🎓 DSS Exam Study Guide

Chapter 6 overview: BI, Data Science, and Data Analytics

Chapter 6 connects decision making with analytics. It explains how BI helps describe the past and present, how data science extends analytics into prediction and prescription, and how descriptive tools such as dashboards and mashups support managers.

Four phases of decision making

1. Intelligence: identify the problem or opportunity, collect information, define goals and criteria.
2. Design: create possible courses of action, build/analyze alternatives, evaluate them against criteria.
3. Choice: select one alternative.
4. Review / implementation / monitoring: execute, monitor, control, and return to earlier phases if needed.

Exam trap: some sources call the fourth stage Review, Implementation, Monitor, or Control. The idea is the same: execute and monitor the chosen solution.

Bounded rationality, satisficing, optimizing

The 3 levels of data analytics

BI vs Data Science

Traditional, modern, embedded, and augmented BI

Seven attributes of modern BI software

1. Speed: real-time questions and answers even with large/diverse data.
2. Visualization: self-service analysis, drill-down, shareable dashboards.
3. Single source of truth: combines data from different sources.
4. Real-time collaboration: live data can be filtered, sorted, discussed, transformed, and shared.
5. Comprehensive governance: safe, trusted, accurate, audited dashboards and reports.
6. Scalability: start small and grow.
7. Mobility: works on phones and tablets.

Data Science lifecycle: 7 stages

Data Science team roles

Software and platforms

• Python: object-oriented, extensible, many free data-analysis libraries.
• R: open-source, statistical and graphical methods, ML, regression, time-series.
• Apache Hadoop: open-source, HDFS, distributed processing, scalable storage.
• Apache Spark: uses RDDs, often much faster than Hadoop because of in-memory processing.
• Descriptive BI leaders mentioned: Microsoft Power BI, Tableau, Qlik, ThoughtSpot.
• Advanced analytics leaders mentioned: Alteryx, SAS, Azure Databricks, Tibco, Dataiku, MathWorks.

Big Data: definition, sources, and 4 Vs

Common sources: social media, sensors, IoT/device data, logs, video, images, web transactions, ERP/CRM, financial data.

Big Data vs Small/Traditional Data

Big Data goals and challenges

• Goals: establish data-driven culture, innovate/disrupt with technology, accelerate services, launch products, improve processes.
• Cultural barriers: data silos, lack of professionals, lack of management support, resistance to change.
• Technical barriers: managing volume/variety/velocity/veracity and presenting results effectively.
• Human expertise remains essential: analytics results need judgment, interpretation, and ethical use.

Four descriptive analytics tools

Data visualization and AR

• Visualization works because humans understand patterns, colors, and spatial relationships quickly.
• Common business value: identify improvement areas, clarify customer behavior, improve product placement, predict sales by location.
• Heat maps use color intensity to reveal patterns; more complex heat maps can analyze website clicks and search behavior.
• Augmented Reality is a high-end visualization form that combines 3D, smart mapping, machine learning, and natural language processing.

Dashboards in Chapter 6

Dashboards combine multiple data feeds and visualizations into a single interactive screen. They are not merely static reports; they connect to accounting, ERP, CRM, web analytics, and other systems to provide current operational visibility.

• Benefits: visibility, continuous improvement, single sign-on/access, budget deviation tracking, accountability.
• Metrics examples: finance: net profit, cash balance, A/R and A/P; sales: leads, proposals, close rate; web: visitors, keywords, traffic sources; operations: inventory, on-time delivery, product output.

Mashups for actionable dashboards

• Mashups reduce the time and effort needed to combine data sources.
• Users can combine fields from different sources and import spreadsheets or competitor data.
• They enable complex queries through drag-and-drop tools for non-experts.
• Important distinction: mashups often work behind the scenes; dashboards are visible and interactive. Mashups can feed dashboards.

Five predictive/prescriptive methods

Time-series analysis: trend, rate of change, and cycles. ML tasks: categorize, predict, identify patterns, detect unexpected behavior.

Chapter 1 overview: decision support, analytics, data science, AI

Chapter 1 explains why organizations need computerized support for managerial decision making, how DSS evolved into BI/analytics/AI, and how analytics and AI fit in a broader ecosystem.

Changing business environment and need for decision support

• Organizations face complex, changing, uncertain, and competitive environments.
• Strategic, tactical, and operational decisions must often be made quickly and sometimes in real time.
• Decisions require data, information, and knowledge; manual processing is often too slow.
• Computerized support helps with speed, consistency, communication, collaboration, and complex analysis.
• Automation is increasingly affecting knowledge work and decision work.

