ClearInsights Max: Change Analysis Overview

Change analysis helps organizations compare two data points in a visualization and identify the key drivers behind the changes. This article explains how Change Analysis works and how to use it.

In this article:
Change Analysis Overview
How to run a Change Analysis
How Change Analysis calculates outliers
Change Analysis in ClearInsights Agent
FAQ
Additional Resources

Change Analysis Overview

Change analysis compares data points using simple or complex measures. The supported aggregates include:

1. Sum
2. Count
3. Sum over sum (provides "what if" percentage insights)
4. Average (provides "what if" percentage insights)
5. Other functions that use a "versus" analysis to show absolute changes grouped by different attributes

---

Change Analysis Limitations

• Queries using growth of or versus keywords are not supported.
• Some complex formulas built on group_* functions are not supported (e.g., sum(group_*) or sum(x)/group_*). Simple formulas using group_* are supported.
• When slicing measures by color, you can only compare values of the same type.
• Runtime filters applied to hidden columns are not applied in Change Analysis.

---

Change Analysis Summary Tab

The summary tab provides a snapshot of all factors and attributes that have changed in a chart. From here, you can:

• View details to see changes by attribute
• Customize attributes to choose which columns to analyze

Note

For charts created with ratio keywords like average, the summary tab does not appear. You will see the previous interface, which analyzes one attribute at a time.

---

How to run a Change Analysis

Contextual change analysis

Contextual change analysis measures the difference between two points for a measure and slices them by another attribute to identify which attributes explain the difference.

ClearInsights supports contextual change analysis for area, line, column, bar, line-column, donut, and KPI charts. 

Contextual change analysis

Steps to run contextual change analysis:

1. Hold shift and click on two points. 
2. Right-click and select “Run change analysis.”

   • Note

     For KPI charts, click the percent change label or select two points from the sparkline. Charts without a sparkline do not support change analysis.

3. A modal displays the top five relevant columns based on past activity.
4. To analyze additional columns, click Customize Attributes and select the desired columns.
5. Review each visualization for detailed comparison.

Tip

Attribute values can account for more than 100% of total change when some values contribute to trends opposite the overall change.

Additional actions:

• Add visualizations to a Liveboard
• Download data as CSV
• Edit or copy the visualization

Keep in Mind

1. ClearInsights automatically highlights values with the highest percentage change or largest contribution to total change. Insights are not available for unique count or ratio measures.
2. Personalized column selections are remembered and can be applied to all users on a Liveboard.

How contextual change analysis is calculated

Contextual change analysis handles different measure types in three categories:

Type 1: Simply decomposable

• Measures like SUM or COUNT
• Changes are compared between timestamps, and thresholds are calculated from the top ten absolute changes. Values outside thresholds are marked as outliers.

Algorithm:

• Compare measurements at two points for each attribute.
• Calculate upper and lower thresholds based on the top absolute changes.
• Any value outside thresholds is an outlier.
• Stop iterating when the combined contribution of a measure at either timestamp exceeds 50%.
• The largest negative change sets the lower threshold, and the smallest positive change sets the upper threshold.

Type 2: Ratio of simply decomposable

• Measures like AVERAGE or SUM/SUM
• Hypothetical percentage changes are calculated to identify attributes that contribute most to overall change.

Algorithm:

• Compute a hypothetical percentage change for each attribute:
  • The overall percentage change if this attribute had remained constant.
  • Smaller hypothetical changes indicate a high contribution to overall change

Type 3: Unknowns

• Measures like UNIQUE COUNT or SUM * SUM
• Changes are treated as a normal distribution. Outliers are identified using Z-scores, with thresholds based on attribute cardinality.

Algorithm:

• Compare measurements at two points for each attribute
• Treat changes as a normal distribution
• Calculate upper and lower thresholds using Z-score: z = x-μ/σ
• Where:
• x = data value
• μ = mean
• σ = standard deviation
• Thresholds: μ ± N * σ, where N depends on attribute cardinality

Z-score threshold table:

Attribute Cardinality	Value of N
<= 100	2.0
500	2.69
2000	3.301
10000	4.0
50000	4.69
>= 100000	5.0

---

Iterative Change Analysis

Iterative Change Analysis lets you drill deeper by selecting an attribute value from your results to recalculate changes. You can analyze more specific drivers and refine insights.

Iterative Change Analysis

Steps to run iterative change analysis:

1. Select two points on a chart and right-click to select Run change analysis.
2. Right-click an attribute value and select Analyze.
3. ClearInsights adds the attribute as a filter and recalculates the analysis.

Additional features:

• Undo: Revert the last action.
• Redo: Repeat the last undone action.
• Reset: Revert all changes in the current session.
• AI-driven suggestions highlight attributes for effective analysis.

Note

Iterative change analysis does not work on visualizations created with custom calendars.

---

How Change Analysis Calculates Outliers

Outliers are detected differently depending on the measure type:

• Type 1: Based on absolute changes and thresholds.
• Type 2: Based on hypothetical percentage contributions.
• Type 3: Based on Z-scores using standard deviations.

This approach ensures significant changes are highlighted, providing actionable insights across all measure types.

---

Change Analysis in ClearInsights Agent

With ClearInsights Max, you can now utilize the full power of our Change Analysis to drill deeper into candidate behavior with the ClearInsights Agent by simply asking 'why.' The Agent will run a Change Analysis based on the request and return:

• Receive an explanation of how the analysis was performed.
• See charts highlighting key contributors to change.
• Review a summary of significant changes.

For a quick demo, see: Change Analysis from ClearInsights Agent.

---

FAQ

Can Change Analysis be scheduled to run automatically on a recurring basis?

No, you can’t schedule Change Analysis yet. You can share Liveboards on a schedule. Learn more in ClearInsights Share Analytics.

Does Change Analysis performance vary with large datasets?

You may notice slower performance with large datasets. We are actively working to improve this experience.

Can you customize how outliers are highlighted in visualizations?

No, you can’t customize how outliers are highlighted at this time.

How does Change Analysis handle missing or incomplete data?

You can choose which columns to include when running Change Analysis. This helps you exclude fields with missing data that could affect results. You can also use filters to refine your selection.

---

Additional Resources

ClearInsights Max: Answer Builder
ClearInsights Max: Chart Types Available
ClearInsights: KPI Chart Overview
ClearInsights Max: Performance Getting Started
ClearInsights Max: Recruiting Getting Started