Data Governance Manager interview questions, with worked answers

How to use these questions

Interviewers often describe a messy situation and ask where you would start. Start small and visible: one domain, a few critical fields, owners who care. Then show how it spreads. Give a real domain you governed, what changed in people's work and what the figures showed.

What a data governance manager is expected to know

• The practice: data governanceSet who owns which data, which rules apply to it and how decisions about it are made., from roles and policies to the council.
• The quality: data qualityDefine, measure and improve the quality of the data the business relies on, rule by rule. rules, scores and fixes at the source.
• The knowledge: metadata and lineageKeep definitions, lineage and owners of data recorded so that people can find and trust it., the glossary and master dataGovern customer, product and supplier records so each exists once and is correct everywhere..
• The rules: data protectionKnow what data protection law asks of data handling, and work with the DPO where it applies. with the DPO, and in banks BCBS 239Know the Basel Committee's principles for risk data and what they ask of banks' data., the Basel Committee's international principles for risk data.

The catalogue tools are named in ads side by side and none is a must: Collibra, Alation, Microsoft Purview, Informatica and Atlan each hold a glossary, a catalogue and lineage. Interviewers care more about whether people use the catalogue than about the product; some ask about DAMA-DMBOKKnow the areas DAMA-DMBOK describes and use its vocabulary with other data professionals. terms, which describe the same ground.

Owners, stewards and quality

What is a data owner, and how do you get the business to take the role?

What they check: the decision right at the heart of governance.

Short answer: "A data owner is the business leader accountable for a set of data, such as customer data: they decide its definitions, who may use it and what quality it needs. I ask the leader whose results suffer most from bad data, show them the cost, and keep the role small: a few decisions a quarter, with stewards doing the daily work."

The full answer: Owners in IT do not work; the person who loses money when the data is wrong should decide about it.

A weak answer: "IT owns the data." Follow-up: "Two directors both claim customer data." Say how you settle it.

What does a data steward do day to day?

What they check: the people who make governance real.

Short answer: "A data steward is the person in the business who looks after a data domain every day: they write and check definitions, watch the quality scores, fix or route problems, and answer questions about what the data means. Most stewards do this as part of another job, so I keep their tasks clear and their time agreed with their managers."

The full answer: Stewards need a few hours a week written into their goals, or the role disappears in the first busy month.

A weak answer: "Stewards are the data team." Follow-up: "Your best steward is moved to another project." Say what you do.

How do you write a data quality rule and measure it?

What they check: turning "good data" into checks.

Short answer: "A data quality rule states what correct data looks like for one field or record, such as every active customer has a valid postal code, with the owner's agreement on the target. I run the rule on the data regularly, show the score per domain and per source system, and send failures to the stewards to fix at the source, not in the report."

The full answer: Pick rules by business impact; a hundred rules nobody acts on cost more than five that change a process.

A weak answer: "We clean the data in the warehouse." Follow-up: "The score goes down after every release of the sales system." Say where you go.

What is data lineage, and when is it worth the effort?

What they check: knowing where figures come from.

Short answer: "Data lineage shows where a piece of data comes from and every step it passes through until it reaches a report or a model. It is worth the effort for the figures that matter most, such as board and regulatory reports, where a wrong number has consequences. I start there, partly from tools that read the pipelines, partly by asking the people who built them."

The full answer: Lineage for everything is never finished; lineage for the critical figures is finished and used.

A weak answer: "The tool draws all lineage automatically." Follow-up: "A key figure passes through a spreadsheet." Say how you capture it.

How do you build a business glossary people actually use?

What they check: shared meaning, the cheapest win in governance.

Short answer: "A business glossary lists the terms the business uses, such as active customer or net revenue, each with one agreed definition, an owner and the data it maps to. I start with the terms people argue about in meetings, agree them with the owners, and link the glossary from the reports, so people meet it where they work."

The full answer: A short glossary that settles arguments beats a long one that copies a dictionary.

A weak answer: "We imported a standard list of terms." Follow-up: "Finance and sales define revenue differently, and both are right." Say what goes in the glossary.

How would you choose and roll out a data catalogue?

What they check: catalogue toolsRun a data catalogue that people search every day, not a list that ages after launch. that people adopt.

Short answer: "I start from who will search it and for what, analysts finding data, owners approving access, then compare tools such as Collibra, Alation, Microsoft Purview, Informatica or Atlan on those uses, the systems they connect to and the cost. I launch with one domain that is complete and correct, not every system half described."

The full answer: The catalogue lives or dies by its first weeks; if people find wrong entries, they do not come back.

A weak answer: "We load all metadata and let people explore." Follow-up: "Six months after launch, nobody searches it." Say what you change.

How do you govern master data such as customers and products?

What they check: master dataGovern customer, product and supplier records so each exists once and is correct everywhere., where errors spread furthest.

Short answer: "Master data is the core records many processes share, such as customers, products and suppliers. I agree which system creates each record, who may change it, the rules for duplicates and the fields that must be complete, and I measure those, because a wrong address or a duplicate customer travels into every system downstream."

The full answer: Fix the way records are created first; cleaning without that only buys time.

