Data Scientists

Job purpose

To transform raw structured and unstructured data into meaningful information using statistical software, data-oriented programming languages and visualisation tools, and to interpret and report the findings, so that business problems and management objectives are addressed by evidence rather than assumption.

Skills

• Applied statistical knowledge across data mining, data modelling, natural language processing and machine learning: feature selection algorithms for predicting outcomes such as sales, attrition and healthcare use; sampling and survey design; and the metrics used to compare models, such as loss functions and explained variance. Writes new functions and applications in programming languages to conduct analyses, and works in specialised visualisation software. No distribution of qualifications actually held by incumbents has been surveyed for this occupation, so no modal category can be given; the preparation expected is degree-level study with several years of related experience and on-the-job training.
• Presents modelling and analysis results, orally and in writing, to management and other end users, and recommends data-driven solutions to key stakeholders. Statistical method and its limitations must be made intelligible to readers who do not share the specialism, and the problems analysis can properly address identified with managers. No supervisory or teaching duties are recorded in the source.
• Analytical work of substance and open character: identifying relationships, trends and factors that could affect results; testing, validating and reformulating models so that they predict the outcome of interest accurately; and comparing competing models. Method is not prescribed — the sampling approach, the features selected and the model chosen all have to be reasoned and defended.
• Sequences an analysis from problem definition through data cleaning, modelling and validation to visualisation and reporting, and keeps abreast of emerging analytic methods through the research literature. Priorities follow the business problems and management objectives identified; no data on time pressure or deadline structure has been surveyed for this occupation.
• Computer-based analytical work using statistical and visualisation software. Posture, dexterity and handling demands have not been surveyed for this occupation and none is asserted.

Responsibilities

• No line management or supervisory responsibility appears in the source. Influence rests on the data-driven solutions recommended to key stakeholders.
• No budget is held and no spend committed. Analysis identifies solutions to business problems such as budgeting, staffing and marketing decisions, which are taken by others.
• Ordinary care of a workstation and of statistical and visualisation software.
• Works on large structured and unstructured datasets, cleaning and manipulating raw data and building the models and the dynamic reports, graphs, charts and other visualisations that convey the results. Accountable for the accuracy of the data as processed and for models that predict the outcome of interest reliably. The source does not address the confidentiality of the datasets handled.

Effort required

• Sustained analytical concentration on the construction, testing and reformulation of models, on the manipulation of large datasets where a fault in cleaning carries through to the result, and on comparison against performance metrics. Judgement is exercised under incomplete information about which method fits the question.
• Accountability is for the technical defensibility of models and of the recommendations put to stakeholders. The source records no exposure to distress, conflict or hostility.
• Computer-based analytical work implies seated work at a workstation; no posture, exertion or lifting data has been surveyed for this occupation, so no physical effort demand is evidenced.

Working conditions

• The tasks describe office-based analytical and reporting work. No work context survey exists for this occupation, so no exposure to hazards, weather, noise, travel, lone working or difficult people is evidenced.
• No data on working hours, schedules, week length or workload peaks has been collected for this occupation; shift, rota, on-call and out-of-hours working are not evidenced and are not asserted. What the tasks show is work directed by identified business problems and management objectives, with the choice of method left to the postholder.