Mixed Model Sequencing: Choose an Order That Works

Home » Resources » Mixed Model Sequencing: Choose an Order That Works

Estimated reading time: 10 minutes

The shift contains the right number of every model, yet the line keeps stopping. Three wiring-heavy units arrive together. The operator falls behind, while the easier units that would allow recovery are still further down the queue.

The daily mix is correct. The order is not.

Mixed model sequencing means choosing the order in which different product variants enter an assembly line so their changing work content can be handled at the stations. A useful sequence does more than spread model names evenly: it gives each constrained station a physically possible pattern of work and recovery.

Then that sequence needs a place in the factory’s wider plan. SkyPlanner APS brings variant-specific operations, material availability and shared resources into one production schedule, reducing the manual work of reconciling an assembly list with separate test and workstation queues.

A balanced mix does not prove a workable sequence

Production leveling, or heijunka, smooths volume and mix over a period. Line balancing assigns work elements to stations. Mixed model sequencing chooses the actual order of units through that arrangement.

These decisions support one another, but answer different questions. Making the required mix every shift does not establish whether two complex units can follow one another. An acceptable average station time does not show how far behind that station gets before a lighter unit arrives.

The distinction matters because “heavy” is local. A variant with extra wiring may be demanding at assembly but quick at testing. Another may be simple to assemble and need a longer test. Alternating models helps only if their actual station loads complement one another.

Start with the model-by-station work content, not a single label for the difficulty of the whole product.

Establish what each station can actually absorb

For each relevant variant, record the work required at each constrained station. Include normal handling and checks, option-specific work, and any setup that changes with the preceding model. Separate unattended machine time from work that occupies an operator; their overlap changes the capacity needed.

Use representative durations from a defined method. The fastest observed cycle is a poor basis for a sequence the team must repeat all day.

Next, establish the planned interval between units arriving at the station. This release interval is the time budget used in the sequence check. Customer takt describes the required demand pace; the interval you actually plan to run must also account for the production losses and recovery margin your line needs.

Finally, agree what happens when a unit needs more time than that interval:

• On a rigidly indexed line, work may have to finish before the next transfer. A lighter unit later cannot rescue an earlier missed transfer.
• On a continuously moving line, an operator may have an approved work zone that allows work to extend beyond one nominal interval. Access, walking, interference and the station exit limit still apply.
• On a buffered, less tightly coupled line, a station may temporarily build a queue and catch up. Check both the queue before it and the risk of starving or blocking neighboring stations.

Only use recovery time that the physical process really permits. It is not permission to skip a check, assume an operator will speed up or borrow a person who is already committed elsewhere.

Example: the same mix, very different peaks

Imagine a line introducing one unit every 60 seconds. It makes three units each of variants A and B. Two stations have these simplified work times:

Variant	Station 1	Station 2
A	78 seconds	42 seconds
B	42 seconds	78 seconds

Each station averages 60 seconds per unit across that mix. The team is comparing two orders: A A A B B B and A B A B A B.

For this illustration, assume each station has its own operator, the unit order stays the same, and the physical arrangement permits limited work to carry into the following interval. Times are fixed, with no additional setups or interruptions. We are checking each station’s local workload, not simulating the whole line.

Track the unfinished work remaining at each next arrival:

Next carryover = the greater of zero and (previous carryover + this unit’s work time − release interval).

Start with the work already outstanding, not automatically zero. Never carry a negative balance forward as credit: spare time before a unit arrives cannot be used to assemble that future unit.

At station 1, starting empty, the grouped order leaves 18, 36, 54, 36, 18 and 0 seconds of carryover after its successive intervals. The alternating order leaves 18, 0, 18, 0, 18 and 0 seconds.

The average has not changed, but the peak has:

Candidate order	Peak carryover at station 1	Peak carryover at station 2
A A A B B B	54 seconds	54 seconds
A B A B A B	18 seconds	18 seconds

Suppose the team has physically validated a maximum carryover allowance of 24 seconds at each station. The grouped candidate exceeds it; the alternating candidate stays within it under these assumptions. That 24-second allowance is an imagined property of this example, not a recommended buffer for other lines.

Station 2 also shows why the end matters. Starting empty, the alternating order finishes its last B interval with 18 seconds still outstanding. If the pattern repeats, the next A provides enough spare time to clear it. If production ends there, the remaining work still needs time to finish. A shift boundary does not erase it.

Both stations are loaded exactly to the release interval on average, so this example has no average capacity margin for disturbances. The alternating pattern is the better deterministic candidate, not proof of a robust production rate. Verify actual variation, transfers and recovery on the line before committing to it.

Choose the sequence against every constrained station

Use the same ordered unit list for the station checks. For each station, compare its highest accumulated work with its permitted recovery limit, and confirm that it recovers often enough to keep running. Include the incoming workload from the preceding production window and the join between repeating patterns.

A useful selection process is:

1. Keep the required quantities and commitments visible. Do not improve the workload trace by quietly dropping difficult units or postponing a customer order beyond its delivery window.

