Project Scoping, Data Acquisition, and Ethics Review

Project Scoping, Data Acquisition, and Ethics Review

Every successful data science project begins long before the first line of code is written. The decisions you make in the first few days—what question you are asking, where your data comes from, and how you will protect the people whose information is in that data—shape every result that follows. This lab walks you through the three foundational tasks: scoping your project, acquiring appropriate datasets, and completing an ethics and data management plan.

Defining a Measurable Analytical Question

A project scope document is not a vague statement of interest. It is a precise agreement between you and your stakeholders about what you will measure, for whom, and how you will know when you have succeeded. A strong one-page project brief includes four elements:

• Business or civic question: State the question in plain language. Prefer questions with a measurable outcome. For example, "Which Chicago zip codes show the highest rates of emergency department visits for asthma?" is specific and testable. "How does health work?" is not.
• Target audience: Identify who will act on your findings—city planners, a nonprofit board, a retail operations team. Audience shapes which metrics matter and how you will communicate results.
• Success criteria: Define what a good answer looks like before you look at the data. A useful criterion might be: "A map visualization that shows statistically meaningful variation across zip codes, with all data sourced from public health agencies dated within the last three years."
• Data sources: Name at least two candidate sources at the scoping stage. You will refine these during acquisition.

This process mirrors the CRISP-DM (Cross-Industry Standard Process for Data Mining) framework, which begins with Business Understanding before any data is touched. IBM's Data Science Professional curriculum emphasizes CRISP-DM as the standard lifecycle model for applied projects.

Locating and Evaluating Datasets

Public repositories give you a wide range of real-world data without licensing barriers. Key sources include data.gov (federal datasets across agencies), the U.S. Census Bureau (demographic and economic data), the CDC (public health statistics), local and municipal open-data portals, and Kaggle (curated datasets often used for practice projects).

Not every dataset you find is appropriate. Evaluate each candidate against a structured checklist:

| Criterion | Questions to ask |

|---|---|

| Provenance | Who collected this data? Is the source authoritative? |

| License | Are you permitted to download, use, and share results? |

| Completeness | What percentage of key fields are populated? Is missingness random or systematic? |

| Temporal coverage | Does the date range match your question? Is it current enough? |

| PII risk | Does the dataset contain names, addresses, social security numbers, dates of birth, or other personally identifiable information? |

CompTIA Data+ Domain 1 (Data Concepts and Environments) covers data sourcing, governance, and the importance of understanding data quality before use—the same principles you apply in this checklist.

Document every dataset you download with a full citation: source name, dataset title, publication or update date, URL, and the date you accessed it. This protects you if the source changes later and demonstrates professional practice.

Data Management and Ethics Planning

Working with real-world data carries responsibility. Even public datasets can contain sensitive attributes—race, income, health status, geographic location fine enough to identify individuals—that require careful handling. This is not a compliance formality; it is the ethical foundation of trustworthy analytics.

Your data management and ethics plan should address five areas:

1. Sensitive attributes: List any variables that could identify or harm individuals or groups if mishandled. Fields such as age, sex, race/ethnicity, ZIP code, and diagnosis codes can be sensitive even when names are removed, because combinations of attributes may re-identify people.

2. De-identification steps: Describe what you will do to reduce PII risk. Common techniques include removing direct identifiers (names, ID numbers), aggregating fine-grained geographic data to a coarser level, and suppressing cells with very small counts.

3. Storage and access controls: State where data will be stored (e.g., a password-protected folder on your institution's network or an encrypted cloud folder), who can access it, and how long you will retain it.

4. Bias and fairness considerations: Note whether the dataset over- or under-represents any population and how that might affect your conclusions.

5. Data-use acknowledgment: Sign a statement confirming you have read the dataset's license, that you will use the data only for your stated project purpose, and that you will not attempt to re-identify individuals.

This structure is modeled on Institutional Review Board (IRB) documentation used in academic and clinical research. While your capstone project may not require formal IRB approval, adopting the same disciplined thinking builds habits that every professional data scientist needs.

Safety Considerations for Data Work

Data science does not involve physical machinery, but it does involve risks that practitioners must manage responsibly:

• Data security: Never store datasets containing PII on unencrypted personal devices or unsecured cloud services. Treat any sensitive data with the same care you would expect if your own information were in the file.
• Phishing and unauthorized access: Only download data from verified, official URLs. Malicious files can be disguised as CSV or Excel downloads on unofficial mirrors.
• Intellectual property: Confirm licensing before using any dataset. Creative Commons licenses, open government licenses, and proprietary licenses have different terms. Violating a data license is both an ethical and a legal risk.
• Bias as a safety issue: Biased data used to make decisions about housing, employment, or healthcare can cause real harm to real people. Identifying and documenting bias in your ethics plan is a harm-reduction step.

Advancing to the Next Phase

You may not proceed to exploratory data analysis until your instructor has approved both your project brief and your data management and ethics plan using the 10-point rubric, and you have at least two datasets downloaded with full citation documentation. The PII risk section is reviewed independently and must pass on its own. This gate exists to protect you, your data subjects, and the integrity of your final project.