Production scheduling sequences work orders on specific resources (machines, lines, work centers) over short horizons (typically days to weeks). It's distinct from supply planning (longer horizon, less detailed) and from execution (real-time shop floor management). This guide covers the concepts that matter for understanding modern production scheduling.
The concepts apply differently across industries: process industries (chemicals, food, pharma) emphasize sequence-dependent setup and campaign planning; discrete manufacturing emphasizes finite capacity and bottleneck management; mixed-mode operations apply concepts from both.
Mature production scheduling applies these concepts together. Finite capacity scheduling produces executable plans within resource limits. Sequence-dependent setup optimization minimizes total changeover time. Campaign planning balances changeover cost against inventory cost. Bottleneck management focuses optimization where it matters most. APS platforms enable all of this with sophistication basic ERP scheduling lacks.
Horizon supports production scheduling concepts at mid-market scale ($100M-$3B revenue, manufacturing operations). Where Horizon doesn't optimize for: highly specialized scheduling needs (semiconductor fabs, refinery operations) where specialist scheduling platforms typically fit better, or process-industry-specific complexity (chemicals with extreme sequence-dependent setup, pharma with regulatory cleaning) where process-industry specialists (OMP, AspenTech) typically fit better.
Production scheduling decisions affect throughput, cost, and customer service significantly. Bad scheduling produces excess changeovers, underutilized bottlenecks, late orders, and inflated WIP. Good scheduling optimizes throughput, minimizes changeover loss, hits delivery commitments, and runs efficient resource utilization. The gap between basic scheduling and advanced scheduling represents substantial operational performance difference.
Traditional MRP-based scheduling assumes infinite capacity. The system calculates what should be produced when based on demand and lead times, without checking whether resources have capacity to produce it. Result: schedules that may exceed actual capacity, requiring manual reconciliation to become executable.
Infinite capacity scheduling persists because: it's mathematically simpler, it's adequate when capacity is rarely constrained, it's what ERP systems traditionally provide. Most companies running only ERP-based scheduling have infinite capacity scheduling.
Finite capacity scheduling respects actual capacity constraints. Calculates feasible schedules within capacity limits, identifying when demand exceeds capacity and surfacing the choice (defer demand, add capacity, change priorities). The output: executable schedules that don't require manual capacity reconciliation.
The benefit: more reliable plans, less manual rework, better visibility into capacity decisions. The cost: more sophisticated systems (APS rather than basic MRP), more complex configuration, more data requirements.
Setup time depends on what was running before. Example: switching paint production from white to black requires minimal cleaning; switching from black to white requires substantial cleaning to prevent contamination. The setup time varies based on the sequence — going A-to-B may take different time than going B-to-A or A-to-C.
Process industries with cleaning requirements: chemicals (residue contamination concerns), food and beverage (allergen separation, flavor cross-contamination), pharmaceuticals (regulatory cleaning requirements), paint and coatings (color contamination). Discrete manufacturing with significant tooling changes: machining (different parts requiring different tooling), assembly (line reconfiguration between products).
Sequence-dependent setup makes scheduling decisions interact. The optimal sequence minimizes total setup time across the schedule. This is a combinatorial optimization problem — for N products, there are N! possible sequences. Basic scheduling typically ignores sequence dependency; APS platforms optimize sequences explicitly.
Groups similar products into production campaigns to minimize changeovers. Instead of running A-B-A-B-A-B (alternating, requiring multiple changeovers), run A-A-A-B-B-B (campaigns, requiring single changeover). The trade-off: longer time between cycles of each product, requiring more inventory to bridge between campaigns.
Process industries with significant changeover cost: chemicals (cleaning, qualification), food and beverage (allergen segregation, flavor switching), pharma (regulatory cleaning). Industries where changeover-driven cost is substantial relative to carrying cost. Industries where machine throughput is constrained by changeover time more than by run time.
