Business Analytics - Lecture notes, lecture 1 - 10 PDF

Preview

Summary

Download Business Analytics - Lecture notes, lecture 1 - 10 PDF

Description

Lecture 1 – Introduction to business analytics Business analytics is the ability of firms to collect, analyse and act on data. It is the ability to generate knowledge such as: - What products customers want? - What prices will customers pay? - How many will each customer buy? - Why do they buy? Why do they buy more? - How do problems affect the bottom line? - To predict demand, performance and problems. - Identify opportunities, discover new segments. It is the use of data, information technology, statistical analysis, quantitative methods and mathematical or computer-based models, to helps managers make better decisions. It is a combination of (STAP) skills, technologies, applications and processes used by organisations to gain insight into their business based on DISQM-C to drive or guide business planning. Competing on Analytics In order for organisations to compete on analytics, Davenport suggests they must have data, personnel and procedures. Where is the data in an organisation? Operational systems: They are used to run day to day business operations – also known as Transaction processing systems. It uses OLTP databases to store daily business transactions. OLTP is a computer system where time-sensitive, transaction-related data is processed immediately and is always kept current. It is used for order entry, financial transactions, and CRM and retail sales. It is a class of information systems that facilitate and manage transaction-oriented applications, typically for data entry and retrieval transaction processing. Characteristics of data under this system: - Transaction-oriented - May be inconsistent and incomplete - Volatile (changing continuously) - Current. Databases are everywhere. - Efficient transaction processing Automatic data capture Easy retrieval Data quality, errors, inconsistencies - Automating business processes Production, Marketing, Accounting, Sales Databases are great, but

- Too many of them Everybody wanted 1 or 2 or more Production, Marketing, Accounting, Sales - Everybody got what is best for them IBM, Oracle, Access, Excel, file drawers - Created the problem databases were meant to solve Inconsistent, incompatible, inaccessible data As a result, data are not effectively employed and hence opportunities for improving performance are missed and there are problems with analysing data to improve decisions and performance. Organisations need to store data for day to day operations. Data is being stored in many databases (usually duplicating data across them, and usually incompatible with each other). Organisations could make use of this data to improve the decision making process through some sort of business analytics or business intelligence. This could be accomplished via some sort of computer system(s). What is decision making? It is a thought process of selecting a logical choice from available options. It is a process of identifying and selecting – solve a specific problem – choice making which is part of decision making, i.e., select one option from a set of alternatives. There are programmed decision (structured decisions for routine problems) and nonprogrammed ones (unstructured decisions for unique and unusual problems – rational decision making). Simon (Nobel Lecture) on rational decision making in organisations: “ We do know how the information processing system called Man, faced with complexity beyond his ken, uses his information processing capacities to seek out alternatives, to calculate consequences, to resolve uncertainties and thereby – sometimes, not always – to find ways of action that are sufficient unto the day, that satisfice”. Examples of analytics Experiments - test alternative strategies, product designs - different interest rate v/s fees - ad messages - incentives (cash back v/s loyalty points) Identify best customer segment - for each product design, promotion - identify optimum price Simulate effect on financial performance Decisions justified by data - Compensation, rewards, advertising, pricing, R&D, mergers, acquisitions - “In God we trust. All others bring data”

- “Do we think this is true? Or do we know?” - Measuring, testing and evaluating quantitative evidence (Internal and external evidence, analysis) - Companywide practice - Gotaas-Larsen shipping corporation in late 1970s It used a decision support system (DSS) for preparing and revising a 15 month operation plan for its cargo ships. The model-driven subsystem supported cash flow and pro forma analyses on a per ship, per voyage, per division and company-wide basis. The DSS helped users simulate results. The computer system aggregated plans for individual feasible voyages to help managers assess whether the overall plan would be effective. - Port of Melbourne Authority – Lotus 123 spread sheet designed to simulate the arrival, unloading and loading of cargo ships from the point (took >24 hours to run each time). It helped to plan arrival times of ships and also helped to estimate when and where workers needed to be (within various union guidelines). What personnel (with the right skill sets) do we need? 

Expert employees Essential for running BA as a company-wide practice. Expertise with numbers, or training. Arm with best available data and best quantitative tools. Better decisions (big and small, everyday; modelling, optimization, opportunities)

What procedures do we need? Organizational practices  



Multiple applications - Not one “killer” application, all parts of the organisation Centralised resource, easily shared - Expertise, army of PhDs with quantitative, statistical, modelling, and optimisation skills - “bilingual experts” who understands business as well as analytics - Integrated data from multiple sources, DW Apply to entire supply chain

Some success stories Business analytics in action at MARRIOTT They introduced an application called the RPO (Retail pricing optimizer) to predict consumer behaviour and optimise product availability and price to maximise revenue growths. The RPO helped hotels pinpoint where opportunities are exactly. Data about their business trends to their market conditions are complex data and BA simplify them to make good business decisions. RPO helped determine the optimal, transient rate, using analytically-driven,

market-based methodology. Basically, it provides a price elasticity model and helps make recommendations. It helped increase efficiency in analysing and decision-making processes.  Revenue Management - Establishing the optimal price for guest rooms - Analysis of own past data, competitor data (purchased) using quantitative models - Squeeze the most revenue from your inventory of rooms - Aids decisions of front line managers  Tools and expertise widely available  Consistently superior performance Types of BA analysis 





