A research question turns a stakeholder problem into a clear, answerable target for analysis. It should define the scope, settle ambiguous terminology, identify the relevant data, and remain documented for transparency and reproducibility. section 1.2.1
Develop a workable question by:
- Clarifying ambiguous terms and deciding which dates, events, entities, or categories count. section 1.2.1
- Defining boundaries, such as whether an analysis of cars also includes vans. section 1.2.1
- Checking the survey, data dictionary, columns, missing data, and outliers before committing to an analysis. section 6.3
- Starting with a minimum viable answer, then revising the question when the available data cannot support the original version. section 1.3 section 8.4
- Using cautious language such as “the data suggests” when evidence does not robustly support a stronger claim. section 6.3
Developer attitudes toward AI tools
A vague concern becomes a research question by specifying the outcomes and comparisons: investigate whether AI use differs by experience level, and whether experience, job role, or tool usage relates to opinions and trust. The analysis then selects relevant survey columns, inspects data quality, tests appropriate relationships, and reports only conclusions supported by the data. section 6.3 section 6.4
Table 1.1 A summary of the eight projects in the book
|
Project
|
Data analysis skill(s)
|
Chapter(s)
|
|---|---|---|
|
Analyzing customer retail spend in different geographic areas
|
Extracting structured data from free text
|
2
|
|
Extracting unique customer records from e-commerce transactions
|
Data modeling
|
3
|
|
Defining and finding best-performing products on an e-commerce store
|
Defining metrics
|
4
|
|
Analyzing the effects of the coronavirus pandemic on the film industry
|
Extracting data from PDFs
|
5
|
|
Investigating developer attitudes toward AI tools from a survey
|
Handling categorical data
|
6, 7
|
|
Identifying potential improvements to cycling infrastructure
|
Working with time series data
|
8, 9
|
|
Building a proof-of-concept application to explore the Welsh property market
|
Rapid prototyping
|
10, 11
|
|
Continuing another analyst’s work and creating customer segments using mobile phone activity
|
Iterating on someone else’s work Customer segmentation, clustering
|
12, 13
|
Compare research questions across realistic projects, from customer records and product performance to film trends, cycling infrastructure, and property markets. section 1.3
