What a fractional head of data actually does in the first 30 days

"Fractional head of data" is a growing category and a vague one. Businesses know they need senior data capability without a permanent hire; far fewer know what they should expect in return.

Here is what the first month should contain — and, as importantly, what it should not.

What it is not

It is not a month of building dashboards. If someone starts by building you a dashboard, they have skipped the part that determines whether the dashboard is worth anything.

It is not a tooling review either. The stack is rarely the binding constraint, and replacing tools before understanding the process usually relocates the problem.

Week one: find out which numbers the business runs on

Every business steers by a handful of numbers. Rarely more than ten. They appear in the board pack, the weekly sales meeting, the marketing review.

The first job is to write them down and, for each one, answer four questions:

• Who produces it, and how long does it take them?
• Which system is it pulled from?
• What is the definition — the actual one, not the label?
• Who would notice if it were wrong?

That last question is the most revealing. A surprising number of regularly-produced numbers have no one who would notice. Those are candidates for deletion, and deleting them is real work delivered.

By the end of week one you should have a list of the numbers that matter, and a shortlist of the ones nobody can currently defend.

Week two: trace two of them end to end

Pick the two most important numbers and follow each backwards through every system to its origin.

This is unglamorous and it is where the actual findings come from. Typical discoveries:

• A metric filtered by a rule one person added years ago and never documented
• Two systems using the same word for different objects
• A manual step nobody mentioned in week one, because it has become invisible
• A field that stopped populating after a form change, silently

This is also your test of whether the engagement is working. By the end of week two, a fractional data lead should be telling you at least one thing about your own numbers that you did not know. If they are still asking for access and orienting themselves, that is a warning sign.

Week three: write the definitions down

One page per core metric: object counted, filters applied, timestamp used, timezone and calendar.

The document is not the deliverable. The agreement is. Getting marketing, sales and finance to sign up to shared definitions is the highest-leverage thing available, and it is political work as much as technical work — which is a large part of why it needs someone senior rather than an analyst.

Expect this to be uncomfortable. Definitions are where teams discover they have been reporting different things to the same board for years.

Week four: sequence the work and fix one thing

Two outputs.

A prioritised plan — what to fix, in what order, with an honest view of effort. Ordered by what unblocks the most, not by what is easiest. Usually identifier capture comes early, because everything downstream depends on it.

One thing actually fixed. A month of analysis with nothing shipped sets a bad precedent. Pick something small and visible — a report that was manual and is now automatic, a broken field now populating, a reconciliation that used to take a morning.

What you should hold at the end of the month

• A list of the numbers the business runs on, and who owns each
• Definitions agreed in writing by the people who use them
• Two metrics traced end to end, with the breaks named
• A prioritised plan with effort estimates
• One visible fix in production

If a month produces only a strategy document, you bought a document.

How the ongoing arrangement should work

After the first month the shape usually changes: a smaller recurring commitment for oversight, definitions governance and reviewing the work of whoever implements — internal team, agency or contractor.

The reason to keep it fractional is that the scarce skill is judgement about what to measure and why, and that is not a full-time need in most businesses. The implementation often is. Those are different jobs and conflating them is how businesses end up with an expensive senior hire writing SQL.

The honest caveat

This model does not suit everyone. If your data function is genuinely large, or the work is continuous rather than periodic, you want a permanent hire and you should make one.

Fractional works best in a specific window: the business has outgrown spreadsheets and disconnected tools, the reporting burden is real, but the volume of work does not justify a full-time senior salary. That window is where the arrangement earns its keep, and it is worth being honest with yourself about whether you are still in it.