Decision-making process and decision modeling

1. Intelligence: scan reality and identify the problem/opportunity.
2. Design: develop and analyze possible solutions/models.
3. Choice: select the solution/model alternative.
4. Implementation: put the solution to work.
5. Monitoring: evaluate results and feed back into earlier phases.

Chapter 1 uses Simon’s model and emphasizes that decision making/modeling is iterative, not a simple straight line.

Why decision making is difficult

• Data may be unavailable, expensive, imprecise, subjective, or insecure.
• Important data may be qualitative or “soft.”
• There may be information overload.
• Outcomes can appear over a long time, making evaluation hard.
• Future conditions may differ from historical data.
• Many decisions involve multiple objectives and constraints.

Managers, decisions, and support needs

Gorry-Scott-Morton DSS framework

The framework combines two dimensions: type of control and degree of structuredness, producing a 3×3 matrix.

• Structured decisions: repetitive, routine, standard procedures, can often be automated.
• Semistructured decisions: some parts are clear/model-based, others require judgment.
• Unstructured decisions: complex, fuzzy, novel, require human judgment.

Decision Support System (DSS)

1. Supports semistructured/unstructured decisions.
2. Supports managers at all levels.
3. Supports individuals and groups.
4. Supports interdependent/sequential decisions.
5. Supports all phases of decision making.
6. Supports different decision processes and styles.
7. Is flexible and adaptable.
8. Uses friendly interfaces, graphics, and sometimes natural language.
9. Improves decision effectiveness more than mere efficiency.
10. Leaves the decision maker in control.
11. Can be developed partly by end users.
12. Includes models.
13. Can access many data sources.
14. Can be stand-alone, integrated, or web-based.

DSS subsystems

DSS vs BI

Evolution of computerized decision support

BI methodology and architecture

BI typically combines data sources, data warehousing, business analytics, business performance management, and user interfaces such as dashboards.

Analytics overview from Chapter 1

Prescriptive analytics often builds on descriptive and predictive analytics. It is usually not wise to jump to prescription without understanding and prediction.

Selected analytics domains

• Sports: player recruitment, pricing, training, strategy, injury prevention, Moneyball-style insights.
• Healthcare: risk prediction, care management, patient analytics, fraud prevention, resource utilization.
• Retail: assortment, segmentation, pricing, promotions, location, inventory, market basket analysis, customer behavior.

Artificial Intelligence overview

• AI can learn and improve over time.
• AI can understand human language, answer questions, recognize patterns, and automate work.
• Benefits include lower cost, faster work, consistency, productivity, profitability, and competitive advantage.
• Applications include cybercrime detection, e-commerce decisions, high-frequency trading, agriculture, healthcare, robotics, smart devices, and decision automation.

Types of AI by autonomy

Why analytics initiatives can fail

• Predictive models may have unintended effects.
• Models must be used ethically, responsibly, and mindfully.
• Analytics may work well for some applications but not others.
• Garbage-in, garbage-out: data and assumptions determine quality.
• Data can be incomplete, obsolete, inaccurate, inconsistent, or low quality.
• Data from different sources can vary in format and quality.

Convergence of analytics, AI, IoT, and Big Data

• Big Data empowers AI through cheaper processing, large online data sets, and scalable/deep-learning algorithms.
• IoT produces sensor/device streams that feed analytics and AI.
• AI and analytics together can support speech recognition, anomaly detection, underwriting, claims processing, smart buildings, autonomous systems, and connected products.
• Convergence matters because many modern systems combine data collection, prediction, automation, and action.

Analytics ecosystem

The analytics ecosystem includes technology providers, accelerators/intermediaries, and user organizations.

Number to memorize: the ecosystem model has 11 sectors grouped into 3 categories.

Chapter 3 scope: Nature of Data and Preprocessing

The exam instruction excludes the statistical modeling part. This guide focuses on data nature, data taxonomy, preprocessing, business reporting, visualization, visual analytics, and dashboards.

Data-to-knowledge continuum

Analytics exists to move from raw data toward useful knowledge and decisions.

Analytics-ready data: 10 characteristics

Simple taxonomy of data

Art and science of data preprocessing

Preprocessing is essential because analytics projects fail when data problems are ignored. The four main tasks are consolidation, cleaning, transformation, and reduction.

