2.8: Detailed Figure Captions - Chapter 2

Data Types, Data Sources, and Data Quality — Figure Specifications

Overview

• Chapter: 2 — Data Types, Data Sources, and Data Quality
• Total Figures Proposed: 8
• Figure Types Used: Comparison Table, Conceptual Illustration, Architecture Diagram, Flowchart, Infographic, Network Diagram

---

Figure 2.1.1

Title: Structured, Semi-Structured, and Unstructured Data Spectrum 

Detailed Caption: A horizontal "Rigid Schema → No Schema" gradient bar runs across the top, beneath which three columns sit side by side under colored header pills — navy "Structured Data," teal "Semi-Structured Data," and warm-gold "Unstructured Data." The Structured panel lists examples (SQL Databases, Spreadsheets, Sensor Readings) above traits "Fixed rows & columns, SQL-queryable" and a footer "~10–20% of enterprise data." The Semi-Structured panel pairs JSON Documents, XML Files, and Email Messages with "Tags/keys provide hierarchy, Flexible schema, NoSQL" and "Growing rapidly with APIs & web." The Unstructured panel pairs Text Documents, Images & Photos, and Video & Audio with "No predefined model, Requires NLP/CV to analyze" and "~80–90% of enterprise data." By aligning examples, technical traits, and enterprise share for all three categories on a single canvas, the figure makes vivid the chapter's quiet but consequential point: most of an organization's data sits in formats that don't fit a SQL table, which is why modern analytics stacks need much more than a relational database. 

Figure 2.1.2

Title: Levels of Measurement Hierarchy 

Detailed Caption: Four ascending stair-step tiles climb from lower-left to upper-right alongside a vertical "Increasing Quantitative Power" arrow on the left margin. The bottom step, light-gray "NOMINAL," lists "Categories only" with examples "Blood type, Gender" and operations "Mode, Frequency counts." The next step, gray "ORDINAL," adds "+ Rank order" with "Survey ratings, Class standing" and "+ Median, Percentiles." Step three, teal "INTERVAL," adds "+ Equal intervals" with "Temperature °C/°F, Calendar year" and "+ Mean, Std deviation." The top step, warm-gold "RATIO," adds "+ True zero" with "Weight, Revenue, Distance" and "+ Ratios, All arithmetic." Each higher tier inherits the operations of those below — the "+" notation is the figure's pedagogical hinge, signaling that ratio data can use everything nominal/ordinal/interval can plus more. By rendering measurement levels as a literal staircase, the figure helps students remember that moving up the hierarchy unlocks more statistics rather than substituting one set for another, and that the operations available on a variable are constrained by its measurement level — not by its data type. 

Figure 2.2.1

Title: Internal Data Sources Within an Organization 

Detailed Caption: A central navy "ORGANIZATION (Internal Data Sources)" tile anchors the diagram, with six teal-bordered satellite cards radiating outward in a hub-and-spoke pattern, connected by teal lines. Each card pairs a warm-gold circular icon with a category name and a one-line example list: Operational Data (Transactions, inventory, logs); Customer Data (CRM, feedback, behavior); Employee Data (HR, payroll, performance); Financial Data (Revenue, budgets, P&L); Marketing Data (Campaigns, web traffic, clicks); and Production Data (Machine performance, quality). The radial layout makes the breadth of internal data tangible — six categories, generated by entirely different parts of the business, all owned by the same organization — and reinforces the chapter's argument that even before reaching for external sources, most companies sit on a much richer internal data footprint than they actively analyze, with each domain typically managed by a different team and stored in a different system. 

Figure 2.2.2

Title: External Data Sources Available to Organizations 

Detailed Caption: A navy "ORGANIZATION" building icon sits on the left, with a large dashed-bordered "External Data Environment" rectangle on the right enclosing six teal-bordered category cards arranged in two rows of three. The top row holds Public/Government (Census, economic indicators, SEC), Social Media Data (Posts, sentiment, engagement), and Web Data (Scraping) (Prices, reviews, news). The bottom row holds Sensor/IoT Data (Weather, satellites, traffic), Partner/Third-Party (Supplier data, credit scores), and Open Data (Research datasets, open portals). Six teal arrows fan inward from the cards back toward the Organization, signalling that all external sources are inputs that flow into the company. By visualizing external data as a separate environment that the organization deliberately reaches into, the figure complements its internal-source counterpart and makes the chapter's point clear: serious analytics rarely runs only on first-party data — augmenting with external sources is what lets a company benchmark against the market, react to events outside its walls, and quantify factors it doesn't directly produce. 

