Business analytics - Cours de Diego Kuonen. PDF

Preview

Summary

Cours de Diego Kuonen....

Description

Business analytics 1. Introduction to business analytics and motivation Business collect a lot of raw data. They need to treat them in order to make better decisions and choices. Big data = productivity growth, innovation, consumer surplus Many businesses are drowned in data and they lack professionals to seize this competitive advantage. Business analytics = methodology employed by an organization to make optimized decisions based on data. With statistical thinking, the business improves its products, services, operations, human resources, financial management and supply chain. Types of analysis -

Descriptive analysis  what is happening? Diagnostic analysis  why did this happen? What are the trends? Predictive analysis  what will happen? Prescriptive analysis  what to do? How to make it happen?

Statistical thinking principles: -

Sequential approach  continuous analysis of several data sets Strategic thinking  having a strategy and business objectives Data pedigree  considering the data quality before, during and after the analysis Subject matter knowledge  knowing the business context in which the analytics is applied

2. Need for business improvements Management practices need to change. The prosperity of a firm highly depends on its ability to improve its activity. Today’s consumer demand more and global competition has grown. Organisations will not survive if they continue to make product or services how they did before. They have two jobs: do their work and improve how they do their jobs. The managers should assess continuously their practice in the seek of improvement but analytics and statistical thinking is an aid to thinking and should not replace it. Businesses activities are subject to variation and uncertainty. If they want to improve their operations, they need tools to measure and understand this uncertainty. One of these tools is statistical thinking. Theses tools allow to take data-driven decisions and reduce the importance of the opinion of the highest paid person in the room. They also reduce the potential harm of cognitive bias. -

Anchoring bias Availability heuristic confirmation bias Overconfidence Ostrich bias Selective perception

Overall model for business improvements

Doing activity: Represented above by the “business process” chart. It is a sequence of action that transforms inputs into outputs. Improving activity: There are many ways to improve the activity but we’ll focus on process improvement and problem solving. Process improvement: a series of activity that aims at improving the performances of the process. Problem solving: address specific problems that are not part of the typical behavior of the process. Data are the link between the doing activity and the improvement activity. The doing activity generates data that will fuel the analysis conducted in order to improve the process. They document the process performance, identify problems and evaluate the impact of proposed solutions.

3. Statistical thinking It is a philosophy of learning and action based on these principles: 

All work occurs in a system of interconnected processes Variation exists in all processes Understanding and reducing unwanted variation is key to success Using data whenever possible to guide actions and improve decision making

From a statistical point of view, improving activity can be made in two ways: -

Reducing variation (variance) Improving overall level (average)

Quality is inversely proportional to variation. A customer would prefer receive his bill in a predicable period of time. Data-driven decision making

4. Principle of statistical thinking First principle: all work occurs in a system of interconnected processes Process: one or more activities that transforms input into output Process map/system map: charts that show the overall work from suppliers to customers that helps implementing improvements.  SIPOC: suppliers – input – process steps – output – customer

The process itself can be represented in a flowchart. This will help identify the part of the process that generates problem or that can be improved.

This is Business Process Management (BPM). We can divide processes into sub-processes to go into more details. In this chart we include side activities that contribute to a high quality service. For example, the greetings of the customer, customer service, product development, etc. “blame the process, not the people”  85% of problems are caused by the process and not by the people working in it. Process measurements and operational definitions An operational definition is a set of specific instructions about when, where and how to obtain data.  Instructions that allow people to look at the same thing at the same time and have the same measurements  For example, we can ask people to tell how many countries there are on earth, but if they have different definitions of a country they will arrive at different numbers. To determine if a plane arrive on time, we must define the delay that we accept before declaring it late. Process measurement tracked over time: In order to evaluate the process over time we need to measure data regularly. -

Assess current performance

-

-

Determine if it has changed over time o If it must be adjusted (minor changes) o If it must be improved (major changes) Predict future performances

Second principle: variation takes place in all processes Variation is the enemy of quality and must be understood and controlled to improve process and offer higher quality service or product. Variation is desirable when it is intended because it offers choice.

