What Is Data Quality, and Why Should You Care?

Data quality isn't a technical problem. It's a governance problem.

And the difference matters more than you think.

When we talk about dirty data, the most common reaction is to point at the engineering team. "The pipeline's broken." "The data model is a disaster." "Nobody documented anything."

But here's the uncomfortable truth: most data quality problems aren't a technical failure. They're a governance failure.

And if you don't understand that difference, you'll keep putting out the same fire over and over.

What is data governance, really?

So what is data governance, really? It's the blueprint for how your organization controls, organizes, and moves its data. It spans everything from raw system logs to the polished dashboards your CEO uses to make decisions.

It's not a committee. It's not a policy sitting in a PDF nobody reads. It's the structure that defines who owns the data, how it's validated, how it's transformed, and how it's used.

And data quality is a direct consequence of how well that governance is implemented.

"Garbage in, garbage out" doesn't cut it anymore

For years, the mantra was simple: if garbage goes in, garbage comes out. The blame fell on the user who entered the data wrong.

That doesn't hold up anymore.

If your system lets garbage in, the system is the problem. And you built the system.

We data engineers have to own this: we're the architects of the rules that validate (or fail to validate) the data. When a date field accepts "99/99/9999," that's not a user error — it's a design error.

Medallion Architecture: the framework that starts governance

One of the best tools for implementing governance with data quality is the Medallion Architecture a layered approach. The idea is simple: you organize data into layers with increasing levels of trust.

• Bronze: raw data, exactly as it arrives from the source. No transformations. The faithful historical record.
• Silver: cleaned, validated, and enriched data. This is where the quality rules apply.
• Gold: business-ready data. Certified metrics, official KPIs, the stuff that feeds dashboards and ML models.

The key isn't the architecture itself — it's what it represents: every layer has a quality contract. Nobody should be able to skip Silver and use Bronze straight in a business dashboard.

If your architecture allows that, you've got a governance problem, not a data problem.

What about AI? Here's the Achilles' heel

Everyone wants AI. Everyone's investing in AI. And almost nobody is investing enough in the quality of the data that's going to feed that AI.

A Generative AI (Gen AI) model or a machine learning model trained on unvalidated Bronze data will:

• Hallucinate with more confidence than ever
• Perpetuate biases that exist in the historical data
• Generate predictions nobody can explain or audit

AI doesn't solve data quality. It amplifies it — for better or worse.

Three questions you should be asking

If you're in a data leadership role, these three questions are your starting point:

1. Where does that number come from? Before acting on any metric, trace its lineage. Bronze? Silver? Was it certified? Who's the data owner?
2. Do we have well-defined data layers? Or is everything one big lake where raw and processed data get mixed together and nobody knows what to trust?
3. Who's responsible for each data domain? Governance without ownership is just theory. Every data domain needs a steward — someone who answers for the quality of that data and has the business context to define the rules.

The mindset shift towards Data Quality and Data Governance

Data quality isn't the exclusive responsibility of engineers, nor of the business. It's shared — and that requires governance.

As data professionals, we have to stop acting like plumbers who only fix what bursts and start acting like architects who design systems where quality is the default, not the exception.

This is where data governance best practices earn their keep: modern data classification systems, combined with frameworks like the Medallion Architecture and a clear governance model, let you automate a big chunk of the validation and make quality scale with volume.

But the first step is always the same: stop treating data quality as an isolated technical problem and see it for what it really is — a function of governance.

How is your organization handling data governance? Do you have well-defined layers, or does everything live together in the same swamp? Share your experience in the comments.

‍