1.3 Types of Analyses
So far, we have covered the data life cycle, provided a conceptualization of data quality, and explained how data becomes wisdom. But a key step in converting data to wisdom are the various data analyses that guide that wisdom. There are many, many types of business data analytics that need to be performed in companies of all types and sizes. How do we make sense of them all? That is what we will tackle next.
The image above provides a framework for understanding each of the analyses you will learn in a business data analytics program as descriptive, diagnostic, predictive, or prescriptive. These types of analyses vary both in terms of the degree of complexity and automation allowed as well as the value or competitive advantage they provide. Generally speaking (with some exceptions), analyses that provide more value become more complex and harder (but not impossible) to automate. Let's dive into these types one at a time. For each type, you can refer to the figure above to see what tools are commonly used.
Descriptive Analytics
Descriptive analytics are intended to reveal "what happened" in the past. They are summaries of historical data that identify basic patterns and trends. They provides insights into past events and helps organizations understand what has happened in their business operations. For example, a line chart of total sales over the past year, a bar chart of total sales separated by product line, and a pie chart of total sales by customer segment each describe what has happened over the last year in terms of sales. Common techniques used for descriptive analyses include:
Data Aggregation: Combining data from multiple sources to provide a comprehensive view.
Data Visualization: Using charts, graphs, and dashboards to represent data visually, making it easier to interpret.
Reporting: Generating regular reports that summarize key metrics and performance indicators.
Statistical Analysis: Applying basic statistical methods (mean, median, mode) to summarize data.
Diagnostic Analyses
Diagnostic analyses help us understand "why did it happen?". They go a step further than descriptive analytics by exploring the reasons behind past outcomes. It aims to identify the root causes of observed patterns and trends.
Drill-Down Analysis: Examining data at a more granular level to understand underlying factors.
Correlation Analysis: Identifying relationships between different variables to understand how they influence each other.
Data Mining: Extracting useful information from large datasets using various techniques such as clustering and classification.
Root Cause Analysis: Investigating the primary causes of specific problems or events.
Predictive Analyses
Predictive analyses helps us know "what will happen next?" They use historical data and statistical models to forecast future outcomes. It helps organizations anticipate trends, behaviors, and events.
Regression and Classification Analysis: Modeling the relationship between a dependent variable and one or more independent variables to make predictions.
Clustering: Identifying groups of similar objects to be studied and labeled for further understanding.
Time Series Analysis: Analyzing data points collected or recorded at specific time intervals to forecast future values.
Machine Learning: Using algorithms to learn from data and make predictions. Techniques include decision trees, random forests, and neural networks.
Prescriptive Analyses
Prescriptive analyses help us decide "what should we do?". They go beyond prediction by recommending specific actions to achieve desired outcomes. It provides actionable insights and suggestions for decision-making.
Optimization: Finding the best solution from a set of feasible options, often using linear programming or other mathematical techniques.
Decision Trees: Modeling decisions and their possible consequences to identify the best course of action.
Simulation: Evaluating the impact of different strategies under various scenarios to recommend the optimal approach.
Heuristics: Applying rule-based methods to make decisions in complex situations where an optimal solution may not be feasible.