the contract is awarded, using ranges and explicit award states. It shows cash and opportunity cost separately, tests downside and working-capital limits, and admits strategic value only when it is a named outcome with an owner, route to use, evidence, expiry and approved value cap. Its work product is a bid_investment_record containing the tender and lot, decision date, evidence versions, sunk and remaining cost, commercial scenarios, strategic-value record, switching tests, authority, spend cap, decision state and recheck triggers. It does not set the tender price, predict the award, approve contract risk, allocate delivery capacity or authorize a bid submission.
Teams often approve a bid because the contract has a large headline value, because work has already started or because the opportunity is called strategic. None of those statements tells the approver whether the next pound should be spent. Revenue can conceal a thin or negative contribution. A single win probability can conceal a range that crosses the decision boundary. Hours already consumed can create pressure to finish even though they cannot be recovered. “Reference customer” and “market learning” can become unlimited credits with no owner or usable asset. The opposite error also occurs: a high response budget can look wasteful even when the contribution, competitive position and downside remain favorable across credible scenarios. Without a dated marginal decision, the meeting rewards confidence and persistence rather than evidence.
Make the decision at the next reversible spending boundary. Retain all pursuit cost for learning, but exclude irrecoverable spend from the arithmetic that decides what to do next. Include the opportunity cost of employees or assets already paid for when they still have a credible alternative use. Use contribution after contract delivery cost rather than revenue, and keep award probability as an approved range tied to a forecast event. Model no award, cancellation, partial award, delay and adverse contract outcomes where they are material. Run switching tests so the approver can see which assumption must change to justify continuation. Strategic value is a separate, capped case and cannot repair failed eligibility, unapproved contract exposure or unavailable delivery capacity. An agent may join authorized records, check units, calculate scenario ranges and show a counterfactual. It may not invent margin or probability, assign monetary value to reputation, spend budget, displace another pursuit, accept contract risk or approve the decision.
Decision boundary
Decide about the next avoidable spend, not the work already done
A useful record begins at a decision point: before solution workshops, before an external legal review, before a mandatory sample, before a presentation, or whenever the current budget is nearly consumed. Name the next block of work and the latest time at which it can be stopped. “Finish the bid” is too broad. The team needs to know whether it is deciding on £8,000 of discovery, £31,000 through submission or £12,000 for a conditional presentation stage.
Attach the exact procurement object. A framework maximum, one call-off and one lot can have different economics. Record the buyer, procedure identifier, lot, controlling document version, current stage, deadline and forecast event. Import the current eligibility, delivery-fit, contract-risk, pricing, response-cost and award-probability records. If one is missing, show the missing input. Do not silently build a replacement during the meeting.
Hard gates run before return. An ineligible bidder, an unapproved unlimited liability, a delivery plan with a blocking gap or a prohibited data use does not become acceptable because expected value is positive. The investment record may return indeterminate while an authorized owner resolves the gate. It must not price an exception that the organization has not authorized.
| Field | Required content | Stop condition |
|---|---|---|
| decision_scope | Buyer, procedure, version, lot and stage | Mixed or unidentified opportunity |
| decision_gate | Decision time and avoidable work through the next gate | No reversible spending boundary |
| mandatory_gates | Eligibility, delivery, contract, security and authority states | Failed or prohibited state |
| cost_record | Approved estimate version and actual-to-date cut | Missing remaining-cost basis |
| commercial_record | Award state, contribution range and pricing approval | Revenue-only case |
| forecast_record | Probability range, event, evidence date and owner | Unsupported point estimate |
Cost basis
Keep two cost views because they answer different questions
The full-pursuit view starts when opportunity-specific work began and retains internal effort, external cash, evidence work, partner effort and later stages. It answers how much this pursuit consumed and supplies the baseline for estimating later bids. The next-choice view starts now. It includes only cash and capacity affected by the current choice. An invoice already paid and an irreversible workshop already delivered belong in the full record, but they cannot make continuation more attractive.
Committed does not always mean sunk. A purchase order may be cancellable with a fee. A specialist may be booked but movable. Record the cancellation date, recoverable amount and authority. The avoidable cost is the difference between stopping and proceeding, not the original order total. Conversely, an employee already on payroll still carries opportunity cost if the same hours could protect a client delivery, support a stronger pursuit or complete funded product work. Name that alternative and its time window. Hypothetical future revenue is not an opportunity-cost fact.
