Bracken Offshore · Decision brief · 25 May 2027

Tessa, choose both mitigations without the survey.

Recommend package B: repair preparation + weather-capable vessel, and request a £822k P90 budget. For the twelve structures, the supplied four-state model completes every state by day 11, with no backup use. The cheaper feasible survey policy saves only £10.58k in expectation and leaves a late, unfunded tail. O1 O2–O7

Recommendation — not approvedFictional planning estimatesNo survey result received

Pay £10.58k for the stronger planning assurance

Unconditional B costs £606k expected, versus £595.42k for survey + → B / − → N. The £10.58k premium (1.8% of the survey policy's expected cost) raises model completion by day 14 from 97% to 100%, lowers the P90 request by £25k, and removes the modelled backup and funding tail. This is planning assurance within this fixture, not real-world safety assurance. O2–O8

Recommended funding request

£822k covers 100% of the supplied B outcomes. It is £216k above expected cost and £28k below the board's £850k P90 ceiling. Modelled overrun above £822k: £0, probability 0%. The £20k backup reservation is still paid. O1 O3 O7

If the board instead prefers the survey saving

Authorise the whole +B/−N rule before the result, request £847k, and explicitly resolve the 3% chance of £1,125k total cost: £278k above the request, completing on 21 June. That tail is not automatically funded; no client extension is agreed. O1 O5–O8

Quality challenge: if defect sensitivity is 65%, +B/−N completes by day 14 only 94.75% of the time, below the 95% commitment. Its £847k P90 still passes the budget cap, illustrating why the budget test alone is insufficient. B remains at 100%. No probability is assigned to this alternative quality. O8

Ephemeral planning sandbox

Test the complete decision rule

Controls change a hypothetical policy, never source records, approvals or vessel orders. All totals are £000 excluding VAT. Reset or reload restores the original 25 May comparison: survey +B/−N at 80% sensitivity. The recommendation above stays fixed and recoverable. O6–O8

A result is information, not actual condition. The same +/− action applies to every compatible state; neither weather nor condition is known at package commitment. O5–O6

What the survey can and cannot tell Tessa

Joint outcomes, vessel handovers and dates O2–O6

CN/CD = calm ordinary/defects; RN/RD = rough ordinary/defects. “Joint chance” is the ex-ante probability of state AND result. Day 1 is 1 June; finishing day d means the end of d June. Backup, if used, starts on 18 June. Support is charged once.

Cost components for every outcome O3–O5

Cumulative total-cost distribution O7

Equal-cost outcomes are combined first. P90 is the smallest total covering at least 90%; it is neither a 10% uplift nor an average of branch budgets.

Find the useful cost, deadline and P90 frontier

All four fixed packages and all sixteen survey rules are enumerated. A frontier policy has no alternative with lower-or-equal expected cost and P90, and higher-or-equal timely completion, with at least one strict improvement. The residual maximum-cost tail is shown separately; this definition does not treat P90 as full funding. O6–O7

“Meets both” applies ≥95% timely completion and P90 ≤£850k to the entire policy before the result. “+B / −N” means B after a positive and N after a negative; it never means choosing after observing actual weather or condition.

Challenge the survey quality

Only B can finish the rough-defects state within the primary window. A survey rule with B after + and any other package after − succeeds with probability 1 − 0.15(1 − sensitivity). It therefore needs sensitivity at least 66⅔% to meet 95%. At 65%, only unconditional B and the redundant survey +B/−B meet both requirements. Buying a survey and always choosing B adds £25k without changing the action. O2–O8

This is a separate stress case, not a weighted forecast. False-positive likelihood stays at 20% under both survey qualities.

