Ch 8 - decision making and business intelligence

What is decision making in management?

Choosing from a range of alternatives.

• It’s a key part of management, often done naturally, like deciding how to spend your time.

Why is rational decision making complicated?

1. Rationality is hard to define. 2

2. Good outcomes can come from bad decisions, and vice versa.

3. Humans have limited ability to process all options, known as 'bounded rationality'.

Bounded rationality

Herbert Simon’s idea that humans try to be rational but can’t consider all options due to limited brainpower. We satisfice, or settle for a good-enough option.

Satisficing

Choosing a solution that is good enough instead of perfect.

Ex: Picking a used car that mostly fits your needs, rather than searching endlessly for the perfect one.

Ackoff's criticisms of early MIS assumptions?

1. More data doesn't always help. 2

2. Managers have too much info, not too little (information overload). 3

3. Managers often don’t know what data they actually need.

Information overload

Having too much information to process effectively.

• It causes confusion and poor decision making.

• Ex: Asking for more data than needed just in case.

Why does data growth matter in decision making?

Huge amounts of data are being created fast (e.g., 44 zettabytes by 2020).

If used properly, this data can improve decisions, but it can also overwhelm managers.

What are examples of bad data quality issues?

• Dirty data (e.g., misspelled colors)

• Missing values

• Inconsistent formats (like phone numbers)

• Unintegrated data from different sources

• Inappropriate granularity

Data granularity

Refers to how detailed data is.

• Fine = very detailed (e.g., click-by-click)

• Coarse = summarized.

• Too fine can be combined, but coarse data can’t be split further.

How can information systems help with decision making?

They can process and summarize large data sets, helping managers find the right information and avoid overload, if the data is high quality and relevant.

OLTP (Online Transaction Processing)

OLTP systems collect and process business transactions electronically and in real-time or batches.

Ex: banks, stores, and airline systems.

Real-time vs batch processing

Real-time processing: updates the system immediately after each transaction (e.g., Ticketmaster)

Batch processing: waits to group transactions before updating (e.g., gas stations at day-end).

How do OLTP systems support decision making?

They provide raw, up-to-date data on transactions and operations, which organizations can later analyze for strategic decisions.

OLAP (Online Analytical Processing)

Helps turn data from OLTP systems into meaningful insights using sums, averages, and comparisons.

• It allows flexible, interactive reports.

What are OLAP measures and dimensions?

Measures: the numerical values being analyzed (e.g., total sales)

Dimensions: the categories for analysis (e.g., time, region, product).

What is an OLAP cube?

Shows data in multiple dimensions (like product, region, and time).

• It’s interactive, allowing users to drill into or rearrange the data.

What does 'drilling down' mean in OLAP?

Breaking data into more detail.

Ex: looking at sales by state, then zooming into individual cities.

Why are OLAP servers needed?

Standard databases aren’t made for complex analysis. OLAP servers handle the heavy lifting of sorting, calculating, and formatting large OLAP reports.

• OLAP itself isnt a server, but servers are used

What is the data resource challenge?

Even though companies collect lots of data (e.g., from scanners), they often don’t use it well. The challenge is treating data like a valuable asset and using it for decision making.

How can treating data as an asset improve decision making?

If data is seen as a resource, like money or property, companies will focus more on managing it well and using it to make smarter decisions through business intelligence.

Business Intelligence (BI) system

Provides information to improve decision making by analyzing data and offering insights that create competitive advantage.

Group Decision Support System (GDSS)

Allows multiple decision makers to collaborate (often anonymously) across different locations and times, reducing biases in group decisions.

What is the competitive advantage of GDSS?

GDSS improves outcomes by minimizing group biases and improving collaboration across locations and time zones.

What are reporting systems in BI?

These systems integrate and process data by sorting, grouping, and summing, then deliver formatted reports to users.

What advantage do reporting systems provide?

They improve decisions by providing relevant, accurate, and timely information to the right people.

What is a data-mining system?

A system that uses statistical techniques (e.g., regression or decision trees) to find patterns and relationships in data.

How do data-mining systems improve decisions?

They predict outcomes and identify patterns, like predicting which customers are likely to buy or donate.

Knowledge Management (KM) system

Share knowledge (e.g., best practices, product details) among employees, managers, and customers to create value.

What advantage do KM systems offer?

They foster innovation, improve customer service, increase responsiveness, and reduce costs by leveraging collective knowledge.

What is an expert system in BI?

A system that encodes human knowledge using If/Then rules to make recommendations or diagnoses.

What advantage do expert systems provide?

They help non-experts make better decisions by applying expert knowledge automatically.

What is RFM analysis?

