SLAs

⚙️ SLAs and Automated KPI Evaluation

GermainUX Service Level Agreements (SLAs) evaluate KPI data against defined expectations. They can monitor individual events, aggregated performance, changes from a baseline, predicted outcomes and abnormal behavior.

🎯 SLA Scope

SLAs can be applied to preconfigured or custom KPIs across:

When an SLA condition is met, GermainUX can change the KPI's health status and trigger alerts, reports or automated actions.

📋 Choose the Appropriate SLA

Aggregate, Color Percentile and Percent Change SLAs are statistical SLA types. Multi-Metric Evaluation is an SLA capability rather than a separate statistical model.

🔍 Fact-Based SLA

A Fact-Based SLA evaluates each individual data point against configured thresholds.

Examples include:

Use a Fact-Based SLA when each occurrence is important and should be evaluated immediately.

🔔 Stateful SLA Alerts

Stateful alerting is designed for ongoing conditions such as server, database or service availability. It provides a violation-and-recovery notification pattern similar to event-management systems.

When Stateful is enabled for a KPI, GermainUX remembers the last state of each monitored resource:

1. An alert is triggered when the resource enters a violating state.

2. No additional alert is sent while the same resource remains in that state.

3. A recovery notification is triggered when the resource returns to a compliant state.

This is appropriate for continuously collected measurements. It is usually not appropriate for independent events for which every occurrence should generate a separate action.

🔧 Stateful Configuration

Activation Count can be used with or without Stateful behavior. It reduces noise from isolated violations but does not produce recovery notifications by itself.

📊 Behavior Matrix

The key distinction is:

• Stateful creates the violation-and-recovery lifecycle.

• Activation Count delays or de-noises the initial violation action.

✉️ Stateful Alert Template Fields

The alert template context includes fields that can be used to produce different violation and recovery messages.

Example body template for Server Availability:

<#if context.violation>
GermainUX was not able to reach the following server: ${context.fact.system.hostname!}
<#else>
GermainUX reconnected to the following server after ${context.violationCount} failed attempts: ${context.fact.system.hostname!}
</#if>

Example subject template:

Server ${context.fact.system.name!} is <#if context.violation>not available<#else>available</#if>

Use context.violation to select the appropriate message and context.violationCount when the notification should report the number of failed attempts.

The shared <#include "SLA-SUMMARY"> template can also display context.summaryMetrics.attempts when a generic summary is preferred.

📊 Statistical SLAs

Statistical SLAs evaluate a collection of facts over a defined period rather than judging each individual occurrence independently.

🧮 Aggregate SLA

An Aggregate SLA calculates a measure for a time period and compares the result with a fixed threshold.

Examples include:

Use an Aggregate SLA when the overall period result matters more than a single event.

💚 Color Percentile SLA

A Color Percentile SLA evaluates the proportion of facts classified as GREEN during a period and compares that proportion with a threshold.

For example:

At least 95% of User Clicks must remain within the GREEN performance range.

This model is useful when a limited number of degraded experiences is acceptable but the majority must meet the expected performance level.

🔁 Percent Change SLA

A Percent Change SLA compares a current period with a previous baseline period.

It can detect:

For example:

Alert when User Error volume increases by more than 30% compared with the previous equivalent period.

🔄 In Range Operator

The In Range operator can be used with a Percent Change SLA to determine whether the result remains within an acceptable positive or negative range.

For example, a range of -30% to +30% can be treated as normal, while a result outside that interval can be treated as a significant change.

🧾 Filter Percent Change Results

A large percentage change can be misleading when both the baseline and current volumes are very small. Statistical SLAs support a Threshold Expression that can use the underlying current and baseline values to suppress low-volume noise.

Available variables include:

The following expression prevents an alert when both counts are below 20, even if the percentage change is large. It also treats values below the configured threshold as compliant:

(current < 20 && baseline < 20) || (value < threshold)

Threshold Expressions allow the SLA to consider both relative change and business volume.

🔮 Predictive SLA

A Predictive SLA uses historical data and trends to forecast a future value or outcome.

Examples include:

Predictive SLAs support proactive action before a threshold is actually violated.

💡 Smart Insights SLA

Smart Insights identifies abnormal trends and noteworthy deviations relative to comparable business periods.

Comparisons can reflect the operating context, such as:

Examples include:

Smart Insights can be enabled for active KPIs involving technology, processes, user behavior and business outcomes.

🔗 Evaluate Multiple Measures on One Fact

GermainUX can evaluate multiple measures from the same data point, each with its own threshold.

For example, a User Click can be evaluated for:

Measure	Details / Media
Failure status
Interaction duration
Memory footprint

Another use case may evaluate both Active Duration and Inactive Duration to understand how employees spend time while completing an HR workflow.

Thresholds are evaluated in order. The worst resulting health color is applied to the fact. For example, if one threshold returns ORANGE and another returns RED, the final result is RED.

This makes it possible to represent the overall health of a data point without losing the individual conditions used to evaluate it.

🚨 Alert Examples

📝 Configuration Recommendations