Decision, funding and vessel-call sequence

25 May · decide whether information is worth buying

Tessa recommends no survey and seeks campaign-board authorisation for B and the £822k request. No weather, condition or survey result is known. The survey-order cutoff is today, so waiting until 27 May to decide whether to survey would lose that option. If the board chooses the survey alternative, authorise its expenditure and the complete contingent rule today, with explicit sponsor treatment of its £278k tail. O1 O6–O8

26–27 May · execute only the authorised rule

If bought, the survey completes 26 May; the inspection lead accepts the evidence. + and − update probabilities but do not reveal the true state. By 27 May the board authorises the policy/budget and package choice; the marine lead confirms the primary vessel and firm £20k backup reservation. Repair preparation is finished before mobilisation, with no extra campaign delay. No new pre-commitment signal is available. O1 O4–O6

1 June onward · respect the booked windows

Mobilise day 1. Standard vessel leaves after day 14; W leaves after day 12 even if nominal work remains. The marine lead calls the reserved backup when residual scope requires it: no work in the gap, restart day 18, pay £80k call-out plus £55k per residual day. There is no priced option to combine vessels or add discretionary mitigation mid-campaign. Inspection evidence and safety decisions stay with authorised people under the supplied procedures. O1 O3 O6

Sponsor decision · identify who funds a tail before committing

For recommended B, the £822k request covers the entire model; the maximum state is RD at £822k and day 11. For +B/−N, do not interpret a £847k request as authority to spend £1,125k: record whether the sponsor provides a separate £278k backstop or consciously accepts the unfunded exposure, with an escalation owner. No such decision or client extension is recorded. Under 65% sensitivity, that rule also needs a different package policy to meet the existing deadline requirement. This page records proposals only. O1 O7–O8

Workings and evidence boundaries

Recorded: fictional instruction, offers, windows, prices, authority and decision dates in O1–O8. Forecast: O2 probabilities and mitigated durations; 100% means every supplied state, not certainty about real offshore operations. Derived: exact enumeration below. Proposed: B and its funding request. Hypothetical: all control selections and displayed survey branches. No survey result, approval, call-out or vessel order has occurred. O1–O8

d = nominal days − 3 × [S chosen AND defects] − 4 × [W chosen AND rough] window = 12 if W chosen, otherwise 14; rate = £42k or £40k respectively primary days = min(d, window); backup days r = max(0, d − window) completion = d if r = 0; otherwise 17 + r cost (£000) = 100 mobilisation + 20 reservation + 50[S] + 70[W] + primary days × rate + support + 55r + 80[r > 0] + 25[survey] p(state, +) = p(state) × P(+ | condition); p(state, −) = p(state) × (1 − P(+ | condition)) expected cost = Σ probability × total cost P90 = smallest cost b with Σ probability(cost ≤ b) ≥ 0.90 reserve = P90 − expected cost; tail = probability(cost > P90) max overrun = max(0, largest outcome cost − P90)

State probabilities remain joint: calm-defects is 15% and rough-defects is 15%; weather and condition are not factored into independent marginals. The survey modifies only the likelihood weights. No additional weather multiplier is applied to backup residuals, specialist support remains unchanged, and no cost/penalty outside the fixture is added. O2–O5

Window check: RD with W alone takes 18 − 4 = 14 nominal days, but W leaves after day 12. Two residual backup days end on day 19; cost £190k fixed + £504k primary + £110k backup + £80k call-out + £120k support = £1,004k. RD with B takes 18 − 3 − 4 = 11 days, £240k fixed + £462k hire + £120k support = £822k. O2–O4

Read the supplied evidence

Original fictional records, unchanged. Source dates refer to May 2027; this is a synthetic decision exercise.