A method for ranking customers based on how Recently (R) they purchased, how Frequently (F) they buy, and how much Money (M) they spend.

How is RFM scoring done?

Customers are divided into 5 groups for Recency, Frequency, and Money spent (1 = top 20%, 5 = bottom 20%). These scores are used to rank and target customers.

How does BI relate to sports decision-making?

Teams use BI tools (like statistical analysis) to pick athletes based on data, reducing reliance on gut feelings. This 'Moneyball' approach improves accuracy.

Why don't analysts use operational data directly for BI?

Operational data is optimized for transactions, not analysis. Using it directly can risk errors, harm performance, and slow down operational systems.

What is a data warehouse?

A data warehouse is a system that extracts, cleans, organizes, and stores data from multiple sources so it can be used for business intelligence analysis.

What are the functions of a data warehouse?

1. Obtain data

2. Cleanse data

3. Organize and relate data

4. Catalog data

What kind of data goes into a data warehouse?

Data can come from operational systems, purchased external data (e.g., credit reports), and other external sources.

What is metadata in a data warehouse?

Metadata is 'data about data'. It includes information about where the data came from, its format, and how it should be interpreted or used.

What is the role of the Data Warehouse DBMS? (data base management system)

It stores and manages cleaned data for analysis. It may use a different system than the one used for daily operations (e.g., SQL Server vs Oracle).

What do business intelligence tools do with warehouse data?

BI tools analyze, visualize, and report on data from the warehouse to help managers make informed decisions.

Who manages a data warehouse?

A small team: technical staff manage storage and systems, while business analysts ensure the data meets user needs.

What does the data warehouse diagram show?

Data comes from operational, purchased, and external sources. It's cleaned and processed, stored in the warehouse DBMS, and used by BI tools to support users.

What is a data mart?

A smaller data collection designed to serve a specific business function, problem, or opportunity.

What is an example of a data mart?

An e-commerce company may have a data mart for analyzing website clicks (clickstream), one for analyzing sales patterns (market-basket analysis), and another for managing inventory layout.

How is a data warehouse different from a data mart?

A data warehouse is a central storage that collects, cleans, and organizes data from many sources. A data mart is a smaller, specific portion of that data designed for a particular business need.

What is a data warehouse?

A system that stores large amounts of data collected from operational systems and external sources. It prepares the data for business intelligence (BI) tools.

What analogy helps explain the relationship between data warehouses and data marts?

A data warehouse is like a distributor in a supply chain, it stores and prepares data. A data mart is like a retail store, it focuses on specific needs using data from the warehouse.

What kinds of professionals typically work with data warehouses?

Data warehouse staff are experts in data management, cleaning, and transformation, but not necessarily experts in specific business functions.

Who typically uses data marts?

Business analysts or department specialists who understand a specific area (like sales, marketing, or inventory) use data marts for targeted analysis.

What are the three examples of data marts?

1. Web Sales Data Mart (for analyzing clickstream data) 

2. Store Sales Data Mart (for store management using sales data) 

3. Inventory Data Mart (for planning inventory layout)

Why might smaller organizations only use data marts?

Data warehouses and data marts can be expensive to build and operate. Smaller organizations may only need a simple data mart for specific analysis needs.

How do BI tools relate to data marts?

BI tools analyze the data in data marts to help with tasks like webpage design decisions, store management training, and inventory planning.

What is data mining?

The process of using statistical techniques to find patterns and relationships in large sets of data.

What fields contribute to data mining?

Data mining combines statistics, mathematics, artificial intelligence, machine learning, data management technology, cheap computing/storage, and business expertise.

What are the two types of data mining?

1. Unsupervised: No model or hypothesis beforehand; patterns are discovered from data.

2. Supervised: A model is made first, and data is used to test or build it.

What is cluster analysis?

A type of unsupervised data mining that groups data into clusters based on similarities, like grouping similar customers.

What is regression analysis?

A supervised data mining technique that predicts a value based on the influence of other variables, using a formula.

What are neural networks in data mining?

Neural networks are complex sets of equations used to make predictions or classifications, often used in supervised learning.

What is market-basket analysis?

A data mining method used to find which items are frequently bought together (e.g., masks and fins).

What is support in market-basket analysis?

Support is the probability that two items are bought together in a transaction.

What is confidence in market-basket analysis?

Confidence is the conditional probability that someone buys item Y given that they bought item X.

What is lift in market-basket analysis?

Lift compares confidence to the base probability of buying an item.

• A lift > 1 means the likelihood increases.

What is big data?

Big data refers to very large, diverse, and fast-changing datasets used to find patterns and help with better decisions.

Why is big data controversial?

Because it’s expensive, vague, can lead to misleading patterns, and often collects too much personal data.