Learning Analytics
Learning analytics comes in four forms: Descriptive, Diagnostic, Predictive and Prescriptive.
Next, we will examine each one, considering its purpose, common methods, and the role it plays in informed decision-making.
Descriptive Analytics
Descriptive analytics summarizes past activities to show what happened by focusing on key metrics and trends.
Diagnostic Analytics
Diagnostic analytics identifies likely contributing factors and candidate root causes behind past outcomes.
Predictive Analytics
Predictive analytics uses historical data patterns and advanced algorithms to forecast what is likely to happen next.
Prescriptive Analytics
Prescriptive analytics recommends actions to improve outcomes based on analytics and design judgement.
Descriptive → Diagnostic → Predictive → Prescriptive
The sections that follow examine the methods, techniques, tools, and reporting formats commonly associated with each type of analytics.
Descriptive Analytics
Descriptive analytics focuses on understanding what happened by summarizing past activity and outcomes, providing the foundation for all other types of analytics by revealing patterns and trends in the data. It uses basic measures, tabular and visual summaries, simple comparisons, and rankings to summarize data.
Basic Measures
Mean, median, standard deviation, and percentiles are examples of basic measures.
Tabular & Visual Summaries
Frequency tables, and histograms are examples of tabular and visual summaries.
Predictive analytics uses historical data patterns and advanced algorithms to forecast what is likely to happen next.
Simple Comparisons & Rankings
Month over month, year over year, and pre and post are examples of simple comparisons, and Top-N is an example of ranking.
Dashboards, Scorecards, & Pivot tables
Interactive dashboards, key performance indicator scorecards, and pivot tables, let users filter, slice, and drill-down to monitor KPI’s, compare segments, and spot trends and outliers.
Descriptive Analytics → Diagnostic Analytics
Descriptive analytics shows what happened. It highlights patterns in engagement, completion, and performance using summaries, trends, and comparisons.
Next, we investigate contributing factors and relationships to explain what’s driving the results. This is the role of diagnostic analytics.
Diagnostic Analytics
After understanding what happened with descriptive analytics, the next step is to explain why it happened.
Diagnostic analytics uses exploratory and statistical methods to identify likely contributing factors and candidate root causes behind past outcomes.
Diagnostic Analytics Techniques
Techniques include drill-down analyses, anomaly detection, and correlation and regression analysis to test hypotheses about why metrics changed.
Diagnostic Views
Business-intelligence platforms such as Microsoft Power BI, Tableau, and Qlik Sense provide diagnostic views.
Key Influencers and Decomposition Tree Charts
For example, key Influencers and decomposition tree charts surface key drivers and decomposition paths, highlighting likely causes and their relative impact on the metric change.
Diagnostic analytics → Predictive Analytics
Diagnostic analytics helps explain past outcomes by uncovering likely causes and relationships. Next, we turn to predictive analytics, which uses these patterns to forecast what is likely to happen next.
Predictive Analytics
After explaining why outcomes changed, the next step is to anticipate what is likely to happen next. Predictive analytics involves statistical modeling and machine learning methods to predict future outcomes.
Train Models & Predict Future Outcomes
To train models on historical data and predict future outcomes, software tools and programming languages such as Python and R use time-series analysis and predictive modeling methods, including regression, classification, and forecasting.
Predictive Outputs
Then, Business-Intelligence and Machine Learning platforms score new data, produce forecasts, and flag risks for decisions.
Predictive Analytics → Prescriptive Analytics
Predictive analytics estimates what is likely to happen next. Next, we turn to prescriptive analytics, which uses these predictions to recommend actions to improve outcomes.