The U.S. Federal Acquisition Regulation defines bid and proposal costs for its own contracting context as the cost of preparing, submitting and supporting bids or proposals, while excluding effort required in contract performance. That boundary is useful, but its accounting treatment is not a universal rule. Apply the organization’s own accounting policy. For this decision record, consistency and visibility matter more than forcing cash, capacity and opportunity cost into one total.
| Cost state | Use in next-choice arithmetic | Retention |
|---|---|---|
| sunk_actual | Exclude | Keep in total pursuit history |
| irrevocable_commitment | Exclude except unpaid cash exposure | Keep amount, timing and obligation |
| cancellable_commitment | Include cancellation delta | Keep supplier terms and decision date |
| remaining_internal_work | Include approved loaded cost or capacity unit | Keep role, range and schedule |
| remaining_external_cash | Include full avoidable amount | Keep quote, tax, currency and due date |
| credible_alternative_use | Show separately as opportunity cost | Keep owner, competing use and evidence |
Commercial cases
Compare response investment with contribution, not contract value
Contribution is the amount the awarded work is expected to leave after the costs assigned to delivering it on the approved commercial basis. The finance owner must define which overhead, financing, tax and risk provisions are included. Use the award state that the buyer can plausibly commit. For a framework with no guaranteed spend, the published ceiling is not contribution. For a multi-lot competition, do not multiply by lots that the delivery plan cannot accept. Options and renewals need their own probability, margin and timing rather than automatic inclusion.
Expected contribution before response cost can be expressed as award probability multiplied by contribution if awarded. That compact equation is a check, not a verdict. Show the values at credible matched scenario corners and disclose dependence between inputs. A low price may raise the chance of award while reducing contribution; treating the highest probability and highest margin as the same case can create an impossible upside. If probability already reflects a contract issue, do not deduct the same issue again unless the second line measures a distinct consequence.
Expected value also misses some constraints. A £60,000 bid may have positive expected contribution but still breach this quarter’s cash limit. An award may require a bond, mobilization spend or ninety-day payment cycle that the organization cannot fund. A low-probability liability may exceed risk appetite even when its average looks small. Show cash timing, peak working capital, concentration and accepted downside beside the commercial range. Averages do not grant authority to cross those limits.
| State | Required value | Common mistake |
|---|---|---|
| no_award | Remaining response cost and any close cost | Assuming loss has no cost |
| full_award | Approved contribution, timing and delivery risk | Using revenue as benefit |
| partial_award | Contribution and stranded bid work for awarded scope | Pro-rating full-lot margin mechanically |
| cancellation_or_delay | Cost recoverability, resource release and expiry | Keeping the old probability unchanged |
| option_or_renewal | Separate exercise probability and contribution | Counting the maximum term as committed |
| adverse_delivery_case | Distinct accepted downside and owner | Double counting a risk already in margin |
Strategic value
A strategic benefit needs an owner, a use path and a limit
“Strategic” should introduce a record, not end the discussion. Name what will exist because the organization competes or wins. It might be an approved reference in a defined sector, a reusable test artifact, a partner agreement that opens named pursuits, or evidence that satisfies a future qualification. Then state whether the value depends on bidding, reaching a stage or winning. If the buyer prohibits publicity, a reference-logo claim has no current basis. If the artifact cannot be reused without customer permission, that condition belongs in the record.
Assign the outcome to a person who can realize it. Record the target pursuits or decision it will serve, the evidence for demand, the latest useful date and the action required after this tender. Cap the amount the approver is willing to invest for that option. The cap need not pretend to be a market valuation. It is an authority boundary: “we will spend up to £6,000 for this reusable compliance package if legal confirms reuse rights by Friday” is reviewable. “The logo is priceless” is not.
