One Customer, One Answer: Modeling Data for Consistency

This article is part one of a five-part series exploring the phases of building data readiness that your business (and eventually your AI initiatives) can run on.

Try this at your next leadership meeting. Ask your sales, finance and operations leaders one simple question: how many customers do we have?

If you get three different answers, you don’t have a reporting problem … you have a definition problem.

Sales counts everyone in the pipeline. Finance counts anyone who’s been invoiced. Operations counts anyone with an open order. Meanwhile, the CRM lists a customer as “Acme Corporation,” the ERP has “ACME Corp.,” and billing has “Acme – Buffalo.” Same company, three records, and nobody’s quite sure how to represent them as the same company.

Every answer is technically defensible. None of them match. And that’s the data consistency or modeling problem the second phase solves.

Where We Are in the Journey

As a quick refresher, we think about rebuilding the data estate in five phases:

In Phase One, we talked about consolidation: connecting your systems so your data lives in one place. That step is important, but it surfaces something uncomfortable. Once everything is side by side, you can finally see where data consistency breaks down and how much your systems disagree.

Model is where you resolve it.

Phase Two: Model

The goal is simple to say: One Customer. One Product. One Vendor. One Employee.

Each of those is built from the systems that hold a piece of it. The CRM knows the relationship and the contacts. The ERP knows the billing and payment history. Your operational system knows what was actually delivered. Collectively, that information creates a single customer view. Data modeling brings those pieces together into a single, complete picture.

The key to data consistency is deciding, for each piece of information, which system wins. The CRM might own the contact name. The ERP might own the billing address and payment terms. That decision gets made explicitly, written down and applied automatically from then on.

Notice I said decided once. Most organizations have this debate every month, usually at close, usually under pressure. Modeling ends the rerun.

Data Consistency and Modeling Require Leadership, Not Just IT

To be candid, this is the phase where executives tend to feel the most discomfort, and it’s worth naming why.

For many organizations, it’s the first time they have to make data decisions out in the open. Which department’s definition of “customer” do we use? Who has the final say on product data? What happens when two systems disagree?

Those aren’t technical questions; they are business decisions and they need business owners. The technology to match and merge records is the easy part. Getting the right people in a room to agree is the real work.

My advice: plan for the conversation, not just the build. The organizations that move fastest here are the ones whose leaders show up and make the call.

What “Done Enough” Looks Like

Not perfection. You’ll know Model is working when:

Most organizations start with customer and product, because those two tend to block the most reports. Start where the pain is.

Why it Matters for What Comes Next

Model is what makes the remaining phases possible.

You can’t define what “margin by customer” means (Meaning) until you’ve agreed on what a customer is. You can’t assign ownership of a number (Trust) if the records underneath it are in dispute. And you certainly can’t point AI at your data (Run) when three systems give it three different answers. A person reading a report quietly fills in those gaps. AI doesn’t. It just guesses.

Start with One Report

The practical way in is the same as Phase One: pick one number leadership relies on (margin by customer is a good one) and work backward. What do we mean by “customer” for this report? Which system is the authority? Who owns it?

Answer those questions for one report, and you’ve built the first piece of your model, plus the first report you can actually trust. Then do the next one.

You don’t need every definition settled to get started. You just need to stop having the same argument every month.

Next up: Phase Three (Meaning), and why “what does margin actually mean here?” is the most important question you’re not asking.