Industries where changeovers are quick relative to run time. Industries where inventory carrying cost is high relative to changeover cost. Industries with strict service requirements that don't tolerate the longer cycle times campaigns create. The economics depend on changeover cost, inventory cost, and service requirements.
The resource constraining overall throughput. In any production system, one resource is slowest — every other resource has more capacity than necessary because the bottleneck limits the system. Identifying and managing the bottleneck is central to throughput optimization.
Eli Goldratt's framework: identify the constraint, exploit it (use it fully), subordinate everything else to it, elevate it (add capacity if needed), repeat (when constraint changes). The implication for scheduling: schedule the bottleneck optimally; other resources schedule to support the bottleneck, not optimize their own efficiency.
TOC-derived scheduling approach. The drum: the bottleneck sets the production pace for the whole system. The buffer: inventory protecting the bottleneck from upstream variability (so the bottleneck never starves). The rope: production releases tied to bottleneck consumption, preventing WIP buildup. The result: bottleneck runs at maximum throughput, WIP stays low, system flow improves.
The bottleneck isn't always obvious. Indicators: resource with longest queue or highest WIP backlog, resource running at highest utilization consistently, resource whose downtime stops the whole line, resource where extra capacity would most immediately increase throughput. Bottlenecks can shift: addressing one bottleneck often reveals a new one elsewhere.
APS software performs finite capacity scheduling with constraints like sequence-dependent setups, resource capabilities, material availability, and order priorities. Optimizes scheduling decisions across resources jointly rather than scheduling each resource independently.
Modern APS platforms include: finite capacity scheduling respecting resource limits, sequence-dependent setup optimization, alternative resource consideration (when multiple resources can do a job), material constraint awareness (don't schedule production without materials available), priority handling (rush orders, customer-specific commitments), what-if scenario evaluation. The capabilities address common scheduling pain points basic ERP scheduling doesn't handle.
Basic ERP-based scheduling: calculates what should be produced based on demand and lead times, assumes infinite capacity, treats setups as fixed parameters. APS: respects actual capacity, optimizes sequences considering setup dependencies, handles alternative resources, supports what-if analysis. The capability difference is substantial for operations with significant complexity.
Planning: longer horizon (weeks to months), aggregate decisions (capacity allocation, supplier orders, inventory positions), less granular (often weekly or monthly buckets). Scheduling: shorter horizon (hours to weeks), detailed decisions (specific work orders on specific resources), more granular (often hourly or shift-level).
Planning sets the supply commitments scheduling executes. Planning says "make 10,000 units of product X next week"; scheduling decides which lines run which orders in what sequence to achieve that. Bad planning makes scheduling impossible (can't schedule what shouldn't be made); bad scheduling makes plans unexecutable (planned output doesn't get produced).
Different decision timeframes, different responsible roles, different data requirements. Planning decisions involve cross-functional input (sales, finance, operations); scheduling decisions are operations-internal. Planning systems and scheduling systems are often separate platforms even within the same SCP suite.
Overall Equipment Effectiveness combines three components: availability (percentage of scheduled time the equipment is running), performance (actual vs. designed speed), quality (percentage of output meeting quality standards). OEE = Availability × Performance × Quality. World-class OEE is typically 85%+; most operations run 40-60%.
OEE matters most for bottleneck resources. Improving OEE at bottlenecks directly increases throughput. Improving OEE at non-bottleneck resources doesn't improve throughput (other resources already have spare capacity). Companies sometimes invest in OEE improvements at non-bottleneck resources, which improves local metrics but doesn't help system throughput.
Production driven by forecast and inventory positions. Scheduling can be relatively stable, organized around economic batch sizes and inventory replenishment cycles.
Production driven by customer orders. Scheduling must respond to actual orders with realistic delivery commitments. ATP/CTP capabilities matter substantially.
Customer orders specify configurations from standard options. Scheduling considers configuration complexity and material availability for specific configurations.
Customer orders require engineering work. Scheduling integrates engineering capacity, project management, and production capacity. Different scheduling logic from repetitive production.