Forecasting and predictive modelling - Extrapolate historical data, trends - Statistical techniques: regression, time series Simulations - “What ifs” analysis, model - analyse alternative assumptions and scenarios Just give me this information - Information retrieval and visualisation - Probably the most used functionality

Clever thing you can do with BA  Harrah‟s Winners Information Network - For your clients, all of whom will lose money  Customer service relationships built on customer knowledge - learn from customer behaviours and preferences - where they gambled, how often, what games they played, how much they gambled, what offers entice them, target offers - Objective: make them “play” (lose) more and maximise profitability across various casinos More BA examples 



Web crawling - Collect information on all mobile plans - Plans posted on company websites - Organise by areas, competitors and so on - Sell to all competitors - One input in competitive response Mobile usage - Telcos collect information on all calls, internet usage, websites visited, phone numbers called



- detect changes in individual usage patterns - individual promotion, select form canned options Web usage, analysis of clickstream data

Strategic BI: - Product launch  

 

Test market products before launch Estimate national sales - diffusion models project from test market data - fairly complex but reasonably accurate Estimate promotional expenditure Managerial decision - test data support realistic cost benefit analysis

- Market segmentation 





Different buying patterns - Segment on the basis of age, education, income, personality variables, ethnic background and so on. Analyse data to identify segments - Market research, historical data - Progressive Insurance segments by risk Customize marketing programs - Improve response in each segment - Premiums set based on risk segment

Evaluating strategic decision 





Strategic decisions - Commitment of specific resources (managerial, financial, technical and so on); substantial and irreversible; to a course of action; with a view to improving performance Identify strategic vulnerabilities - Changes in external environment (taxes, regulation, government policy, technology), customer taste, competitor actions and so on. Identify pre-emptive and contingent actions - Models, sensitivity analysis

Continuous process improvement 

Input Transformation processes O/p - March and Hevner‟s management 101 primer - Generic model of organizations - Generic context of decision making and BA



Focus on organizational processes - Input processes, transformational processes, output processes - Improving efficiency and effectiveness of business processes, continuous cycles - Data and analysis-based improvements

Steps in achieving BA capability o Integration Operational databases, Internal information, External information o Implementation Data warehouses, Model management, Analytic tools, User interfaces o Intelligence Management, Decision support and Data mining tools o Innovation Corporate vision and strategies, Change management

Lecture 2 – Introduction to data warehousing Concepts and architecture Where is the data? See previous lecture. So what is the problem? We are doing just fine - data are valuable Islands of data; integrated data not available - Improve processes, improve performance - Analysis, decision-making If only we had the tools to analyse data - Competitive performance suffers Our competitors are improving and we are not! -The problem? Local optimisation  Global sub-optimisation Organisations are not able to capture benefits from its data and technology assets> “Data in jail” The solution Look for a technology that     

Optimises information for the whole organisation Looks at information over time Has the data validated Improves data quality Reduce costs of decision making and so on

So, rather than the traditional transaction based database, we want a database about the organisation‟s information – we essentially want a system where we can “play” with that data. Motivations for data warehousing    

Demands on OLTP databases for query processing would be too great. Data warehouse is designed for efficient retrieval. Data in legacy systems is frequently redundant, inconsistent, of poor quality, and stored in different formats. Reduce costs in providing data for decision makers

Characteristics of Transactional v/s Informational database systems Characteristic

Transactional

Informational

Primary purpose

Run the day to day business

Support decision-making

Type of data

Current data – representing the state of the business

Historical data – snapshots and predictions

Primary users

Customers; clerks and other employees

Managers; analyst

Scope of usage

Narrow, planned, fixed interfaces

Broad, ad hoc, complex interfaces

Design goal

Performance and availability

Flexible use and data accessibility

Volume

Many constant updates and queries on a few tables or rows

Periodic batch updates, complex querying on multiple or all rows

Data warehouse description Inmon and Hackathorn: “…subject-oriented, integrated, time-variant, validated and non-volatile collection of data in support of management’s decisions” 







Subject-oriented The DW database is organised by “data subjects” that are relevant to the organisation. Example: Sales, claims and shipments. This may be contrasted with the process orientation of many transaction processing systems. Integrated Data in the DW is structured based on a corporate-wide model, spanning the functional boundaries of legacy systems. This includes naming standards, units of measurements and periodicity. Time-variant Data in the DW is characterised by time-series nature of historical data. The data consists of a series of “snapshots” which are time stamped and record values at a moment in time. This supports trend analysis of the data. Validated Data from various sources are validated before storing them in a DW. Data quality is crucial to the credibility of the warehouse.