Strategic value cannot override a hard gate, hide a loss-making contract or consume capacity without the relevant owner’s consent. It may justify a controlled experiment or a limited next stage when the commercial case alone is inconclusive. In that case, issue proceed_with_cap and state the artifact, owner, maximum incremental spend and evidence needed at the next gate. If nobody owns the outcome, assign zero decision credit while retaining the claim as unverified.
| Field | Question it answers | Reject when |
|---|---|---|
| outcome_or_asset | What specific thing will exist? | Only reputation, learning or access is named |
| realization_event | Does value require bid, shortlist or award? | The event is unstated |
| use_path | Which future decision or pursuit will use it? | No named use exists |
| owner | Who can secure and apply the value? | Nobody accepts ownership |
| evidence_and_rights | What supports demand and permitted reuse? | Use depends on unverified permission |
| expiry_and_cap | When does it expire and how much may it justify? | Value is timeless or unlimited |
Sensitivity
Find the assumption that changes the answer
A central expected value is useful only when the route to it remains visible. Start with lower and upper bounds supported by the imported records. Create coherent cases: a lower contribution may belong with slower volume and a higher delivery cost; a higher award probability may belong with a price that also lowers contribution. Do not cross-pair every minimum with every maximum and call the resulting span a forecast. Explain which combinations are plausible and which are not.
The basic break-even award probability is remaining avoidable response cost divided by contribution if awarded. It works only when both amounts share the same scope, currency, price basis and time cut, and when material downside has not been omitted. Calculate it across the contribution and cost range. Then compare the threshold with the approved probability range. A large gap in the favorable direction supports a stable decision. Overlap means the result depends on evidence that still needs resolution.
Run other switching tests on the variables that the team can investigate or control: the cost cap, minimum contribution, probability after a clarification, allowed contract position, minimum guaranteed volume or value assigned to a reusable asset. The test should produce a question or limit. If the case works only when every uncertain input reaches its optimistic bound, say so. HM Treasury’s Green Book is public-sector appraisal guidance rather than a supplier ROI rule, but its disciplines on sunk cost, sensitivity and switching values are useful here when applied within the organization’s own commercial policy.
| Test | Expression | Decision use |
|---|---|---|
| break_even_probability | remaining cost divided by award contribution | Compare with the approved probability range |
| maximum_response_cost | authorized value less required risk buffer | Set the next-spend cap |
| minimum_award_contribution | remaining cost divided by accepted probability | Return to pricing if the floor is not met |
| strategic_value_needed | commercial shortfall plus approved buffer | Test whether the claimed asset would need implausible value |
| working_capital_limit | peak cash exposure against approved headroom | Keep liquidity outside expected-value averaging |
Worked case
A fleet-routing tender only works in its favorable corner
A software supplier is considering one lot for a city fleet-routing platform. Eligibility, delivery fit and the current contract position are open but not failed. The response-cost record shows £11,000 already spent. That amount stays in the pursuit history and leaves the next-choice arithmetic. Work through submission will require another £34,000 to £49,000, driven by an integration prototype, data-protection review and route-data validation. Finance estimates contribution from the initial committed term at £145,000 to £235,000. The approved award-probability record is 12% to 22%.
Expected contribution before the remaining response cost therefore spans £17,400 to £51,700. After response cost, the crude corners run from negative £31,600 to positive £17,700. The break-even probability spans about 14.5% in the favorable cost-and-contribution case to 33.8% in the adverse case. The approved probability range crosses only part of that threshold range. A central case of £41,000 cost, £185,000 contribution and 17% probability yields negative £9,550 before any separate downside. The case is sensitive, not comfortably positive.
The capture lead claims that a public-sector reference would make the bid worthwhile, but no owner has confirmed publicity rights, named a later pursuit or set a value cap. It receives no decision credit. The unresolved route-data licence could change both prototype cost and contribution. The record returns clarify_then_reassess, with authority for no more than £3,500 to resolve data rights and reprice the prototype by 10 September. It prohibits starting the full build. If the clarification does not lift the contribution floor, lower the remaining cost or support a probability above the matched break-even threshold, the next state is decline.
| Input | Current evidence | Decision treatment |
|---|---|---|
| sunk_actual | £11,000 | Retain for learning; exclude from next choice |
| remaining_response_cost | £34,000 to £49,000 | Use as avoidable investment range |
| award_contribution | £145,000 to £235,000 | Initial committed term only |
| award_probability | 12% to 22% | Approved range; not raised for preference |
| strategic_value | Unowned public-reference claim | Zero decision credit |
| decision_state | clarify_then_reassess | £3,500 investigation cap; expires 10 September |
Decision record
Return a bounded state that another reviewer can reconstruct
The bid_investment_record uses six states. proceed means the authorized evidence supports funding through the named next gate. proceed_with_cap adds a lower spend or commitment boundary. clarify_then_reassess funds only specified evidence work before a new decision. decline stops avoidable pursuit work. indeterminate means a material input or gate cannot yet support a decision. authorization_missing means the analysis may exist, but no permitted owner can make or fund the choice. Each state carries reasons and contrary evidence, not just a label.
