Satellite supply

With financial analysis off, maximize complete satellites ready in their scheduled month, with equal weight. With financial analysis on, retain the exact satellite/month launch plan chosen without costs, then minimize machine-time, one-time activation and inventory holding costs. Launch profit is reported separately and cannot cancel or swap launches. If a buffer is enabled, retain the chosen launches and then fund the buffer as described below. Delayed satellites count once; catch-up launches are not modeled.

• Whole components and satellites. A complete bill of materials is required in the scheduled month.
• Only qualified suppliers with the required machine type and minimum size. Type C makes A, B and C. A and B make only their own types.
• Each machine shares one monthly budget: supplier days × uptime. Using a machine reserves it for the full month.
• “Already on” is guaranteed capacity. Dated machines contribute from their activation month when planned supply is included.
• The optional safety buffer is available only with financial analysis on. It uses next month’s full scheduled SKU demand, rounded up to whole components. It keeps the selected monthly launches fixed, maximizes buffer coverage with spare capacity, then minimizes cost. Any buffer shortfall is shown. Financial mode first preserves the costs-off launch plan and minimizes its cost, then funds the buffer and includes all additional charges. The final month retains the last target. This is an additional scenario policy, separate from the case inputs.
• Inventory starts at zero. Carryover has an editable shared maximum in total component units across all SKUs at each month end, including buffer stock; the default limit is zero. A buffer target cannot override that limit, and any unmet target remains a shortfall. The Guaranteed button selects the base scenario without carryover or new qualifications. Planned reveals carryover and supplier-policy controls; financial analysis also reveals the buffer and restores their saved settings. Already qualified uses the existing SKU–supplier approvals, including dated machines from their availability month when Planned is selected. All capable also permits new qualifications after the chosen lead time. There is no inventory of unfinished satellites.
• The base Guaranteed policy uses current qualifications only. Checked next steps can add specific dated qualifications, inventory or machine reservations to either supply policy. Additional qualification is available with planned supply; when selected, it starts for all capable unqualified supplier–SKU pairs in parallel on January 1, 1973. The qualification lead time is editable from 0 to 24 calendar months; 0 is immediate and 24 completes on January 1, 1975. Completion-month hours are prorated by calendar days and shared across all newly qualified SKUs on each machine.
• Physical planning reports a launch gap. Financial planning reports a cost gap: (current cost − proven cost floor) / current cost. Launch proof and cost proof remain separate; Optimal requires both to complete. Costs cannot remove a physically selected launch. Search continues until proof or Stop search. Financial inputs are user assumptions in USD, separate from the case inputs. Prefilled prices are illustrative model-dollar examples, not real supplier quotes, SpaceX economics or historical 1973 prices. They remain editable. Reset examples restores these prices; Clear financial inputs restores blanks. Launch profit is before the capacity, activation and inventory costs entered here. Choose billing for the whole reserved month (the case rule) or production hours only (a future scenario). A planned machine incurs its activation cost once, in its first production month, and only if used. Already-on machines have no activation charge. Inventory holding is charged per SKU unit remaining at month end, each month until used. The default rate is $100 per component per month; each SKU rate remains editable. No discounting, qualification or other costs are assumed; the result is profit within this scope, not company-wide net income. Fractional production hours can produce fractional-cent charges. Every launch selected by the physical plan is a commitment in financial mode, even when its modeled profit is negative.
• Without a safety buffer, physical planning uses a brief allocation refinement after maximizing on-time launches. With a buffer, it keeps those launches, prepares a checked reserve and gives automatic buffer refinement a 3.5-second budget. Continue buffer allows another 12-second pass. Remaining reserve gaps stay visible; a budget expiry is not an optimality proof. Financial cost search for the protected launch plan continues until proof or Stop search. Launch results appear first. Alternative optimal monthly plans can exist.

Next steps and orders

Next steps tests one additional manager action at a time against the current scenario: qualifying one named SKU at one capable supplier, raising the shared inventory limit by a specific number of units (and enabling carryover if needed), securing an individual planned machine at an earlier date, or increasing a supplier’s operating hours or days throughout the horizon. Tests prioritize current component and shared-machine bottlenecks, timing and storage, with launches ranked first and verified cost savings considered when financial analysis is enabled. This determines test order; only verified improvements become recommendations. Each qualification recommendation names the SKU, supplier and responsible buyer. It approves that SKU only. Qualification starts January 1973 and uses the test lead time. Dates and capacity requests are scenarios to confirm with suppliers. Costs use the current prices; qualification and expediting fees and physical storage space are not modeled. A completion-month qualification shares the remaining hours across all new SKUs. Comparisons retain their gaps; a time-limited test finding no gain does not prove there is no benefit, and the benefits of separate actions cannot be added together. Buffered financial comparisons follow the existing launch-selection and funded-buffer policy rather than claiming a joint launch/buffer cost optimum.