Non-volatile (Read-only – only the operator can modify the data) The DW is not continuously updated (inserts, deletes, changes) like data in OLTP system. Data in the DW is periodically updated at scheduled time intervals.

DWs v/s operational databases So, a data warehouse database      

Is subject rather than application oriented Is integrated rather than only partially integrated Is non-volatile rather than continuously updated Has stabilized data values rather than current data values Provides ad hoc retrieval rather than predictable retrieval Provides slice and dice

Data organisation  Star schema v/s relational Data warehouse development There are a number of ways to develop a DW - Data mart (Bottom up  start small to big); the “Kimball” methodology - Enterprise-wide warehouse first (Top down  make big then little ones); the “Inmon” methodology - When properly executed, both result in an enterprise-wide data warehouse The data mart strategy  The most common approach  Begins with a single mart and architected marts are added over time for more subject areas  Relatively inexpensive and easy to implement (On average 9 months and $1.5m)  Can be used as proof of concept for data warehousing  Can perpetuate the “silos of information” problem  Require an overall integration plan The enterprise-wide strategy     

A comprehensive warehouse is initially built An initial dependent data mart is built using a subset of the data in the warehouse Additional data marts are built using subsets of the data in the warehouse Like all complex projects, it is expensive, time consuming and prone to failure When successful, it results in integrated, scalable warehouse

Data warehouse architecture

Data sources Source data systems Where does the data come from? - Could be existing organisations systems Different types of databases Different languages (Both programming and people languages) - Could be from outside the organisation Example: Stock market data, interest rates, currency exchange rates - Could be structured (Akin to machine language – information with a high degree of organisation and readily searchable by simple straightforward search engines or other search operations – easily understood and readable by computer) - Could be unstructured (Makes compilation a time and energy consuming task – usually for humans who do not interact with information in strict data base format – example – a phone call or all the organisation‟s correspondence). Data staging area (Operational data store) This is where all the work is done on the data. - Data is imported from source systems to the Operational Data Store. An ODS consolidates data from a number of source systems and provides a near real time, integrated view of volatile current data. Its purpose is to provide integrated data for operational purposes. It has add, change and delete functionality. - Data from application systems are „cleaned up‟ and stored in the ODS. - Halfway between the OLTP and DW - Data from the ODS are fed into the DW and data marts. Data Extract, Transform, Load (ETL)

 Data sources - Old applications, ERP systems, external  Data extraction - Custom write or use commercial packages  Data Transformation - Cleansing, standardisation and integration  Data loading - Periodic bulk loading or real time update Data quality As part of the ETL process the concept of data quality is extremely important: - Data quality problems are widespread in practice and have significant economic impacts. - Data bases have significant error rates. Data and metadata storage area This is the “Data warehouse”      

I.e. a BIG STORAGE AREA that stores the data for the DW. May be a single “DW”. May be a “DW” with data marts. A data mart stores data for a limited number of subject areas such as marketing and sales data. It is used to support specific applications. An independent data mart is created directly from source systems. A dependent data mart is populated from a DW.

Essentially this is a database system - Can be relational - Can be dimensional Metadata   

Data about the data - For both IT and business people Includes definitions, sourcing details, refresh schedules, quality indicators… Can be integrated within the data warehouse or in a separate linked product

End user presentation tools (Data access) This is essentially the Analytics area (but not only analytics area) 



Analysis of data - Data mining - Modelling - Ad hoc querying Report writing

 

Visualisation tools Other end user applications

Data analysis tools  Basic query and reporting  Basic question: What happened? Historical focus, limited flexibility. Data source: OLTP database, ODS, data mart, data warehouse.  On-Line Analytical Processing (OLAP)  Basic question: what happened and why? Historical focus, multidimensional - look at the data from many point of views, with medium flexibility (slice and dice, drill down). Data source: data mart, data warehouse  Data mining - Basic question: what is interesting? What might happen? Future focus, high flexibility Extract relationships, patterns and trends, predict future trends Data source: data warehouse Types of analytic analytics and Clever thing you can do – see previous lecture. More clever things  Marriott: optimal room price in real time Loyalty programs, competitive intelligence Yield management  Capital One experiments Predict impact of interest rates, incentives, etc.  Wal-Mart: optimize supply chain flows Simulation, predictive modelling Summary o Data is crucial to business analytics. o Informational systems are designed to support decision making, while operational systems are designed to support daily transactions. o A data warehouse is a subject-oriented, integrated, time-variant, validated and nonvolatile collection of data. o Data warehouse consolidates data from various sources to support decision making. o Data mining tools can provide insights (knowledge discovery) and make predictions based upon historical data. Case study Harrah‟s entertainment – data warehousing supported a successful shift to a CRM oriented corporate strategy.

Background          

Established in 1937 In 1993, the gaming laws changed which allowed Harrah‟s to expand. 25 casinos in 2000 Acquires Caesars entertainment in 2005 Renamed Caesars Entertainment Corporation in 2010 Currently, it is the world‟s largest gaming company They have over 50 properties and all are linked together They decided to compete a brand strategy supported by information techno...