Govern, Manage, Transform: Why Your Data Needs All Three

Why do 70% of data governance initiatives fail? This figure, cited by Gartner in its data management studies, reflects a reality I've observed for 15 years in the field. After supporting organizations through their data management and transformation projects, I've identified a recurring root cause: the three essential layers — govern, manage, and transform — live in separate silos.

You know this scene. Monday morning, executive committee. The VP of Marketing presents their numbers: "The average customer spent $50,000 this quarter."

The VP of Finance frowns. Their numbers say $45,000.

Who's right? Nobody knows. Everyone blames "the data." The CEO asks to "fix it." A consultant is hired. A master data management tool (MDM) costing $500K is purchased. Six months later, the numbers still don't match.

The real problem: three silos

In most organizations, data is managed in three silos that ignore each other:

The result? Rules stay on paper. The field improvises. And automated systems transform data whose quality nobody guarantees.

This is exactly why I created CQSEV with 3 integrated layers.

→ See more real-world cases in the Field section

Layer 1 — Govern: the rules of the game

Data governance is the set of policies, roles, and standards that define how an organization manages its data. It's the "what" and the "who" — not the "how."

Governing is not about creating bureaucracy. It's about answering a simple question:

"Who decides what about the data, and according to which rules?"

Think of the Ministry of Health for drinking water. It doesn't filter the water itself. It sets the standards: "Water must contain less than 0.5 mg/L of chlorine." It appoints a person in charge. It audits the stations.

In business, it's the same. The data committee decides that "active customer" = a transaction in the last 12 months. Period. No debate. Marketing and Finance use the same definition.

In concrete terms, governing means:

• Appointing a Data Owner for each data domain
• Defining a business glossary — not 200 pages, just the 20 terms that cause problems
• Writing the 3 critical policies: quality, access, retention
• Measuring compliance with Law 25 (Quebec's privacy law) and GDPR

To dive deeper into the 5 axes of pragmatic governance, see the complete CQSEV framework.

Layer 2 — Manage: hands in the dirt

Data management is the operational enforcement of governance rules on a daily basis. It's the "how" and the "when" — the field verification that rules are actually followed.

Having rules is fine. Verifying they're applied every day is better.

Think of quality control in a food processing plant. Every morning, an inspector takes samples from the production line. They check the temperature, the expiry date, the lot traceability. If something is out of spec, they pull the product before it reaches the consumer. They don't rewrite the recipe — they enforce the existing standards.

In business, it's the Data Steward (the day-to-day data guardian) who does this work. Every week, they check the duplicate rate in the CRM (customer relationship management software). If it's above 5%, they trigger a cleanup. They don't ask the committee for permission. The rule is clear, they enforce it.

In concrete terms, managing means:

• Measuring quality every week (not every quarter)
• Fixing anomalies as soon as they're detected
• Keeping data descriptions up to date (metadata) so everyone can find what they need
• Running quarterly access reviews: who has access to what?

Layer 3 — Transform: the automated systems

Data transformation refers to the automated processes — often called ETL (Extract, Transform, Load) or ELT — that move, clean, and load data from one system to another. These are the automated pipelines that enforce rules at scale, without human intervention.

Think of a car assembly line. Parts arrive from dozens of suppliers (Extract). Each part is checked, adjusted, and assembled according to a precise plan (Transform). The finished vehicle is delivered to the dealership (Load). If a sensor detects a bolt is loose, the line stops automatically.

In business, it's the same. Your automated pipeline extracts data from the CRM, cleans it, deduplicates it, and loads it into the data warehouse (the central data repository). But if the quality rules defined in governance are not programmed into the pipeline, invalid data slips through.

In concrete terms, transforming means:

• Embedding validation rules directly into the automated pipeline
• Automatically rejecting non-compliant data
• Masking personal information — names, addresses, social insurance numbers (PII) — at extraction
• Alerting when a flow fails — not discovering the problem 3 days later

What CQSEV changes: evaluating all three at once

The CQSEV framework asks 5 questions — but each question is evaluated across all 3 layers simultaneously. Let's take Quality:

If the answer is "no" at any single level, you know exactly where to act.

In practice — the customer address

Let's see how the 3 layers work together on a concrete example everyone can relate to: a customer's mailing address.

That's CQSEV applied in practice. Governing without managing is fiction. Managing without transforming is improvisation. Transforming without governing is blind automation.

Where to start?

For each CQSEV axis, ask your team:

1. Does the rule exist? (Govern)
2. Is someone verifying it? (Manage)
3. Does the automated pipeline enforce it? (Transform)

The empty cells are your action priorities.

→ See the complete CQSEV matrix (5 axes × 3 layers)

Final word

In 15 years, I've never seen an organization fail because it didn't have the right tool. I've seen them fail because:

• ×The rules existed but nobody knew they existed
• ×The verification was done by nobody
• ×The automated systems completely ignored the policies

Governing without managing is fiction. Managing without transforming is improvisation. Transforming without governing is blind automation.

All three together — that's reliability.