Module 1: The Data Science Landscape
What Data Scientists Actually Do
Job adverts for data scientists often read like wish lists: statistics, programming, machine learning, cloud platforms, business sense and storytelling, all in one person. That can make the field look out of reach. In reality, the work follows a recognisable pattern, and most of it is careful, practical problem solving rather than advanced mathematics. Understanding what the job really involves helps you focus your learning on the skills that make the biggest difference.
The core job: turning questions into evidence
A data scientist helps an organisation make better decisions by using data as evidence. The work usually starts with a question from someone who is not technical: a sales manager, a finance officer, a product owner. Typical questions look like this:
- Which of our customers are likely to stop buying from us in the next three months?
- Why did loan repayments fall in the Mombasa branch last quarter?
- How much stock of each item should each shop hold next month?
- Did the new SMS reminder campaign actually change customer behaviour?
Your job is to translate a question like this into something data can answer, find and prepare the right data, analyse it honestly, and explain the result in a way that leads to action. Notice that only one part of that sentence is "analyse". The rest is understanding, preparation and communication.
Where the time actually goes
Practitioners consistently report that a large share of their time goes on finding, understanding and cleaning data, rather than on building models. You will see different percentages quoted in different places, so do not treat any single figure as a law. The principle is reliable, though: real data is messy, spread across systems and full of surprises, and getting it into a trustworthy state is a major part of the job.
A realistic week for a junior data scientist at a Nairobi retail chain might include:
- Meeting the operations manager to clarify what "stock-out" means in their reporting.
- Writing SQL queries to pull six months of sales from the point-of-sale database.
- Discovering that two branches record dates in different formats, and fixing it.
- Exploring the data with summary tables and charts to spot patterns and errors.
- Building a simple forecast and checking how wrong it would have been on past months.
- Preparing three slides and a short written summary for management.
Related roles and how they differ
Data science overlaps with several other roles, and job titles are used loosely. A rough guide:
| Role | Main focus | Typical outputs |
|---|---|---|
| Data analyst | Describing what happened and why | Reports, dashboards, ad-hoc analyses |
| Data scientist | Explaining and predicting, often with statistics and machine learning | Models, experiments, deeper investigations |
| Data engineer | Building reliable pipelines that move and store data | Databases, data pipelines, data warehouses |
| Machine learning engineer | Putting models into production systems | Deployed models, monitoring, APIs |
In small organisations one person may do all four. In large ones the roles are separate. Either way, the foundations in this course apply to all of them.
Worked example: from vague request to clear question
Imagine a savings and credit cooperative in Nakuru. The manager says: "We want to use data to improve our loans." That is not yet a question data can answer. A good data scientist asks follow-up questions:
- What decision will change? The manager explains that loan officers decide whether to approve applications and for how much.
- What does "improve" mean? Fewer loans falling more than 30 days behind, without turning away too many good members.
- What data exists? Five years of loan records, member savings history and repayment dates, held in the core banking system.
- How will success be measured? The share of new loans that fall into arrears, compared with the previous year.
The refined question becomes: "Using past loan and savings records, which application details best indicate a higher risk of arrears, and how could loan officers use that information?" This is specific, measurable and tied to a real decision.
The skills that matter most
You will learn tools in this course, but these habits matter more than any single tool:
- Curiosity and scepticism. Asking "is this number right?" before "what does it mean?"
- Clear thinking about cause and effect. Knowing when data shows a link and when it shows a cause.
- Reproducibility. Writing down steps, usually in code, so someone else can repeat your work.
- Communication. Explaining findings to people who will never read your code.
Common mistakes to avoid
- Jumping into modelling before you understand the business question.
- Assuming the data is correct because it came from an official system.
- Measuring success by how sophisticated the method is rather than whether it helps a decision.
- Producing results nobody asked for and nobody can use.
Keep this picture in mind throughout the course: data science is a decision-support discipline. Every lesson that follows is a step in turning raw data into evidence that someone can act on.
Exercise
Pick an organisation you know well, such as your employer, a family business or a local chama. Write down one vague goal someone there might have, then turn it into a specific question using the four follow-up questions from the worked example: what decision will change, what improvement means, what data exists and how success will be measured. Keep your final question to one or two sentences.
Key points
- Data science uses data as evidence to support real decisions.
- Much of the work is understanding, finding and cleaning data, not modelling.
- A good question names the decision, the measure of success and the data available.
- Curiosity, scepticism, reproducibility and communication matter more than any one tool.