An agent can retrieve the source records, normalize dates and currencies using an approved basis, validate that scopes match, calculate the scenario table and identify a switching point. It should retain the original values and transformations. It should abstain when contribution, probability, cost ownership or a mandatory gate lacks evidence. Access to a document does not imply permission to expose its figures in another system. The record may contain commercially sensitive inputs and must stay inside the authorized boundary.
The human decision owner accepts the spend, displacement and risk. Record the approver, decision time, maximum spend through the next gate, commitments that remain prohibited, and triggers for review. Typical triggers include a tender amendment, clarification response, material cost variance, pricing change, contract exception, capacity conflict, shifted award date, competitor evidence or expiry of the probability record. Submission remains a separate authorized action even after a proceed decision.
| State | Meaning | Permitted next action |
|---|---|---|
| proceed | Evidence supports the named gate within normal authority | Fund work only through that gate |
| proceed_with_cap | Case supports work under a tighter boundary | Fund to the stated cap and conditions |
| clarify_then_reassess | One bounded investigation may change the result | Resolve named facts, then issue a new record |
| decline | Avoidable continuation is not authorized | Stop and close the pursuit record |
| indeterminate | Material evidence or gate is unresolved | Return specific missing inputs |
| authorization_missing | No authorized owner can decide or fund | Escalate without committing spend |
What good looks like
Useful outcomes from tender bid investment decision
- One tender, lot, procedure version and decision gate define the investment question.
- Sunk, committed, cancellable and still-uncommitted response costs remain visibly different.
- The next decision uses incremental avoidable cost while the full pursuit cost stays in the record.
- Contribution after delivery cost replaces headline contract revenue in the return comparison.
- Award probability remains a sourced range with a forecast event, evidence date and owner.
- No-award, cancellation, partial-award and contract-downside states are included when material.
- Cash timing and working-capital exposure are tested outside the expected-value calculation.
- Every strategic benefit names an asset or outcome, owner, use path, expiry and approved cap.
- Sensitivity and switching tests expose which assumptions control the result.
- The decision carries a next-spend cap, authority, expiry and explicit recheck triggers.
Operating model
How to run the work
- 01
Fix the decision gate
Record the tender, lot, current document version, submission or shortlist stage, decision time and the next block of spend that can still be avoided. Import the latest gate records and stop if a mandatory gate has failed.
- 02
Reconcile the response investment
Separate actual sunk cost, irrevocably committed cost, cancellable commitments, remaining internal work, remaining external cash and the credible alternative use of scarce capacity. Preserve both the total-pursuit and next-choice views.
- 03
Import commercial scenarios
Use an approved contribution range for the exact award state, an approved award-probability range and any material cancellation, delay, partial-award, penalty, cash or working-capital exposure. Do not substitute revenue for contribution.
- 04
Bound strategic value
Describe each claimed strategic outcome as a specific reusable asset or unlock condition. Assign an owner, evidence, use path, date, expiry and a maximum value authorized for this decision. Reject unowned labels such as visibility or learning.
- 05
Calculate ranges and switching points
Show matched scenario corners, a documented central case where one exists, break-even award probability and the cost, contribution or probability change that would reverse the decision. Keep correlated assumptions together.
- 06
Apply authority and limits
Compare the evidence with the organization’s risk, cash, capacity and strategic-value rules. Set a decision state, next-spend cap, prohibited commitments, owner and expiry. Escalate any exception to the role that owns it.
- 07
Reopen on a material change
Reissue the record when an amendment, clarification, competitor fact, pricing decision, cost variance, capacity conflict, contract position or award timetable crosses a named trigger. Never let an early proceed state become permanent permission.
Evaluation
Questions that change the decision
- Which exact tender, lot, version and approval gate does this record govern?
- What spend and capacity can still be avoided if work stops now?
- Which commitments are already irrevocable, and which can be cancelled at what cost?