Checking a next step includes it as a reversible scenario overlay. Unchecking removes that overlay; Reset all removes all of them. The original input tables remain editable and intact. Inventory changes respect the displayed Max stored units setting beside Buffer; the storage test increase is editable under Test settings. Older bulk-qualification recommendations are removed so they cannot silently approve every SKU. Next steps stays visible below the main Launch map and starts its tests when scrolled into view. Opening the Sensitivity tests box hides Next steps and pauses its tests; closing the box restores the main map and resumes unfinished recommendations when visible. The list recalculates after main-plan changes and shows verified candidates while tests run. Orders derives machine-month reservations and per-SKU production quantities from the selected verified plan. Each used machine is reserved for the whole month once, including when multiple SKUs share it. Orders are planning instructions only; nothing is sent to suppliers.

Sensitivity tests

Open the Sensitivity tests box on Launch map to expand the controls. They combine into one separate test map below, replacing the main map while the box is open. Closing the box returns to the main map. Next steps is hidden and paused while stress tests are open. The global Save main plan button saves the underlying main scenario, not the separate test. Toggle later dated machine arrivals, lower supplier uptime, slower new qualifications, longer processing times for every SKU, and higher scheduled launch demand. Select individual suppliers to take their entire capacity offline for the full simulation. Already-on machines and existing qualifications remain in place. Demand is increased for each satellite type and month, rounded up to whole satellites; months with zero planned launches stay zero. New satellites, SKUs, suppliers and months follow the same rules. All controls start off, and original inputs stay unchanged. Results use the current supply policy, accepted next steps, finite storage and financial inputs. Financial mode and buffer policy follow the main plan. Valid changes update automatically; Reset stresses clears just these tests. A short search budget keeps the page responsive; Refine gap gives it a longer pass. Only verified whole-unit plans are shown, and unfinished searches retain their gap. Test results are scenarios, not probabilities or guarantees.

Missing source inputs

SKU 7 assumes Type B, size 200 and 2.5 hours per unit, using SKU 8 as a proxy. SKU 16 assumes Type A, size 200 and 2 hours per unit, using SKU 18. Confirm these with supplier process engineers and cycle-time studies.

Reset all restores the workbook inputs with the approved SKU 7 and 16 values already filled, the original timeline, example prices and scenario defaults, with financial analysis off. The financial-analysis switch is on Financial assumptions. All other physical values come from the fictional case workbook.

Scope

Scenarios saves inputs and verified allocations in this browser, with the gap earned when saved. Original case is protected. Loading a scenario keeps a Previous draft, including pending input edits, and Reset all does not delete saved scenarios. Cost columns appear only for financial scenarios. Named snapshots do not change as a new search runs; update one explicitly to save a newer result.

Clicking a launch cell shows the component gaps in that allocation and links to supplier/machine constraints. Check capacity uses the same whole-part, timing, qualification, shared-hour and storage rules, without the profit objective or optional buffer targets. It first keeps every currently chosen launch, then prioritizes the selected cell if needed. Feasible witnesses and ceilings are reported separately. A gap or a red component does not prove an independent capacity bottleneck.

Monthly timing. Within-month launch dates, machine purchase prices, budgets and qualification costs are not modeled. Hourly capacity charges, activation and inventory holding costs are modeled only when financial analysis is on and user prices are complete. Qualification has no modeled concurrency limit or capacity cost. Component gaps compare the selected production plan with the full launch plan; they are not a ranking of independent bottlenecks.

Inputs and saved scenarios are stored in this browser. Chat requests send the current planner context and attached files to OpenAI. The solver runs locally using HiGHS. Identical previously solved scenarios are reused, including verified original-input results. Old profit-based allocations may supply verified route hints, but never financial optimality certificates for the new objective; Best found remains distinct from Optimal. Changing a modeled input requires a new search. Financial searches start with a checked feasible plan and continue until proof or Stop search. Continue to proof retains the best plan and its valid cost floor. Only the selected financial supply case is calculated; switch supply to calculate the other case. The assistant uses OpenAI to interpret requests and attachments and select planner tools. Enter the shared assistant password to connect; credentials stay on the server. Proposed changes require acceptance before they affect the model.