O1 — Campaign instruction, 24 May 2027; director.
Mobilisation begins 1 June, calendar day 1. Completion duration d means end of calendar day d. Commercial commitment: estimated probability of completing all scope by day 14 must be at least 95%. Board considers a requested P90 budget no greater than £850k; costs beyond that request remain an exposure requiring explicit sponsor treatment, not automatically funded. No client extension is agreed. Tessa recommends; campaign board authorises policy/budget, marine lead calls vessels, inspection lead accepts evidence. Actual inspection safety decisions remain with authorised people under supplied procedures; this exercise concerns planning.
O2 — Joint planning states, 24 May; marine and inspection leads.
These four states are exhaustive and mutually exclusive. Probabilities are planning estimates. “Nominal days” include required work and weather delay before applying the listed mitigations/window rule; support cost is charged once per campaign.

| State | Probability | Nominal days | Additional specialist support, £000 |
|---|---:|---:|---:|
| CN: calm, ordinary condition | .45 | 7 | 0 |
| CD: calm, defects | .15 | 11 | 80 |
| RN: rough, ordinary condition | .25 | 12 | 0 |
| RD: rough, defects | .15 | 18 | 120 |

Weather and condition must not be treated as independent. Neither is known at commitment. All fixed costs/unit rates below are £000 excluding VAT. No further cost or penalty is omitted.
O3 — Vessel offers and finite windows, contracts manager, 24 May.
Standard primary vessel is available days 1–14, costs £40/day used and then must leave. Weather-capable primary W is available only days 1–12, costs £42/day used and also must leave. Every policy pays £100 mobilisation and a £20 firm backup reservation. Backup is guaranteed only from day 18, at £55/day used plus £80 call-out/remobilisation if needed. It can finish any residual scope in this fixture. Reservation is not refundable even if unused.

For the supplied planning model, first calculate mitigated nominal days. If they do not exceed the chosen primary window, finish on that nominal day and pay only that many primary hire days. If they exceed it, pay the full primary window, then use exactly nominal-days-minus-window backup days. No work occurs in the gap; backup starts on day 18, so completion is day 17 plus residual days. The quoted residual days already allow for the state's remaining weather/work; do not add another weather multiplier. Both vessels cannot be combined simultaneously to shorten work.
O4 — Mitigation packages, inspection/marine leads, 25 May.
S is pre-campaign repair preparation, costing £50: reduce nominal days by three in each defects state, none in ordinary states. W is choosing the weather-capable primary, adding £70 fixed vessel premium: reduce nominal days by four in each rough state, none in calm states; its rate/window are O3's W values. Both effects add if combined. Additional specialist support and state probabilities remain unchanged. Thus available packages are N (neither), S, W, B (both). Preparation is completed before day 1, with no separate calendar delay. These exact effects are accepted planning assumptions, not a claim about real equipment.
O5 — Optional survey, 25 May; survey contractor and inspection statistician.
Cost £25 regardless of result; completed 26 May before package decisions. Results are positive (+) or negative (−). P(+|defects)=.80; P(+|ordinary)=.20. Conditional on actual condition, result probabilities do not vary with weather; use those likelihoods on each O2 row, retaining the joint weather/condition prior. A negative result is not a clean-condition certificate. The survey changes information only: it does not itself shorten work, remove support costs or eliminate the backup reservation. No other signal arrives before the commitment deadline.
O6 — Decision timeline and policy limits, 25 May; commercial manager.
By 25 May choose whether to buy the survey. By 27 May choose N/S/W/B; all quoted packages remain available at the same prices until then. A no-survey policy chooses one package without observing condition/weather. A survey policy states one package after + and one after −; the same result must lead to the same action in every state. It cannot choose separately after seeing actual weather or condition. All sixteen result-to-package policies are allowed. Choices are fixed before the campaign; the backup call follows O3 automatically if required. No discretionary mid-campaign mitigation beyond that backup is priced or available.
O7 — Funding convention, finance director, 25 May.
Expected cost is the probability-weighted total campaign cost, including survey when purchased. P90 is the smallest budget covering at least 90% of total policy outcomes; combine equal-cost outcomes before interpreting coverage. Report achieved coverage, reserve above expected cost, maximum overrun beyond the requested budget and probability of that overrun. Do not substitute 10% contingency or average conditional budgets. The deadline requirement and P90 limit both apply to the entire ex-ante policy, not only the branch after a favourable result. A policy can pass them while still having a residual unfunded tail.
O8 — Challenge for decision, sponsor, 25 May.
“Survey expenditure needs a demonstrable benefit. Compare savings with the downside of acting on a false negative. I may prefer an unconditional package if its assurance premium is modest, but show the trade-off.” Test sensitivity with P(+|defects)=.65 instead of .80, keeping P(+|ordinary)=.20 and all state probabilities/costs unchanged. Reassess which policies meet the deadline requirement; no probability is assigned to this alternative test quality. No survey result has actually been received for this decision brief.
Read the complete supplied case · BPB2-05
# BPB2-05 — Offshore inspection campaign — choose responses and contingency

