[Blog]()Production # Production scheduling: dates that respect available capacity A calendar with bars does not establish that its workload fits the available bottleneck. **[Jonathan Perez]()**Co-founder · Design, product and sales October 8, 2026 · 8 min read **Short answer** Schedule against usable time on the resource that limits the process, including setup, breaks, maintenance and committed work. Check materials and skills as well as machines. If work cannot fit, show what is split or delayed and who accepts the change; retrying plan publication must not create extra reservations or hide the earlier plan. ## Usable capacity and plan constraints Ten tests of minutes, setup, order splitting, materials and publication without duplicate reservations. [Download CSV worksheet]() In this guide - [How many minutes can you actually use?]() - [Example: two full orders cannot fit on Monday]() - [Does an available machine mean an order is ready?]() - [What happens when the resource breaks down?]() - [What should a capacity test demonstrate?]() - [What data should you bring to a diagnosis?]() ## How many minutes can you actually use? Start with resource calendars, shifts and unavailable periods. An eight-hour shift is not always eight hours of manufacturing. Identify setup, cleaning, breaks, maintenance and prior reservations without deducting a period twice when already excluded from the calendar. Distinguish machine duration from operator attendance: one shared person may constrain two seemingly available machines. Microsoft documents finite capacity that considers reservations and moves work when capacity is insufficient. It requires enabled plans and resources, a defined horizon and attention to model limitations. [Finite capacity](). Do not assume a configuration models every variable shift; demonstrate how it represents your calendars and which constraints it leaves out. ## Example: two full orders cannot fit on Monday **Synthetic plan CP01** A resource has 480 shift minutes, a 30-minute break and 60-minute maintenance: 390 usable minutes remain. There are no other reservations. Each A or B unit needs 2 minutes, with different setups. These are illustrative, not observed times. | Work |Setup |Processing |Total | | --- | --- | --- | --- | | A: 120 units |30 min |120 × 2 = 240 min |270 min | | B: 80 units |20 min |80 × 2 = 160 min |180 min | | Both complete |50 min |400 min |450 min | | Available Monday |480 − 30 − 60 |No other reservations |390 min | Completing both needs sixty more minutes. In this exercise A goes first, B may split and its setup carries into Tuesday without extra time. A uses 270, leaving 120. Setting up B uses twenty and making fifty uses one hundred: Monday completes A and fifty B units. The remaining thirty need sixty minutes Tuesday. If restarting B requires another setup, Tuesday needs that time as well; recalculate the load and planned finish against Tuesday’s calendar. Do not hide that assumption inside a Gantt bar. ## Does an available machine mean an order is ready? Check eligible material, the preceding operation, tools and people qualified for the task. A promised purchase order is not received material. For sequential operations, define whether the entire lot must finish before the next starts or partial transfer is allowed. Avoid multiplying machines by hours if they all wait for the same mold or operator. The effective recipe and routing must explain quantities and duration. SkyPlanner publishes APS scheduling using orders, capacity and materials alongside planner controls. [Production scheduling](). This is vendor-declared coverage; its case figures and speed are neither Nightly results nor evidence for your factory. An algorithm can propose a plan, but its input constraints and priorities still need review. ## What happens when the resource breaks down? Retain the published plan and create a new proposal using actual unavailability. Show affected orders, changed dates and reservations released or reassigned. If one person edits priority while another schedules, compare against the base revision and resolve the conflict before publication. Changing the planned date does not establish customer notification or acceptance of a new delivery; those decisions need their own evidence. ## What should a capacity test demonstrate? - Normal: CP01 shows 390 usable minutes and explains 270 for A plus 120 for Monday's part of B. - Exception: unavailable material or tools hold work under the defined rule; work cannot consume production capacity as if already started. - Change: a breakdown creates a revised proposal with visible affected orders and an owner for accepting new priority. - Retry: repeated P01 publication keeps one published revision and one reservation per operation. - Assumption: prohibiting a B split or requiring Tuesday setup changes the result; the system cannot preserve an impossible date to appear compliant. ## What data should you bring to a diagnosis? Complete the [minutes and constraints worksheet]() for one bottleneck. The [production guide]() covers progress; [equipment maintenance]() explains unavailability and closure. To scope [production control](), bring two orders competing for the same resource and their setup assumptions. You need not invent timings to begin: unknown durations can remain items for measurement. ## Bring two orders competing for one resource In a free diagnosis we can identify missing calendars, assumptions and constraints for scheduling. - The bottleneck calendar with breaks and maintenance. - Two orders with quantities, routing and known or estimated timings. - A recent priority change, unavailable material or breakdown. [Book a free consultation]()[Ask on WhatsApp]() Related - [Production control]() - [Manufacturing software]() ## Frequently asked questions ### Do total hours across machines establish capacity? Not when a shared resource, preceding operation, mold or operator limits flow. Test the bottleneck and its dependencies. ### Can I always split an order between days? It depends on the process, transfer rules and setup. The example explicitly permits it; your test should show allowed and prohibited cases. ### Does an AI scheduler guarantee the best date? No. Proposals depend on data, constraints, horizon and priorities. Require explanation and review of a feasible result before publication. ### What if actual times are unknown? Identify uncertainty, retain estimates as estimates and define measurement. Do not present an assumed time as observed or a calculated date as an accepted commitment. ## Sources 1. [Finite capacity planning and scheduling]()Microsoft Learn 2. [APS capacity and material scheduling]()SkyPlanner Last updated: October 8, 2026 ## Keep reading [ProductionOct 8, 2026 ### Preventive maintenance: calendars, readings and work-order closure]()[ProductionOct 8, 2026 ### Production rework: a linked order, consumption and traceable closure]()[OperationsOct 4, 2026 ### How to know where every order stands without walking the floor]() --- Canonical: https://nightlysoftware.com/en/blog/production-capacity-scheduling Updated: 2026-10-08 Description: Test usable minutes, setup, materials and rescheduling before promising dates that a machine or shift cannot meet in practice.