SKU Changeover Sequence Optimization: How Order Decides How Much Time You Lose

Key takeaways

• Changeover sequence = the order in which different SKUs are scheduled on the line.
• Changeover time between SKUs A→B is often different from B→A and very different from A→C.
• Optimizing the sequence to favor low-cost transitions cuts total changeover by 20-40% without SMED.
• The math is a traveling-salesman problem on the changeover matrix.
• Most plants schedule by demand, not by changeover cost. Mixed scheduling captures both.

Short answer: Changeover times between different SKU pairs are not equal. A→B may take 30 minutes; A→C may take 5; B→D may require a deep clean. The total changeover time across a week depends heavily on the SKU sequence. Optimizing the sequence to favor cheap transitions can cut total changeover by 20-40% without any SMED work. Plants typically schedule by demand and miss this. See also SMED vs Changeover.

What sequence optimization is

Given a set of SKUs to produce in a week, there are many possible sequences. Each pair-wise transition has a changeover cost (time). The total changeover time is the sum across the sequence.

Two different sequences with the same SKU mix can produce very different total changeover times — the difference is sequencing.

Why changeover times vary by pair

Several reasons:

• Tooling. Some SKU pairs share tooling; others require change.
• Cleaning. Allergen, color, or material changes require deep clean.
• Setup. Some SKUs need full re-calibration; others a minor parameter change.
• First-piece confirmation. Variable per SKU.

The changeover matrix captures all of this.

Building the changeover matrix

For N SKUs, an N x N matrix where cell (i,j) is the changeover time from SKU i to SKU j.

• Off-diagonal cells contain real changeover times.
• Diagonal is zero (no changeover from a SKU to itself).
• The matrix may not be symmetric (A→B vs B→A can differ).

Filling the matrix requires observation. Estimates from memory are usually wrong.

What sequence optimization does mathematically

The problem: given the SKU mix to produce, find the sequence that minimizes total changeover time.

This is a variant of the Traveling Salesman Problem (TSP). For small SKU counts (under 10), brute force works. For larger sets, heuristics (nearest neighbor, simulated annealing) find good sequences quickly.

Commercial scheduling tools handle this. Spreadsheet approximations work for many plants.

The constraints

Pure sequence optimization may produce schedules that violate other constraints:

• Customer due dates. Cannot defer SKU X to next week if it ships tomorrow.
• Quality requirements. Some sequences are prohibited (allergen, color contamination).
• Material availability. Cannot run SKU before raw materials arrive.
• Equipment limitations. Some sequences require longer breaks.

Real optimization respects these. The math becomes constrained TSP.

How much savings are possible

Plants without sequence optimization typically have 20-40% more total changeover than optimal. Capturing half of that gap (10-20%) is achievable without process change.

For a plant with 8 hours of changeover per week, a 15% reduction is 1.2 hours per week — about 60 hours per year of additional production time.

Common patterns

1. Demand-driven scheduling. Schedule by customer due date; accept the changeover cost. Common, leaves money on the table.

2. Mixed scheduling. Schedule by demand windows; within each window optimize sequence.

3. Frequency-based sequencing. Frequently-produced SKUs cluster together; rarely-produced ones get their own slots. Reduces total changeovers.

4. Family-based sequencing. Group SKUs by family (similar materials, similar processes). Transitions within family are short.

How to start

1. Build the changeover matrix. Observe pair-wise times for 2-4 weeks.
2. Identify the worst transitions. Pareto by cost.
3. Add constraint list. Quality, customer, material.
4. Try a manual sequence optimization. Spreadsheet often enough for small SKU sets.
5. Measure the gain. Compare new vs old schedule changeover total.
6. Standardize. Build the optimization into recurring scheduling.

Common mistakes

1. Estimating changeover times. Observation is essential; estimates are usually wrong.

2. Ignoring asymmetry. A→B and B→A often differ significantly.

3. Optimizing without constraints. Mathematically optimal sequence violates customer or quality requirements.

4. Treating sequence as fixed. Sequence should be optimized each scheduling cycle, not set once.

How OEE relates

Changeover is Availability loss. Sequence optimization reduces this loss without changing the changeover process. Combined with SMED (which reduces per-transition time), the gains compound.

How a modern OEE platform supports sequencing

A modern OEE platform measures changeover times per SKU pair, builds the changeover matrix automatically, and integrates with scheduling tools for sequence optimization.

Fabrico's OEE module captures changeover times per SKU pair, builds the changeover matrix, and surfaces optimization opportunities for scheduling.

See how Fabrico captures this automatically — explore OEE for manufacturing or book a demo.

Related reading

• SMED vs Changeover

Frequently asked questions

How big a SKU set does this work for?

From 3 SKUs to 50+. For 50+, automated scheduling tools become essential.

Should I do SMED or sequence optimization first?

Sequence optimization is usually faster to capture. Then SMED on the remaining largest transitions.

How often should I rebuild the changeover matrix?

When SKUs, tooling, or process change. Otherwise every 6-12 months.

Does this matter for low-mix plants?

Less so. The savings scale with SKU mix and changeover frequency.

What software supports this?

APS (Advanced Planning and Scheduling) tools handle this natively. Spreadsheet approximations work for small SKU sets.