INFS1000 Lec 11 Business Intelligence

Question 1

What is BUSINESS INTELLIGENCE?

Answer 1

Methodologies / technologies that transform raw data into useful info for business purposes.

Question 2

Why do organizations need business intelligence? (2)

Answer 2

1. Business collect massive amounts of data: so need to provide right info to right person at right time.
2. Data can show patterns and relationships: info that can help (1) better decision making and (2) forecast changes (e.g. drop in site views).

Question 3

What are the 2 ways BI can help businesses?

Answer 3

1. Data warehouses: storing and organizing large amounts of data.
2. BI Tools / Systems: making sense of data from DWs.

Question 4

How does BI achieve this? (3)

Answer 4

1. DW: Integrating data: from multiple internal and external sources (e.g. departments, customers, suppliers).
2. DW: Preparing data: data clensing.
3. BIT: Analysing data: (e.g. summing, grouping, averaging).

Question 5

How does a DATA WAREHOUSE work? (3)

Answer 5

• Regularly copies data from OPERATIONAL DATABASE to create massive stock of HISTORICAL data for analysis purposes. (e.g. sales from WW every night).
• Collates data from EXTERNAL RESOURCES (e.g. demographic, weather data).
• Creates meta-data for these large pools of data.

Question 6

What problem does OPERATIONAL DATA pose for BI systems?

e.g. sales transactions, customer records, employees data etc.

Answer 6

• Operational data is “RAW”, collected to meet IMMEDIATE BUSINESS OBJECTIVE (e.g. generate a report), usually “DIRTY”, hence unsuitable for sophisticated / large-scale BI analysis.
• Doing analyses on operational databases might slow them down and affect day-to-day business.

Question 7

What are the 4 types of DIRTY DATA?

Answer 7

1. Incorrect data
2. Values missing (e.g. postcode missing from address).
3. Data INCONSISTENT: product data changes over time (i.e. diff colours / versions), time-zone specific data in international sales.
4. GRANULARITY inappropriate: too much / little detail.

Question 8

What are the 3 types of BI Methods / Tools?

Answer 8

1. Reporting: (1) filtering (2) sorting (3) grouping (4) simple calculations.
2. OLAP - Online Analytical Processing: MULTIDIMENSIONAL aggregation / visualization of historical data.
3. Data Mining: sophisticated statistical tehcniques to discover patterns / relationships used to predict future outcomes.

Question 9

What are the advantages of OLAP?

Answer 9

• Very quick (i.e. in real time), thus enables executives to make timely decisions.
• Can view information in different configurations.

Question 10

Define the 7 main applications of data mining:

1. Consumer clustering
2. Customer churn
3. Fraud detection
4. Direct marketing
5. Interactive marketing
6. Market basket analysis
7. Trend analysis

Answer 10

1. Consumer clustering: identify common characteristics of customers who tend to buy same products / services.
2. Customer churn: identify reason customer switch to competitors, predict why customer are likely to do it. (e.g. complaint history, number of days to contract expiry, phone model).
3. Fraud detection: identify characteristics of transactions that are most likely to be fraudulent.
4. Direct marketing: identify which prosepctive clients should be included in mailing lists to obatin highest response rate.
5. Interactive marketing: predict what each individual accessing a website is most likely to be interested in seeing.
6. Market basket analysis: understand what products / services are commonly purchased togehter, and on what days of the week.
7. Trend analysis: reveal difference between typical customer this month and last month.

Question 11

What are the 6 things that drive data mining?

Answer 11

1. Statistics / maths.
2. AI / Machine learning.
3. Huge databases.
4. Cheap computer processing and storage.
5. Sophisticated marketing, finance, and other business professionals.
6. Data management tech.

Question 12

What are does MARKET BASKET ANALYSIS help with?

Answer 12

Identify CROSS-SELLING OPPORTUNITIES.

Question 13

Define probability, support, confidence, lift.

Answer 13

P(Mask)

Support P(Mask AND Fin)

Confidence P(Mask / Fin) = P(Mask AND Fin) / P(Fin)

Lift: ratio of confidence to the base probability of just buying items P(Mask / Fin) / P(Fin).

Question 14

What are the 7 main applications of data mining?

What are some extra ones?

Answer 14

1. Consumer clustering
2. Customer churn
3. Fraud detection
4. Direct marketing
5. Interactive marketing
6. Market basket analysis
7. Trend analysis

Other: inferring demographics (e.g. Amazon), loyalty programs (e.g. casinos and hotels).