Figure 2.3.1

Title: The Seven Dimensions of Data Quality 

Detailed Caption: A central navy circle labeled "DATA QUALITY" anchors the figure, with seven dimension tiles fanning outward as petals connected by thin colored spokes. Each tile carries a bold label and a one-line definition: Accuracy (Data reflects the true state), Completeness (All necessary data is present), Consistency (Uniform across all systems), Timeliness (Available when needed), Validity (Conforms to formats & rules), Uniqueness (No unintended duplicates), and Integrity (Relationships preserved). The petal colors cycle through navy, teal, and warm gold to differentiate dimensions without ranking them. By rendering the seven dimensions as equal radii of a single hub rather than a checklist, the figure communicates the chapter's structural claim that data quality is multidimensional — a dataset can be perfectly accurate and still fail because it's incomplete, late, or duplicated, so quality programs must measure and govern all seven simultaneously rather than optimizing any one in isolation. 

Figure 2.3.2

Title: Consequences of Poor Data Quality Across Domains

Detailed Caption: A warm-gold warning triangle at the top labeled "POOR DATA QUALITY" sends five downward navy arrows to a row of teal-headed consequence cards: Trust & Decisions (Stakeholders distrust reports; 84% of CEOs concerned about data), Operations (Duplicate mailings, stockouts, wasted resources), Compliance & Risk (Regulatory penalties, audit failures), Analytics & ML (Biased models, unreliable predictions, garbage-in-garbage-out), and Reputation (Wrong orders, misspelled names, eroded customer trust). All five cards funnel a single navy arrow downward to a final navy banner: "Solution: Proactive Data Quality Management." The fan-out-then-converge layout is the figure's pedagogical hinge — it shows that poor data quality isn't a single contained problem but a root cause that branches into five different categories of harm, and the only durable response is a proactive quality program, not five separate downstream fixes. 

Figure 2.4.1

Title: Types of Metadata 

Detailed Caption: A central navy banner reads "METADATA — Data About Data · Makes data discoverable, usable, governable," with six teal-bordered cards arrayed around it in a 2×3 grid and connected by thin teal lines. Each card carries a warm-gold dot icon and a short definition: Descriptive (Titles, authors, keywords · Helps discover data), Structural (Organization, hierarchy · File relationships), Administrative (Access rights, ownership · Retention policies), Technical (File format, size, encoding · Data types), Semantic (Business definitions, ontologies · Meaning of fields), and Provenance (Origin, lineage · Transformation history). The hub-and-spoke composition emphasizes that the six metadata types serve distinct but complementary purposes — descriptive metadata helps you find a dataset, semantic metadata explains what its fields mean, administrative metadata says who can use it, and provenance metadata tells you where it came from — and a serious data catalog needs all six layers, not just the descriptive tags users typically see in a search bar. 

Figure 2.4.2

Title: Data Governance and Data Stewardship Relationship 

Detailed Caption: Three horizontal bands stack vertically. The top navy band, "DATA GOVERNANCE (Strategy & Policy)," lists Governance Council, Policies & Standards, Roles & Responsibilities, and Metrics & Monitoring. The middle teal band, "DATA STEWARDSHIP (Execution & Operations)," lists Quality Management, Metadata Curation, Access Control, and Policy Enforcement. The bottom warm-gold band, "METADATA FOUNDATION," lists Data Catalogs, Data Dictionaries, Lineage & Provenance, and Quality Dashboards. Two arrows link the top two bands — a navy "Sets direction" arrow flowing downward and a dashed gold "Reports metrics" arrow flowing upward — and a vertical gold "Supports" arrow rises from the foundation into the stewardship band. A right-side bracket labeled "Ensures: Data Quality, Compliance, Trust" spans all three layers. The stacked composition encodes the chapter's working analogy: governance defines the rules from above, stewardship operates them from the middle, and metadata supplies the working surfaces from below — and the bidirectional governance/stewardship arrows show why neither layer functions in isolation, since strategy without operations is paper and operations without strategy is improvisation.