Suite BPB2 v0.3-pilot; Foray 247, third essai.

### Scenario

Bracken Offshore must inspect twelve structures before a commercial June deadline. A short-window weather-capable vessel looks attractive, but it may strand unfinished work; a preparatory survey offers useful but imperfect information about defects. Campaign director Tessa Roan must choose a practical pre-campaign decision policy and funding request by 27 May 2027. She needs the cost of that policy across possible outcomes, its chance of timely completion and its residual tail, rather than an average-duration promise or an idealised perfect-information response.

### Exact task prompt

Produce one usable in-browser answer page for the named practitioner, using the supplied fictional evidence. Keep observed facts, forecasts, assumptions and proposed decisions distinguishable, and cite source IDs beside material claims. No external research or real-world certification is required.

Prepare Tessa's campaign decision and contingency brief. Compare the four no-survey packages with feasible survey policies that choose a package after a positive or negative result. Identify the useful cost/deadline/budget frontier and recommend an implementable policy, explaining any premium paid for assurance. Calculate expected cost, probability of completion by day 14, the discrete P90 budget and exposure beyond it, using the joint outcomes and actual vessel windows. Show what is known when each decision is made and how false positives/negatives affect action and funding. Test the stated survey-performance sensitivity. Do not treat a survey result as actual condition, ignore remobilisation, or claim real-world safety assurance.

### Supplied records

Decision brief is prepared on 25 May, before the survey-order cutoff.

**O1 — Campaign instruction, 24 May 2027; director.** Mobilisation begins 1 June, calendar day 1. Completion duration d means end of calendar day d. Commercial commitment: estimated probability of completing all scope by day 14 must be at least 95%. Board considers a requested P90 budget no greater than £850k; costs beyond that request remain an exposure requiring explicit sponsor treatment, not automatically funded. No client extension is agreed. Tessa recommends; campaign board authorises policy/budget, marine lead calls vessels, inspection lead accepts evidence. Actual inspection safety decisions remain with authorised people under supplied procedures; this exercise concerns planning.

**O2 — Joint planning states, 24 May; marine and inspection leads.** These four states are exhaustive and mutually exclusive. Probabilities are planning estimates. “Nominal days” include required work and weather delay before applying the listed mitigations/window rule; support cost is charged once per campaign.

| State | Probability | Nominal days | Additional specialist support, £000 |
|---|---:|---:|---:|
| CN: calm, ordinary condition | .45 | 7 | 0 |
| CD: calm, defects | .15 | 11 | 80 |
| RN: rough, ordinary condition | .25 | 12 | 0 |
| RD: rough, defects | .15 | 18 | 120 |

Weather and condition must not be treated as independent. Neither is known at commitment. All fixed costs/unit rates below are £000 excluding VAT. No further cost or penalty is omitted.

**O3 — Vessel offers and finite windows, contracts manager, 24 May.** Standard primary vessel is available days 1–14, costs £40/day used and then must leave. Weather-capable primary W is available only days 1–12, costs £42/day used and also must leave. Every policy pays £100 mobilisation and a £20 firm backup reservation. Backup is guaranteed only from day 18, at £55/day used plus £80 call-out/remobilisation if needed. It can finish any residual scope in this fixture. Reservation is not refundable even if unused.