A weak answer: "We merge duplicates once a year." Follow-up: "Three systems create customers." Say which one should.

Policies, the council and critical data

Which policies would you write first, and how short would they be?

What they check: rules people can follow.

Short answer: "A data governance policy of a few pages that sets the roles and the decision rights, then short standards for quality, definitions and access. Each says who does what and how it is checked; the detail lives in the catalogue and the procedures, which change more often."

The full answer: A policy nobody has read protects nobody; test it by asking a steward what it asks of them.

A weak answer: "A complete policy framework first." Follow-up: "The board wants the policy in a month." Say what it contains.

Design an operating model for data governance in a company with several business units.

What they check: the structure that makes it last.

Short answer: "A small central team that sets the method, the tools and the reporting; owners and stewards in each unit for their domains; a data council where owners decide what crosses units. Central where consistency matters, such as definitions shared by all, and local where the units know best."

The full answer: Write down who decides what; an operating model is mostly a list of decision rights.

A weak answer: "The central team does all governance." Follow-up: "One unit refuses to name owners." Say what you do.

How do you run a data council that decides things?

What they check: governance that ends arguments.

Short answer: "The council is the owners of the main domains, meeting monthly with a short agenda of decisions prepared in advance: a definition, a quality target, an access question, an issue that crosses units. Each decision is written down and published; anything a single owner can decide stays out."

The full answer: If the council only receives updates, it will stop meeting; give it decisions.

A weak answer: "We present our progress to the council." Follow-up: "Half the owners send deputies." Say what you change.

What is a critical data element, and how do you pick them?

What they check: focusing effort where it matters.

Short answer: "A critical data element is a field whose errors have serious consequences, such as a figure in the board report, a field a regulator checks or one that drives payments. I pick them with the owners from the key reports and processes, then give each a definition, an owner, quality rules and lineage first."

The full answer: Keep the list short enough to govern well; it can grow when the first set is under control.

A weak answer: "All data is critical." Follow-up: "Every unit wants its fields on the list." Say how you decide.

A report shows the board a wrong figure. How do you handle the data issue?

What they check: issue management under pressure.

Short answer: "I log it as a data issue with an owner, find the root cause through the lineage, agree the correction with the owner of the report and of the data, tell the board what was wrong and what changed, and add a quality rule so the same error is caught before the next report."

The full answer: The board forgives one error explained well; it does not forgive the same error twice.

A weak answer: "The analyst corrects the report." Follow-up: "The cause is in a system no team wants to change." Say who decides.

How do retention and deletion fit into data governance, and where does the DPO come in?

What they check: the boundary with data protection.

Short answer: "Retention says how long each kind of data is kept and when it is deleted. Governance makes sure each data set has a retention period, an owner who applies it and a record in the catalogue. For personal data, the data protection officer advises on what the law allows, under rules such as the GDPR in the EU; governance makes the decision work in the systems."

The full answer: Keeping data forever is a cost and a risk; deleting it on time is part of good governance.

A weak answer: "We keep everything in case we need it." Follow-up: "Legal and the analysts want opposite retention periods." Say who decides.

Adoption and AI

How do you get adoption, and how do you measure the maturity of data governance?

What they check: results, not documents.

Short answer: "Adoption comes from solving problems people feel: start with the domain where bad data costs most, show the fix, and let those owners tell others. I measure it with the use of the catalogue, quality scores of critical data, issues closed and decisions taken, and I use a maturity model, such as one based on DAMA-DMBOK or DCAM, once a year to show where to go next."

The full answer: A maturity score impresses once; quality figures that keep improving keep the budget.

A weak answer: "We counted the policies published." Follow-up: "After a year, the maturity score did not move." Say what you look at.

What does data governance need to cover when the company uses AI?

What they check: today's new demands on data.

Short answer: "AI needs the same basics with higher stakes: known sources, owners who allow the use, quality and bias checks on training data, lineage from the data to the model, and records of what each model was trained on. In the EU, the AI Act asks for data governance of the data used by high-risk AI systems."

The full answer: Many AI projects stall on data nobody owns; governance that is already working becomes a head start.

A weak answer: "AI is the data scientists' job." Follow-up: "A model was trained on customer data without the owner knowing." Say what you change.

Senior rounds

As a senior data governance manager, how would you grow a data steward into a governance role?

What they check: growing the network of people governance depends on.

Short answer: "I give a strong data steward a cross-unit problem, such as one definition two units disagree on, with my coaching, then a domain of their own and a seat at the council preparation. I teach the parts that are new: running meetings with senior owners, writing short policies and reporting results."

The full answer: A senior data governance manager or head of data governance is judged by the network of owners and stewards they leave behind.

A weak answer: "Stewards learn on their own." Follow-up: "A unit has no steward and no budget for one." Say what you do.

Your CV for these questions

Each job on your CV invites the questions above: which domain, which owners, which rules and what changed. Give those stories a line each, written so a head of data wants to hear more:

Put your CV beside a data governance manager job you want with the free score; the area it finds thin is the one to rehearse first, with a domain you can describe from owners to quality scores.