When data decisions are piling up faster than anyone senior can make them, but you cannot yet justify, or attract, a full-time data executive. It fits best just before an expensive choice, such as a platform, a first data hire, or an AI plan, when getting it wrong costs more than a few months of senior time.
The signs
- Nobody owns the numbers. Metrics come from whoever built the dashboard, and leadership meetings start with an argument about which figure is right.
- Your data people report to someone who is not a data leader. A strong engineer or analyst reporting to the CTO or CFO often lacks someone who can set direction, review their work, and represent data in leadership decisions.
- A big decision is due. A warehouse migration, a platform contract, a first data hire, or a board question about AI. Each is hard to reverse, and each is easy to make on a vendor's advice alone.
- Regulation is arriving. An audit, a new market, a healthcare or payments customer, or a young player base. Governance becomes a requirement rather than a nice-to-have.
- The warehouse bill is growing faster than the business, and nobody can explain why.
What a fractional head of data does
The same things a full-time one does, at a smaller scope. They own the data roadmap and the budget line. They set standards for how data is modeled, tested, and governed. They help hire and grow the team, including writing the first job descriptions and interviewing. They sit in leadership meetings as the person accountable for the numbers. And they make the platform decisions, with a written rationale the company keeps.
What they should not do is become a very expensive individual contributor. If most of their time goes into building pipelines, you need an engineer, not a leader.
The first 90 days
| Weeks | Focus | You should see |
|---|---|---|
| 1 to 2 | Listen and read the platform: people, pipelines, costs, access | A written current-state read, including the risks nobody had named |
| 3 to 6 | Agree on the handful of metrics that matter and who owns each one | One definition per headline metric, signed off by finance |
| 7 to 10 | Fix the most expensive or riskiest problem first | A visible result: a lower bill, a closed access gap, or a trusted dashboard |
| 11 to 13 | Roadmap, budget, and hiring plan for the next two quarters | A plan leadership can approve, with the first role open if one is needed |
How it compares
| Fractional head of data | Full-time hire | Project consultant | |
|---|---|---|---|
| Accountable for outcomes | Yes, for the length of the engagement | Yes | For the deliverable only |
| Time to start | Weeks | Often months of recruiting | Weeks |
| Sits in leadership meetings | Yes | Yes | Rarely |
| Builds and manages the team | Hires and sets it up | Yes | No |
| Best when | The role is not yet clear or not yet full time | The role is clear and the budget is there | The problem is well defined and bounded |
How to know it is working
By the end of the first quarter, leadership meetings should argue about what to do with the numbers, not about which number is right. The team should know what it is building and why. The warehouse bill should have an owner and a trend line. And there should be a written record of every major decision, so that none of it depends on one person remembering.
How it ends
A good fractional engagement is designed to end. Most end with a full-time hire, either a head of data the fractional leader helped recruit, or a senior member of the team grown into the role. The handoff includes the roadmap, the decision record, and a period of overlap. The other good ending is that the company decides it does not need a full-time data executive yet and moves to lighter advisory support. Either way, the test is that the company is stronger without them than it was before they arrived.
Sid SrivastavaFounder of Millwright Data. He has run data platforms in semiconductor manufacturing, digital health, hospitality, and AAA games, on Snowflake, Databricks, and AWS.
