Production Scheduling Expert Skill - Constraint Optimization and Dispatching

Scheduling: A Small Task That Can Really Be a Killer

Anyone who has worked in production scheduling knows that this job looks simple on the surfacebut actually doing it can drive you baldEvery day you face dozens or even hundreds of work orderseach with different routingdifferent due datesdifferent changeover requirementsand you also have to consider when machines are down for maintenancewhich shift workers are onand whether materials can keep upHonestlywhen I was an intern in a factory back in the daywatching the veteran schedulers at work was like watching magicThey seemed to have a living map in their heads—which job goes firstwhich job can waitwhich machine is the bottleneck—they knew it all cold.

But the problem isthat kind of experience is very hard to replicateWhat happens when the veteran retires? Newcomers have to stumble and fumble through years of trial and error to get up to speedSo when I came across this Production Scheduling skillI was genuinely excited—it takes the knowledge of a scheduler with over 15 years of experience and codifies it into a knowledge base that can be called upon at any timeIt's basically like loading a veteran scheduler's brain into a computer.

What Can This Skill Actually Do?

In simple termsit helps you make scheduling decisionsIt doesn't just hand you a "theoretically feasible" plan like an ERP does; insteadit thinks like a real schedulerconsidering actual constraints and delivering a schedule that can actually be executed on the shop floorIt mainly tackles these types of problems:

• Bottleneck Identification: Uses OEE data and capacity utilization rates to help you identify the true constraint resource in the system—that isthe "Drum."
• Job Sequencing: Selects appropriate dispatching rules based on order priority (overdueconstrained supplynormal)such as EDD (Earliest Due Date)SPT (Shortest Processing Time)or setup-aware EDD.
• Changeover Optimization: Uses a setup matrix combined with nearest-neighbor heuristics and 2-opt optimization to reduce wasted changeover time.
• Disruption Response: When equipment breaks downmaterials run shortor workers call in sickhow to quickly reschedule without destroying the already-frozen plan.
• Line Balancing: Distributes workload across multiple production lines to avoid some lines being overloaded while others sit idle.

It's especially well-suited for discrete and batch manufacturing companies with 3-8 production lines and 50-300 direct laborersand it works for machiningassemblypaintingor packaging operations alike.

A Few Things That Made Me Go "Wow"

Firstit distinguishes between forward scheduling and backward schedulingI honestly hadn't paid much attention to this distinction beforebut after reading the explanationit clicked: by defaultit uses backward scheduling (working backward from the due date)which preserves more flexibility and reduces work-in-process inventoryHoweverif it detects that the latest start time for a given order has already passedit switches to forward schedulingstarting from today and pushing forwardThis approach is very practical—many novice schedulers stumble precisely because they "only know how to schedule forward."

Secondit emphasizes the difference between MRP's infinite capacity planning and finite capacity schedulingThe plan generated by MRP only tells you "what needs to be done," but whether it can actually be doneand whenmust be verified through Finite Capacity Scheduling (FCS)This skill constantly reminds you: don't ever take an MRP plan and execute it directlyor the shop floor will definitely descend into chaos.

And then there's the Drum-Buffer-Rope (DBR) method—an absolute classicThe Drum is the bottleneck resourcethe Buffer is the time buffer that protects the bottleneck (note: not an inventory buffer)and the Rope is the mechanism that controls the rate of material releaseThe entire system operates at the rhythm of the bottleneck: upstream must not over-release materialand downstream must ensure the bottleneck never goes hungryApplying this logic to scheduling delivers immediate results.

How Do You Get Started?

Installation is simpleOpen your terminal and run:

npx skills add https://github.com/agentmatters/mullai-bot --skill production-scheduling

Once installedrun it inthen feed it the scheduling data you have on handFor example:

• Work order list (with due datesroutingand BOMs)
• Line capacity (number of machinesshiftsmaintenance windows)
• Changeover time matrix
• OEE data

It will provide a complete set of scheduling recommendationsincluding which orders to run firstwhich to run laterwhich can be delayedwhere the bottleneck liesand how to optimize changeovers.

A Real-World Example

Let's say the workshop you're responsible for has three linesand suddenly the CNC machine on Line 2 breaks down and needs four hours to repairMeanwhileyou have 15 orders waitingand three of them are already past dueWhat do you do?

The skill guides you through it: firstidentify the bottleneck (Line 2 might be the bottleneck itself)then check which orders can be transferred to Line 3 (alternative routing)which ones absolutely must waitand finally re-sequence the remaining queue to minimize total tardinessThe whole process follows a clear framework—no panickingno flailing.

Another example: you need to decide between batch cycling and mixed-model schedulingYou have 15 orders spanning four product familiesand changeovers take 45 minutesAt this pointyou need to calculate the crossover point between changeover cost and holding costThe skill provides a clear calculation method for this as well.

Let's Be Realistic

This skill isn't one of those "install it and you're done" toolsIt's more like a knowledge framework and a decision toolboxTo truly get the most out of ityou'd be well advised to first understand concepts like DBROEEand SMEDand then adjust the parameters to fit your factory's actual conditions.

Alsonote that it emphasizes the concept of a frozen window (typically 24-48 hours)In other wordsorders that have already been promised to customers should not be lightly changed during the frozen periodor it will trigger a chain reactionThis point is especially important—many schedulers reschedule too casuallychanging things today and changing them again tomorrowand as a resulton-floor execution effectiveness takes a serious hit.

Anywaymy experience using it has been this: it has saved me a lot of detoursAt the very leastI now know what order to think through things when unexpected situations ariseIf you're also doing production schedulingor if you're losing sleep over bottlenecks and changeoversI sincerely recommend giving this skill a try.

Alrightthat's all I have to say for nowI need to go schedule tomorrow's plan 😄