In this flowchart the part of the process where the variation is most likely to occur are designated with a black square.

Third principle: understanding and reducing variation are key to success There are two types of unintended variation that we must differentiate: -

-

Common cause variation o Ordinary variance of the process o Chance cause of variation o Process improvement approach Special cause variation o Variation due to a problem o Assignable cause of variation o Problem solving approach

Special cause variation is produced by causes that are not inherent to the process. Common cause variation is a natural part of the process. We need to distinguish one from the other because we must take different actions.

Three key business analytics questions: -

-

-

How am I doing ? o Understand and agree on key process metrics o Key performance indicators (KPIs) What drives my business? o Once the KPIs are identified, you must understand what influences the performance o What are the drivers/factors that influence the KPIs Who are my customers? What are their needs? o Understand customers and customize their offering, messaging, marketing channel o Drive the KPIs in the right direction

it is important to address the improvement properly. To do so, write an objective improvement statement. The statement must include attainable goals, the time required, the level of efforts needed. Use the method “SMART” to write the objective improvement statement:

Example: customers complain about the delivery time of pizza “the objective is to reduce the number of pizzas delivered in more than 30 minutes from 55 percent to less than 5 percent by June 30 of next year. Doing so will further enhance the pizzeria's image of excellence and will save the company over $60'000 annually.”

Red beans experiment The red beans experiment shows that it’s not a good idea to predict future performance of people based on past. The success of workers is directly linked to the system.

5. Statistical thinking strategy The use of statistical thinking helps organization to improve their operations and management. It can be used in every part of the organization. A manager wants to understand the variation to prevent rework, loss of time, improve the quality of the work. Soccer example A soccer coach used statistical thinking in order to improve the quality of his team: 1) 2) 3) 4) 5)

Study the situation (find out the character and skills of each players) Define quality performance (numbers of game won) Analyze performance (divide goals into smaller subgoals) Focus on the main problems Monitor and evaluate progress

Statistical thinking strategy

Business process: identifying, documenting, understanding the process Use subject matter knowledge (theory) to gather information and form hypothesis. It helps us to predict which bias can occur and avoid them. Once the data are collected, statistical tools and techniques are used to quantify variation. Some hypothesis are confirmed and new ones emerge. We gather additional data and repeat the cycle.

Most business application are sequential studies involving a series of cycle. At each time, we increase our understanding and improvement of the process. The main issue is that business processes are not static but dynamic and evolve over time.

Relationship to the scientific method Stated hypothesis  experiment  observation confirm or infirm the hypothesis Two analytics paradigms: Deductive reasonings -

Starts with hypothesis Make an experiment and collect data to test hypothesis Explanatory and confirmatory analysis

Inductive reasonings -

Start with data Analyze existing data to generate and idea Exploratory and predictive analysis

We need both methods iteratively in order to use the scientific method at its full potential.

Relationship to the Plan-Do-Check-Act cycle PCDA cycle = Deming cycle = Shewhart cycle

Comparison with statistical method: Plan -

Set the project goals and expectation Thinking about which data we need based on our subject matter knowledge

Do -

Implement the plan Conduct the experiment/survey in order to gather data

Check -

Assess major problems and root causes Analyzing the date

-

Fix the obvious and decide how to approach the problem Interpreting the data according to our hypothesis The results determine if we need to repeat the cycle

Act

Synergy between data and subject matter knowledge Understanding the subject is important to collect the right data and no waste resources to gather useless data.

Statistical engineering Statistical thinking is not a collection of statistical methods, but rather a way of implementing improvements and use statistical tools. It provides an overall improvement strategy. In order to think before making calculations, we need a framework to helps us apply the methods in a consistent way. Statistical engineering provides frameworks and tactics to takle a problem.

6. Common cause vs. Special cause S

7. Control chart

Phase I and Phase II...