Fractional data science jobs · hire a fractional head of data science, or hire one for good

Fractional data science jobs: what the work is

The government’s Digital and Data capability framework describes data science as a broad and fast-moving field spanning maths, statistics, software engineering and communications. In practice it means using data to predict, classify or optimise something the business cares about: demand, pricing, churn, fraud, risk, or which customer to call first. Fractional data science jobs put someone senior on that work for part of the week.

Two kinds of seat

The leadership seat sets direction: which problems are worth modelling, which tools and methods the team uses, how models are tested, released and watched, and who is hired. The individual contributor seat builds: cleans the data, writes the code, trains and validates the models and explains the results. A small company’s first senior data hire often does both. Say which one you want to hire.

When a business brings one in

The usual moments: a first model the business wants to rely on, a team of analysts or junior data scientists with nobody senior to review their work, a model already in production that nobody trusts or can explain, or an investor or buyer asking how the numbers are produced. None of these needs a full-time executive on day one.

Fractional data scientist: an individual contributor, a lead, or both?

Many roles advertised as fractional data scientist are hands-on: a senior person who builds the models, writes the code and presents the findings, a few days a week. That is an individual contributor seat, whatever the day rate. Others are the lead seat under a smaller title, and in a start-up the first data scientist is often both.

The government framework draws the line usefully. Its data scientist level works independently and develops complex solutions; its principal data scientist is described as a leader of data science, quite often with responsibility for managing and developing teams; its lead and head levels set direction and build capability (DDaT framework). Work out which level your problem needs before you brief.

The individual contributor seat also raises sharper employment questions. Someone embedded in your team, whose manager decides what they work on and how, looks more like an employee on GOV.UK’s indicators. Where they work through their own company, the off-payroll rules apply: a medium or large client makes the determination, and status turns on how the engagement runs. We recruit the leadership seat, and the board below lists leadership roles.

When models make decisions about people: what the head of data science answers for

A pricing, credit, hiring or fraud model that scores people brings data protection with it. The ICO’s guidance on AI and data protection covers fairness and statistical accuracy as well as lawfulness and transparency; the ICO notes it is under review after the Data (Use and Access) Act. That Act opens up the lawful bases for significant automated decisions about people, provided appropriate safeguards are in place, but not for special category data.

The government’s Data and AI Ethics Framework was written for the public sector, but its principles for the responsible development, procurement and use of data and AI are a sound checklist for any team. A head of data science should be able to show where fairness and accuracy are tested in the release process, and who signs a model off.