Preprocessing exam traps

• Aggregated data cannot always be safely disaggregated.
• Missing values are not always safe to delete; deletion can bias results.
• The most challenging reduction is often column/variable reduction because it can remove explanatory power.
• Data cleaning is not just formatting; it includes missing values, noise, duplicates, and validity issues.
• Lab work often tests whether you can inspect data quality before visualizing it.

Business reporting

Five functions of business reports

1. Ensure departments are functioning properly.
2. Provide information.
3. Present results of analysis.
4. Persuade others to act.
5. Create organizational memory.

Three major categories

1. Metric management reports: SLAs, KPIs, Six Sigma/TQM-style performance tracking.
2. Dashboard-type reports: multiple performance indicators on one page, often with traffic-light colors.
3. Balanced Scorecard-type reports: financial, customer, business process, learning and growth perspectives.

Good reporting qualities: clarity, brevity, completeness, correctness.

Data / information visualization

• Visualization is central to BI and analytics because it reduces cognitive load and reveals patterns.
• It is connected to information graphics, scientific visualization, exploratory data analysis, and human-computer interaction.
• Historical examples include Playfair’s line/bar/time-series charts and Minard’s Napoleon campaign chart, praised for encoding multiple dimensions.

Chart selection: basic charts

Chart selection: specialized charts

Abela chart taxonomy: choose by message

Information visualization vs visual analytics

Data storytelling

1. Think of the analysis as a story with a narrative structure.
2. Be authentic; the story should flow naturally from the data.
3. Be visual; think like a film editor and remove distracting material.
4. Make it easy for both audience and presenter.
5. Invite and guide discussion.

Storytelling matters because reports and dashboards should lead to understanding and action, not just display data.

Visual analytics platforms: key value

• Empower a broader audience with data exploration and approachable analytics.
• Use interactive web interfaces to broaden access.
• Answer complex questions faster.
• Improve collaboration and information sharing.
• Free IT from building every report by giving users controlled self-service access.
• Allow growth from simple reporting to advanced analytics at the organization’s pace.

Information dashboards

Eckerson’s 3 information layers

Dashboards are used for executives, operations, sales, marketing, finance, HR, logistics, healthcare, education, and more.

Dashboard design: good characteristics

1. Use visual components such as charts, sparklines, gauges, and stoplights appropriately.
2. Be transparent and require minimal training.
3. Combine data from multiple systems.
4. Allow drill-down and drill-through.
5. Provide dynamic views with timely refresh.
6. Use minimal custom code where possible.

Dashboard best practices

1. Benchmark KPIs against industry or organizational standards.
2. Wrap metrics with contextual metadata.
3. Validate design with usability specialists or representative users.
4. Prioritize alerts and exceptions.
5. Enrich the dashboard with business-user comments where useful.
6. Present information in three levels: dashboard, static report, self-service cube/detail.
7. Pick the right visual construct for the message.
8. Use guided analytics to move average users toward expert use.

Typical comparisons in BI dashboards

• Current values vs past values.
• Actual values vs forecasted values.
• Actual values vs target values.
• Values vs benchmarks/averages.
• Multiple instances of the same measure.
• One measure vs another, e.g., revenue vs cost.

Lab checklist: before building a dashboard

1. Define the business question and audience.
2. Identify KPIs and the decision/action each KPI supports.
3. Check data source reliability, accuracy, timeliness, granularity, and relevance.
4. Clean and transform data before visualization.
5. Choose charts by purpose: relationship, comparison, distribution, composition.
6. Avoid clutter: one screen should support quick understanding.
7. Use filters/drill-downs when users need exploration.
8. Test with users and revise based on confusion or wrong interpretation.

Flashcards

Click a card to flip it. Use search to filter terms.

Quiz

Answer each question and read the explanation. You can reset or reveal all answers.

Must-memorize numbered lists

High-probability exam traps

• BI is mainly descriptive and handles known unknowns; Data Science predicts/prescribes and handles unknown unknowns.
• Predictive says what is likely to happen; prescriptive recommends what to do.
• DSS supports decision makers; it does not replace them.
• DSS is usually for specific problem support; BI monitors and identifies problems/opportunities.
• OLTP is transaction processing; OLAP is analytical processing.
• A histogram is about frequencies/distributions, not categorical comparison like a bar chart.
• Pie charts are only good for simple proportions with few categories.
• Heat map uses color intensity; highlight table adds numbers.
• Mashups are often behind the scenes; dashboards are visible.
• Data quality and preprocessing come before visualization.
• Dashboard design must match KPIs and decisions, not just look attractive.
• The PDF says to exclude statistics and emphasize visualization and dashboards.