For the supplied planning model, first calculate mitigated nominal days. If they do not exceed the chosen primary window, finish on that nominal day and pay only that many primary hire days. If they exceed it, pay the full primary window, then use exactly nominal-days-minus-window backup days. No work occurs in the gap; backup starts on day 18, so completion is day 17 plus residual days. The quoted residual days already allow for the state's remaining weather/work; do not add another weather multiplier. Both vessels cannot be combined simultaneously to shorten work.

**O4 — Mitigation packages, inspection/marine leads, 25 May.** S is pre-campaign repair preparation, costing £50: reduce nominal days by three in each defects state, none in ordinary states. W is choosing the weather-capable primary, adding £70 fixed vessel premium: reduce nominal days by four in each rough state, none in calm states; its rate/window are O3's W values. Both effects add if combined. Additional specialist support and state probabilities remain unchanged. Thus available packages are N (neither), S, W, B (both). Preparation is completed before day 1, with no separate calendar delay. These exact effects are accepted planning assumptions, not a claim about real equipment.

**O5 — Optional survey, 25 May; survey contractor and inspection statistician.** Cost £25 regardless of result; completed 26 May before package decisions. Results are positive (+) or negative (−). P(+|defects)=.80; P(+|ordinary)=.20. Conditional on actual condition, result probabilities do not vary with weather; use those likelihoods on each O2 row, retaining the joint weather/condition prior. A negative result is not a clean-condition certificate. The survey changes information only: it does not itself shorten work, remove support costs or eliminate the backup reservation. No other signal arrives before the commitment deadline.

**O6 — Decision timeline and policy limits, 25 May; commercial manager.** By 25 May choose whether to buy the survey. By 27 May choose N/S/W/B; all quoted packages remain available at the same prices until then. A no-survey policy chooses one package without observing condition/weather. A survey policy states one package after + and one after −; the same result must lead to the same action in every state. It cannot choose separately after seeing actual weather or condition. All sixteen result-to-package policies are allowed. Choices are fixed before the campaign; the backup call follows O3 automatically if required. No discretionary mid-campaign mitigation beyond that backup is priced or available.

**O7 — Funding convention, finance director, 25 May.** Expected cost is the probability-weighted total campaign cost, including survey when purchased. P90 is the smallest budget covering at least 90% of total policy outcomes; combine equal-cost outcomes before interpreting coverage. Report achieved coverage, reserve above expected cost, maximum overrun beyond the requested budget and probability of that overrun. Do not substitute 10% contingency or average conditional budgets. The deadline requirement and P90 limit both apply to the entire ex-ante policy, not only the branch after a favourable result. A policy can pass them while still having a residual unfunded tail.

**O8 — Challenge for decision, sponsor, 25 May.** “Survey expenditure needs a demonstrable benefit. Compare savings with the downside of acting on a false negative. I may prefer an unconditional package if its assurance premium is modest, but show the trade-off.” Test sensitivity with P(+|defects)=.65 instead of .80, keeping P(+|ordinary)=.20 and all state probabilities/costs unchanged. Reassess which policies meet the deadline requirement; no probability is assigned to this alternative test quality. No survey result has actually been received for this decision brief.

### Browser deliverable and completion conditions

A decision page exposing no-survey options and survey result-to-action policies, with their branch probabilities, vessel transfers, costs, completion dates and cumulative cost distribution. A user must be able to choose the +/− actions and survey sensitivity, see the full ex-ante recalculation and locate any deadline or budget breach. Keep a visible distinction between recommendation, observed evidence and hypothetical result. Show a practical call/approval sequence and residual funding decision. Include a clear return to the gallery; simulation is optional, exact enumeration is sufficient.

The page must work in the supplied browser/runtime; a download button or file preview alone is insufficient. Include a clear “Return to gallery” link using the shared shell route, or ../index.html if no route is supplied. Required controls must change the dependent results, not just their appearance. Original evidence and the initial dated position must remain recoverable; hypothetical changes do not create approvals or alter source records. Optional exports are secondary.