How to Hire a Head of Data for the Mandate You Have

If you need to hire a Head of Data, start by deciding which problem the person must own. A leader hired to repair a fragile data platform is different from one hired to make analytics useful, and both differ from an executive expected to build machine-learning products.

The practical answer is to write a one-year mandate, choose the profile that matches it, and assess every candidate against the same evidence. The title matters less than the decisions, team and outcomes attached to it. This guide shows how to make those choices before a polished candidate makes them for you.

Define the Head of Data mandate before the job description

A generic specification usually combines data engineering, business intelligence, governance, data science and AI into one impossible role. Replace it with a short mandate. Name the business problem, the current state of the data estate, the authority the leader will hold and what must be observably better after one year.

The UK Government's Chief Data Officer capability framework is a useful reference even outside government. Updated in 2025, it frames the senior data role around strategy, organisational alignment, architecture, governance, risk and capability building. That breadth is a warning: your mandate must rank priorities rather than copy every responsibility.

• State the most important outcome: trusted reporting, a dependable platform, data products, governance, or commercial use of data.
• List the decisions the Head of Data owns and the decisions shared with the CTO, product leader, security lead and business executives.
• Describe the inherited team, critical vacancies, major vendors and technical debt honestly.
• Set review points for the first quarter, mid-year and year end, using outcomes rather than activity counts.

Treat this mandate as the source for the scorecard, interview questions and references. The same discipline applies to adjacent roles, but the dividing lines differ; our guides to hiring a CTO and hiring a VP of Engineering explain those boundaries.

Choose between platform, analytics and ML leadership

Most Head of Data searches lean toward one of three profiles. A platform leader builds reliable ingestion, storage, transformation, access and observability. An analytics leader turns messy business questions into shared metrics, decision routines and self-service reporting. An ML leader creates the path from experimentation to deployed models, including evaluation, monitoring and responsible use.

All three need enough range to lead the whole function, but depth should follow the mandate. If outages, undocumented pipelines and slow access block every team, prioritise the platform operator. If executives argue over numbers and analysts operate as a ticket desk, prioritise the analytics translator. If the product strategy depends on models in production, prioritise the leader who has shipped and governed ML systems, not simply managed research.

Do not use team size as the only seniority test. Ask what complexity the candidate personally handled: regulated data, several business domains, real-time systems, international teams, or model risk. A leader from a larger company may have owned a narrow slice, while a candidate from a smaller firm may have built the operating model end to end.

Use a simple decision rule: hire for the hardest problem of the next year, then require credible fluency in the other two areas. If your supposed priority changes whenever a new stakeholder enters the room, the company is not ready to open the search.

Build a scorecard for how to hire a Head of Data

A good scorecard separates evidence from charisma. Give every competency a clear definition, observable signals and a weighting tied to the mandate. Interviewers should record evidence independently before the debrief. This prevents a famous employer, a fashionable stack or an energetic presentation from substituting for relevant achievement.

1. Business judgement. Can the candidate connect data investment to revenue, cost, risk, customer experience or operating speed? Ask for a decision they changed and how they knew it mattered.
2. Technical judgement. Can they make sound trade-offs across architecture, build-versus-buy choices, quality and reliability without becoming the team's principal implementer?
3. Operating leadership. Can they set ownership, planning, incident and quality routines that survive beyond personal intervention?
4. People leadership. Can they hire complementary specialists, develop managers and address weak performance directly?
5. Influence and governance. Can they win commitment from executives and domain owners when the data team has responsibility but limited formal control?

Governance deserves explicit weight. The UK Government's data ownership model distinguishes accountable data owners from operational stewards and technical custodians. A strong candidate can translate that principle into a workable model for your company rather than trying to centralise every decision inside the data team.

Run an interview loop that tests real leadership

Use the same core loop for every finalist. Begin with a mandate conversation so the candidate understands the problem. Follow with structured interviews on business impact, technical judgement, operating leadership and people management. Add a case based on sanitised information, then complete references that test the same scorecard.

For the case, give candidates a fictional company with inconsistent metrics, brittle pipelines and pressure to launch an AI feature. Ask what they would diagnose first, what they would defer, who must own each decision and what they would expect to change over two quarters. Do not reward the longest architecture diagram. Reward prioritisation, questions, trade-offs and the ability to explain risk in plain language.

AI governance is now part of the leadership test when models are in scope. The NIST AI Risk Management Framework playbook connects AI controls to organisational governance, data quality, model testing and documented responsibilities. The EU AI Act's Article 10 on data and data governance also sets detailed expectations for data practices around high-risk AI systems. You are not testing legal recall; you are testing whether the candidate knows when to involve legal, privacy, security and product owners.

Probe past behaviour with specifics. Who disagreed? What decision did the candidate own? What failed? Which metric moved? What did a direct report say about the change? Follow vague answers until the boundaries of personal contribution are clear.

Source beyond people already carrying the title

The right candidate may currently be a Director of Data, VP Analytics, Head of Data Platform, Chief Data Officer or senior engineering leader with a data mandate. Search by scope and evidence, not title alone. Map companies with a similar data problem, business model, regulatory burden or stage, then identify the person who actually led the relevant change.

Calibrate compensation, location and reporting line before outreach. Senior candidates will ask whether the role reports to the CEO, CTO, product leader or another executive, and each answer changes the perceived authority. They will also test whether budget and headcount match the promised mandate. If those answers are unresolved, outreach creates interest that the process cannot sustain.

A focused market map also reduces familiar-name bias. It gives the panel a view of different backgrounds and makes trade-offs visible early. If you need help reaching leaders who are not active applicants, our executive search service covers market mapping, direct approach, assessment and reference work.

Red flags to investigate before the final decision

• They prescribe a new platform before asking how the business uses data or what is failing.
• They talk about AI ambition but cannot explain data quality, evaluation, monitoring or ownership.
• They claim broad outcomes but cannot separate their contribution from that of engineering, product or business teams.
• They treat governance as paperwork owned by compliance rather than a decision and accountability system.
• They need executive sponsorship for change but have no example of earning it.

No single concern should decide the hire. Look for a pattern across interviews, the case and references. References should include a former manager, a peer outside data and someone who reported to the candidate. Ask each person about the same mandate risks rather than requesting a general character endorsement.

Frequently asked questions

Should the Head of Data report to the CEO or CTO?

The reporting line should follow the mandate. A platform-heavy role often fits under the CTO, while an enterprise role spanning commercial decisions and governance may need direct executive access. More important than the box on the chart is clear authority across technology and business domains.

Do we need a Head of Data or a Chief Data Officer?

Choose scope before title. A Chief Data Officer normally signals enterprise-wide strategy, governance and executive accountability; Head of Data can describe anything from a hands-on team lead to the same broad mandate. Use the title that your market understands, then make the authority unambiguous.

How technical should a Head of Data be?

Technical depth must match the main risk. The leader should challenge architecture, quality and ML decisions and earn the team's trust, but need not be the best coder. For a small team, expect more direct design work; for a larger function, judgement, organisation design and influence become more important.

Hiring a Head of Data is easier when the company is honest about the problem and disciplined about evidence. Define the mandate, weight the scorecard, run one comparable process and verify claims through references. If the search requires confidential access to senior data leaders across markets, ArgusRecruit can help you build and assess the field.