- What credible alternative would use the same scarce people or cash during the same period?
- What contribution remains after the full cost and risk of contract delivery?
- Which award state does that contribution describe: full lot, minimum volume, partial lot or option?
- What evidence supports the lower and upper award-probability bounds?
- Which cancellation, delay, no-award, penalty or working-capital states matter separately?
- Is a claimed strategic benefit a controlled asset with an owner and a plausible route to use?
- What award probability, contribution or remaining cost would switch the decision?
- What amount may the team spend before another approval is required?
- Which event expires the result even if the spend cap has not been reached?
Failure modes
Where teams lose control
Headline revenue may be compared with response cost while delivery cost and contract risk disappear.
Spend already incurred may be counted as a benefit of continuing.
An employee’s salary may be treated as sunk even when the person has valuable alternative work.
Cost, contribution and award probability ranges may be paired only at their favorable corners.
A central expected value may look precise even though every input is conditional.
Bid probability may be raised to make the desired answer pass.
The same contract risk may reduce margin, probability and a separate downside line three times.
Maximum framework or option value may be mistaken for committed award contribution.
Working-capital pressure may be hidden inside a positive expected contribution.
Vague strategic labels may become an unlimited override for poor economics.
A proceed decision may be treated as permission to exceed budget or submit the bid.
Private pricing, personnel or competitor information may be exposed to an agent without authority.
Measurement
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.
- decision records tied to one tender version, lot and dated gate
- remaining cost supported by an approved response-cost record
- commercial cases using contribution rather than contract revenue
- award ranges linked to an approved forecast event and evidence date
- decisions with matched downside, base and upside assumptions
- records with break-even probability and at least one switching test
- strategic-value claims with an owner, use path, expiry and cap
- proceed decisions carrying a current next-spend limit
- reassessments triggered before an obsolete assumption drives more spend
- actual response investment and decision outcome retained for later calibration
Questions
Common questions
Should bid costs already spent be included in the decision?
Keep them in the total pursuit record for accountability and calibration, but do not let irrecoverable spend justify the next choice. The decision uses cost and capacity that can still change.
Is employee time a sunk cost because salaries are already paid?
Not automatically. If those hours have a credible alternative use in the same period, show that opportunity cost separately. Do not claim hypothetical revenue as certain loss.
Why use contribution instead of contract revenue?
Revenue does not show the delivery cost, contract provisions or financing needed to earn it. The finance-approved contribution range gives the response investment a comparable commercial benefit.
Can expected value make the decision automatically?
No. It is one view. Cash limits, working capital, concentrated downside, failed gates, capacity and strategic authority can control the decision even when the average is positive.
How is break-even award probability calculated?
Divide remaining avoidable response cost by contribution if awarded, using the same scope, currency, price basis and decision date. Calculate a range and add omitted downside before relying on it.
May strategic value justify a commercially weak bid?
Only within explicit authority. Name the asset or outcome, realization event, owner, use path, evidence, rights, expiry and value cap. It cannot override a hard gate or an unapproved loss-making contract.
How should framework maximum value be treated?
Do not treat a ceiling as awarded revenue or contribution. Model guaranteed or credible call-off states separately, with their own volume, margin, timing and probability evidence.
When should the decision be reopened?
Reopen it when a named trigger changes a mandatory gate, remaining cost, contribution, probability, cash exposure, capacity, timetable or approved strategic case, or when the record expires.
Can an AI agent approve or fund the bid?
No. It may calculate from authorized records and identify missing evidence or switching points. A permitted human owner accepts the commercial choice, budget, displacement and risk.
Sources
Primary references
- The Green Book 2026 HM Treasury
- Cost Estimating and Assessment Guide U.S. Government Accountability Office
- FAR 31.205-18: bid and proposal costs U.S. General Services Administration
- The Sourcing Playbook UK Cabinet Office
- Risk Allocation and Pricing Approaches guidance UK Cabinet Office
- TPE/PME: se développer grâce aux marchés publics Direction générale des Entreprises
About the author
Tony Kim
Tony writes about applied AI, dependable product engineering and the systems that turn complex response work into controlled delivery.
Zelius
Managed tender intelligence and bid execution for teams that want the commercial outcome.
Suppliers, founders and commercial teams pursuing public or private opportunities. Start with the workflow, constraints and evidence you already have.
