🔍 GermainUX anomaly detection overview
GermainUX includes several complementary mechanisms for detecting unusual behavior and finding important issues hidden within large volumes of data.
These capabilities help teams identify the “needle in the haystack,” including:
|
Example |
|---|
|
A new error among millions of known errors |
|
A record or business object that changed unexpectedly |
|
A technology issue affecting users or business processes |
|
An abnormal increase or decrease in a KPI |
|
A pattern that deviates from its normal business period |
No single mechanism detects every type of anomaly. The most effective configuration often combines categorization, change tracking, error analysis and statistical insight.
⚙️ Detection Mechanisms
|
Mechanism |
Primary purpose |
Example |
|---|---|---|
|
Categorization |
Distinguish new or unusual issues from previously known patterns. |
Identify a new crash signature among thousands of recurring crashes. |
|
Document Audit |
Track changes to records, documents and business data over time. |
Detect an unexpected status, owner or value change. |
|
Error Analysis |
Prioritize errors according to user, workflow and business impact. |
Surface an error blocking checkout while suppressing irrelevant background noise. |
|
Smart Insights |
Detect statistically meaningful increases, decreases and deviations from expected behavior. |
Identify an abnormal error spike relative to the same business period. |
🏷️ Categorization
Categorization groups similar events and distinguishes new categories from known ones.
It is particularly useful for high-volume technical data such as:
|
Data type |
|---|
|
Errors and exceptions |
|
Application crashes |
|
Log messages |
|
Browser console events |
|
Failed transactions |
For example, an application may generate millions of errors each day, most of which are already known. Categorization allows teams to focus on a newly appearing error pattern instead of reviewing every occurrence individually.
Categorization can help answer:
|
Question |
|---|
|
Is this issue new or previously known? |
|
How many distinct issue categories exist? |
|
Which categories are growing in frequency? |
|
Which users, applications or environments are affected by a new category? |
See Categorization.
📁 Document Audit
Document Audit tracks changes to records, documents and structured business data. It captures what changed, when it changed and, when available, the previous and new values.
It can help identify:
|
Issue |
|---|
|
Unexpected changes to a business record |
|
Records that stopped updating |
|
Status transitions and the time spent in each status |
|
Changes made outside the expected workflow |
|
Differences between updated and unchanged records |
|
Data quality or synchronization problems |
Examples include:
|
Example |
|---|
|
A service request remaining in the same status for too long |
|
An order changing to an unexpected value |
|
Customer or product data no longer synchronizing |
|
A business record being updated unusually often |
Once Document Audit is enabled, it can be combined with Smart Insights to identify abnormal volumes, missing updates or unusual changes over time.
🔧 Error Analysis
Error Analysis helps teams focus on errors that affect users, customer journeys, workflows or business outcomes while reducing noise from less relevant events.
It can help answer:
|
Question |
|---|
|
Is the error visible to users? |
|
How many users or sessions are affected? |
|
Which journey or workflow step is blocked? |
|
Is the error new or known? |
|
How frequently does it occur? |
|
What is its likely business impact? |
|
What happened before and after the error? |
Error Analysis can combine categorization, user-impact measures, journey context and Session Replay to prioritize the issues that matter most.
See Error Analysis.
📈 Smart Insights
Smart Insights detects meaningful changes in KPI behavior, including upward trends, downward trends, spikes and other deviations from expected values.
Unlike a fixed threshold alone, Smart Insights can evaluate behavior in the context of the relevant business period. This helps distinguish a genuine anomaly from a normal daily, weekly or seasonal pattern.
Examples include:
|
Example |
|---|
|
Error volume increasing above its expected range |
|
Conversion falling unusually during a normally stable period |
|
Process duration rising for a specific team or location |
|
A service becoming slower than its historical baseline |
|
Record updates stopping when activity is normally expected |
See Smart Insights.
🤝 Recommended Combinations
|
Investigation goal |
Recommended mechanisms |
Why |
|---|---|---|
|
Find a new error among known errors |
Categorization + Error Analysis |
Separates new patterns, then ranks them by user and business impact. |
|
Detect abnormal record changes |
Document Audit + Smart Insights |
Captures individual changes and detects unusual change frequency or direction. |
|
Identify a new crash spike |
Categorization + Smart Insights |
Distinguishes the crash signature and detects abnormal growth. |
|
Find errors blocking a journey |
Error Analysis + Categorization + Session Replay |
Prioritizes impacted users, identifies whether the error is new and shows what happened. |
|
Detect a process that stopped progressing |
Document Audit + Smart Insights |
Tracks status changes and identifies missing or delayed updates. |
|
Investigate abnormal KPI behavior |
Smart Insights + KPI Relationships |
Detects the deviation and surfaces related signals that may help explain it. |
📋 Typical Investigation Workflow
-
Detect an unusual condition through Smart Insights, categorization, Document Audit or Error Analysis.
-
Quantify its frequency, affected users and business impact.
-
Segment the issue using pivots such as application, page, process, user group, region or version.
-
Correlate the anomaly with related KPIs, events and transactions.
-
Investigate representative instances through the Analysis Dashboard.
-
Replay or trace the affected session, workflow or technology activity when available.
-
Alert or automate the appropriate response when the condition is actionable.
🔧 Configuration Guidance
-
Enable the mechanisms that match the type of anomaly you need to detect.
-
Establish sufficient historical data before relying on statistical baselines.
-
Use business-period comparisons when activity follows daily, weekly or seasonal patterns.
-
Apply categorization to high-volume events with repeated signatures.
-
Use Document Audit for records whose changes or status duration matter.
-
Include user and business-impact measures when prioritizing errors.
-
Validate detected anomalies against representative raw instances.
-
Configure alerts only for conditions that require action.
Service: Analytics
Feature Availability: