What Is “Good Data”?

Picture this: your CEO sits at their desk, excited to test the company’s new AI dashboard. The CEO types, “How’s the quarter looking?” and the chatbot confidently replies: “Revenue is up 20%.”

Great news, right?

Except that one “advance” invoice, booked early to lock in a customer payment, made the numbers look stronger than they really were. A couple hours later, the CEO jumps on a call with a board member and gushes about a record quarter that never actually happened.

Fast forward: the board’s confused, the CEO’s embarrassed, and now every report — and every answer gets triple-checked.

This is what bad data does.

And it’s why I keep writing: AI doesn’t fix data problems, it amplifies them.

The Foundation: AI Works Only as Well as the Data It Touches

AI is not a magic filter that can turn messy, inconsistent, or incomplete information into insight. It works with what it has, and that means only good data can produce good outcomes.

Good data isn’t an accident. It’s the result of deliberate, thoughtful, and forward-looking design, applied consistently across an organization.

So, what exactly is “good data”?

The Six Dimensions of Data Quality

SAP defines quality data through six core dimensions — a framework that’s especially relevant to finance teams today. (IBM also has a similar piece here):

Dimension	Definition	Key Question
Accuracy	Data correctly represents the real-world entity or event.	Does this record reflect the truth?
Completeness	All required data is present.	Are mandatory fields filled?
Context	Data includes meaning and metadata.	Do we understand what this data represents?
Consistency	Data is uniform across systems.	Are values standardized everywhere?
Timeliness	Data is up to date and available when needed.	Is the data current at decision time?
Uniqueness	No duplicates or redundant records exist.	Are there duplicate entries?

A Real-World Example: The October Invoice Report

Let’s apply these dimensions to something every finance team deals with — an invoice report for say October.

Good data would mean:

• Every invoice in the report is accurate.

• The report contains all invoices for the period.

• The report’s purpose is clear — invoices, which is not necessarily the same as sales.

• Entries are consistent across subsidiaries (similar sold items go to the same GL account).

• The report reflects the company’s position at this moment in time.

• No duplicates or double entries exist for the same transaction.

That’s what “quality” looks like.

Now, let’s see how it falls apart.

How “Bad Data” Creeps In

Even simple processes can go wrong in surprising ways:

• Different subsidiaries close books on different schedules — so the report is incomplete.

• One team doesn’t finish invoicing on time.

• Similar products are booked under multiple GL accounts.

• A customer has three accounts in the system and the wrong one is used.

• One subsidiary treats “invoices” as “sales,” another doesn’t due to complex revenue recognition rules.

Just one of these issues can turn your report into fiction.

Now imagine what happens when AI starts drawing insights from dozens of datasets, each with similar inconsistencies.

The CEO’s Chat With Bad Data

Back to our CEO. They’re “chatting with the company’s data,” but behind that polished AI interface are spreadsheets, APIs, and databases filled with inconsistencies.

How many things have to go perfectly across dozens of systems, invoicing, payroll, CRM, procurement, for the answers to be trustworthy?

That’s the uncomfortable question every finance leader should be asking before turning AI loose on their data.

The Takeaway

Good data isn’t about technology — it’s about discipline.

It’s:

• Data that’s accurate and complete.

• Defined consistently across teams.

• Updated in real time.

• Trusted enough that when AI “chats” with it, you can believe the answer.

Because in finance, there’s no such thing as “AI-ready” — only data-ready.