It’s a hard truth: Supply chain design data isn’t perfect. It’s never “ready” for your next network analysis.
This is usually because that data was generated in a separate context, for some other purpose. So, when it's time to optimize cost-to-serve, realign your DC footprint, or respond to supply disruptions, you’ll need a plan for using the data you have and techniques for turning it into what you will need.
For 21 years, I’ve built models using legacy shipment history, forecasts, disconnected routing guides, and the occasional tribal Excel file. It’s never perfect, and it never will be.
So I want to share the simple four-part process I use for every enterprise network modeling project at Spinnaker SCA. It’s the best way to turn the data you have into a supply chain network model that builds confidence and leads to change.
Figure 1. Four steps to turn messy or incomplete data into something actionable for your network design project
1. Start with a conversation and a data checklist
Every supply chain data project should start with a checklist and a conversation, not a model. There are several reasons to use a checklist. (Related: Don't miss our free Supply Chain Design Data Checklist here.)
The first is there is more data required than any one person can hold in their working memory. While it seems like you should be able to, it will inevitably result in errors in your modeling efforts.
Secondly, a checklist enables you to request things from your co-workers without creating resistance. Alot of the data needed for supply chain design can be closely guarded company secrets. And when you give the answer "I’m asking because it’s on the checklist,” it will save more time than you might expect when making requests.
Lastly, and maybe most importantly, you will have something to send to your colleagues (because they will ask). It will be a document to communicate your status and use for follow-ups.
With a data checklist in hand, you can now sit down with planning, logistics, procurement, and IT—not simply to demand files, but to co-create a solution to the problem you’re trying to solve. You will find data gaps. That’s expected. But what matters is how you triage them.
That’s where sensitivity bounding comes in—start by modeling the highest and the lowest possibilities. Then run a sensitivity test. Put a high-end and low-end value in the model and see if the answer changes. If the outcome doesn’t shift, then move on.
2. Don’t try to be right—just don’t be afraid to be wrong
I like to invoke Cunningham’s Law in network modeling conversations. Instead of asking a question, you purposefully give a wrong answer. The goal is for your audience to correct you. And in my experience, it’s faster than any deep-dive audit.
The first version of your model isn’t going to be perfect. But it doesn’t need to be. You just need to show something directional. This is also your opportunity to begin creating buy-in and stakeholder support.
I'm always amazed at how fast stakeholders engage when they see their world reflected back at them (even if you’ve misrepresented something). When they lean in and say, “Hang on, that’s not how we do it in the Southeast...” That friction? That’s where the insights live. That’s the gold. And now you can start collaborating.
After doing several hundred of these projects, I recommend aiming for five iterations of “build, validate, refine, and repeat.” Not because the math changes that much, but because the ownership does.
By the fifth go around, the supply chain design model isn’t yours anymore—it’s theirs. When the planners, operations leaders, and finance team see their work in the assumptions—that’s when the model sticks. By the end, your stakeholders won’t just trust the outputs—they’ll defend them. And that’s when decision confidence takes hold.
3. Normalize or die trying
Turning garbage into gold isn’t just a metaphor. It’s the reality of modern supply chain modeling.
There are two types of supply chain data: Transactional data from the past orders and shipments and your enterprise systems and design data about what could happen in the future or what is already happening outside of your company.
Design data doesn’t live in your ERP and it’s not in your WMS either. Transaction data and design data sets do not talk to each other naturally. They aren’t apples-to-apples out of the gate. So, what do you have to do for your model to make sense? You must normalize. You’ve got to clean, conform, enrich, and blend until your inputs are speaking the same language.
Don’t view this as just cleanup, it’s calibration. This step is vital as it’s what separates a valid supply chain model from a misleading one. And if you skip this step, your model is just math on top of a shaky foundation; it’s comparing apples to oranges.
4. Automate, document and build a model to outlive you
The secret to data validation isn’t a fancy tool—it’s a disciplined process: Normalization, automation, documentation, and an iterative validate-run-accept loop.
Figure 2. Best-in-class process to normalize supply chain data for automation
I don’t care what tool you’re using—Alteryx, Snowflake, Datastar, or some combination of all three. It’s not about the platform. Tools don’t make a model trustworthy—structure does. That means clear schema. Versioned folders. Transformation logic captured in scripts. And a changelog.
Can someone outside the team follow what changed, when, and why? If the answer’s no, you’ve got a black box which rarely go over well with executives trying to make million-dollar decisions. Six months from now, someone’s going to ask, “Why did we recommend closing that DC?” And you will need a clear answer and a paper trail.
So, script it. Version it. Track your assumptions. Write it like someone’s going to read months or years down the road—and probably question it. Documentation isn’t easy. Especially when you’re on iteration four. But it’s what turns your model from a prototype into a scalable decision-making engine.
The big idea here
The pursuit of perfect data is optional, but progress is not. If you’re in charge of designing tomorrow’s network or making investment-grade decisions, what matters isn’t spotless data—it’s using proven data methodologies backed by a solid structure and process.
Today's supply chains are moving too fast for perfection. But you can still find your way to credible decisions—ones that you can put in front of a COO or CFO and say, “Here’s the trade-off. Here’s the why. And here’s what happens if we don’t act.” If you’ve got a structured process, you’ll get there.
I talked about this same topic at Optilogic's OptiCon in Detroit during my session “Garbage In, Gold Out.” If you weren’t in the room, you can watch a replay or download a copy of the slides. If you’ve still got questions, feel free to reach out directly. I am always happy to have a conversation about supply chain data.