2. Identify the tightest station patterns. Look for consecutive high-work options, not only consecutive identical model codes. Different variants can carry the same demanding option.

3. Place genuinely lighter work where recovery is needed. Recheck all affected stations after each change. An improvement at wiring can create an unacceptable run at testing.

4. Compare feasible candidates on the remaining costs. Include setups, component presentation and delivery timing. The fewest changeovers is not the best result if it causes repeated line stops.

5. Prove the candidate under representative conditions. Check standard work, quality, ergonomics and the actual recovery space with the people running the line.

Where an option rule is used, make it explicit. “No more than two units with this option in any four consecutive units” is a moving-window rule, not a daily percentage. Check overlapping windows and the joins between lists. Such a rule can be a useful operational restriction, but it should reflect the station’s demonstrated ability to recover rather than replace that understanding.

If the average workload already exceeds the station’s available time at the required mix, rearranging units will not create the missing capacity. Improve the method, reassign suitable work, provide justified capacity or change the production pace and commitments. Sequencing manages variation; it does not remove work.

Make the variant plan work with the rest of the factory

A locally acceptable sequence can still fail if its parts are unavailable or its shared tester is busy with another order. Maintaining those conditions in separate lists means repeating the same reconciliation whenever the mix or a delivery changes.

This is where we recommend SkyPlanner. Define the actual process steps, durations, eligible workstations and dependencies for the variants, rather than scheduling an averaged product that hides the extra work. Arcturus, SkyPlanner’s scheduling AI, places those operations against configured shifts, capacity and planner-set priorities. The Gantt timeline shows the resulting work across resources.

For a shared test station, that means scheduling the variant’s real test operation alongside the other orders that need the same equipment. Assembly does not get one promise for the tester while another department makes a conflicting promise in its own queue. You can review the effect on completion timing and decide whether to change priority, use another eligible resource or provide more capacity.

With material-aware scheduling enabled, Arcturus also considers maintained inventory and incoming shipments when calculating start times. A changed component arrival can therefore be considered together with the remaining route, rather than repaired in a separate material list after the capacity plan has been made.

Keep setup reduction and workload recovery distinct. SkyPlanner’s Property Grouping can favor consecutive jobs with a selected common property to reduce setups. That is a different objective from spacing heavy options apart; similar jobs can all be demanding at the same assembly station. Apply grouping where it helps the real process, not as evidence that the line’s recovery check has been satisfied.

For a paced line with option-window or conveyor work-zone limits, bring those exact rules into the scheduling design and confirm how they will be represented and checked. The local workload test above and the coordinated factory schedule serve different purposes. You need both a physically acceptable unit order and realistic availability for the resources that execute it.

Once upcoming work has been checked and prepared, SkyPlanner’s Running Time Lock can protect jobs inside a defined near-term window from rescheduling while other work is recalculated. Choose that window around your preparation and response needs. A real shortage still needs a decision; protecting a queue does not make missing parts available.

You can also work through your own AI assistant (for example, Claude, ChatGPT or Microsoft Copilot) connected to SkyPlanner via MCP, the connection that lets the assistant use SkyPlanner’s functions. Describe the planning change you need; the assistant uses its SkyPlanner product knowledge and the relevant functions, reducing the need to navigate each step yourself.

The benefit is less time rebuilding connections between the mix, assembly operations, testing and the rest of production. You set the production priorities and protect committed work; Arcturus handles the linked scheduling calculation.

Give the line one current version of the plan

The sequence also has to reach the people who use it. A last-minute swap can change local workload, component presentation and downstream timing, even if the model counts stay the same. Recheck those effects before releasing a revised order, and make clear which version production should follow.

UTU, a Finnish switchboard and switchgear manufacturer, previously planned with department-specific spreadsheets. Production-sequence changes in its ERP did not always reach the shop floor, and realistic capacity was difficult to see.

With SkyPlanner connected to Oracle NetSuite, UTU brought its Ulvila factory’s lines and departments into one planning system. The team highlights AI scheduling together with the Gantt timeline and capacity visibility, and reports that an updated work queue reaches installers on the line in seconds.

That experience illustrates the value of a shared, current queue. For mixed-model production, pair that visibility with the station checks: the team needs the approved sequence, not merely the latest list of the right quantities.

Coordinate your next mixed-model production window

Choose a model order that the stations can recover from, then put its actual operations into the same plan as the resources and orders around it. This is the practical case for SkyPlanner APS: variant-aware factory scheduling with less manual queue reconciliation and clear control over priorities and near-term work.

SkyPlanner integrates with any ERP or MES through its open REST API. You can keep the system that manages your orders and connect its production data to a coordinated schedule.

Book a SkyPlanner demo with a few real variants, their station times and materials, a shared-resource conflict and any mandatory sequence restrictions. Use those inputs to assess the scheduling setup and how a mix change affects the rest of production. If you prefer to explore first, try SkyPlanner with demo data or use the AI-assisted Getting Started in SkyPlanner Assistant.