When building a supply chain network design model, it’s essential to strike the right balance between detail and simplicity. A well-designed model gives you fast, accurate results. A bloated or overly complex one may grind to a halt or generate misleading insights. This article outlines best practices to guide you in sizing your network design model correctly—keeping it lean, focused, and powerful.
1. Start with the Right Mindset
Optimization is about trade-offs. The goal is to find the minimum or maximum value (e.g., cost, profit, service level) in your supply chain through a mathematical model. The smaller and more focused your model, the faster it will solve. But too much simplification can hide critical insights.
🧠 Key Principle: Include just enough detail to capture the real dynamics of your supply chain, but no more.
2. Rule of Thumb: Simplicity with Purpose
Keep your model as simple as possible—without compromising the quality of your decisions. Don’t model every nuance of the business unless it directly affects the outcomes you're trying to optimize.
Tip: Always ask, “Will this extra layer of detail change my decision?” If the answer is no, leave it out.
3. Know Your Limits
Two practical thresholds define the model’s feasibility:
Excel Threshold
- 1 million rows (if not connected to a backend database)
- Going beyond this limit makes data handling and review difficult, even before optimization begins.
Solver Complexity Threshold ("Rule of 50")
- 50 million combinations of Transportation lanes is a typical upper limit for practical model size.
- This is roughly calculated for outbound models as:
Products × Time Periods × Warehouses × Customers × Modes of Transport
The closer you get to 50M, the slower your model. Above it, expect inefficiencies or failure to solve.
4. Recommended Design Exercise
Here’s how to make an informed decision about model size:
Step-by-step Exercise:
- List all key dimensions: Products, Periods, Warehouses, Customers, Modes.
- Define 3–4 levels of granularity for each (e.g., for Customers: individual → postal code → city → region).
- Pick your preferred level for each, prioritizing aggregation without losing insight.
- Calculate total combinations using the rule of 50 formula.
- Validate the total:
- < 50M: Safe zone.
- Close to 50M: Optimize further.
- > 50M: Re-do the aggregation—your model is too complex.
Example
Dimension | Options | Chosen Level |
Products | SKU → Product Family → Total | Product Family |
Periods | Monthly → Quarters → 1 Year | 1 Year |
Warehouses | Each Site → Region | Each Site |
Customers | Location → City → Postal Code | City |
Modes of Transport | Truck → All Modes Combined | Truck |
Multiply: 500 products × 1 year × 15 warehouses × 1000 customer Cities × 2 modes = 15M combinations
5. What If You're Over 50M?
If your model exceeds 50 million combinations, you’re trying to do too much at once. Here are options to reduce scope:
Segment Your Models:
- Inbound model (e.g., supplier to plant)
- Production & manufacturing model
- Outbound model (e.g., DC to customer)
Use specialized models for each part of the supply chain to reduce dimensionality and tailor granularity.
🧩 Example: One global consumer goods company separated its strategic customer allocation model from its tactical monthly planning model, improving solve time by 75%.
6. Reduce the Number of Time Periods
Time periods are often overrepresented in models. Consider these simplifications:
- Model peak and off-peak seasons only
- For strategic decisions, use just one representative period
- Separate strategic (yearly) and tactical (monthly/weekly) models
🧠 Expert Insight: Many network design decisions don't change significantly whether you use 12 periods or 4, especially for long-term planning.
7. Product Aggregation is Powerful
Many modellers want to include detailed SKU-level data. While valid in some cases, aggregating products often yields equally useful results, especially for high-level questions.
- Group SKUs by volume, weight, or handling requirements
- Sometimes, total volume or total weight is enough to answer strategic questions
Real-world Tip: A global logistics company saved weeks of modeling time by using product cube and weight alone—answering 90% of its strategic sourcing questions.
8. Don’t Overlook Clustering & Node Aggregation
Node aggregation is one of the most effective techniques to simplify your model without losing value:
- Customer aggregation: Group by zip code prefix, city, region
- Supplier aggregation: Use a “global supplier” to represent multiple sourcing options
Case Study: A retailer with over 10,000 delivery points grouped customers into 150 zones by delivery frequency and location—reducing solver time from 6 hours to 10 minutes, with <1% impact on total cost.
Summary Guidelines
Guideline | Benefit |
Simplify dimensions | Faster solve time |
Aggregate wherever possible | Maintain insight, reduce size |
Use the Rule of 50 | Avoid solver overload |
Separate models by flow type | Clearer scope, cleaner results |
Limit time periods | Easier to manage, interpret |
Minimize product detail | Avoid unnecessary complexity |
Final Thought
Modeling is an art as much as a science. Start small. Keep it simple. Then grow only if needed. Your network design model is a decision-support tool, not a digital twin of every transaction. By designing your model wisely, you’ll get faster results, clearer insights, and better decisions.