🔍

Activity Order Classification

Enrichments

Overview

The Activity Order Classification enrichment automatically analyzes the timestamps in your event log to identify cases where the sequence of activities cannot be determined with certainty due to timestamp limitations. This data quality enrichment is essential for process mining accuracy, as uncertain activity ordering can lead to incorrect process models, misleading performance metrics, and unreliable conformance checking results.

Many source systems record only dates without time components, or multiple activities share the exact same timestamp due to bulk data imports, batch processing, or timestamp granularity limitations. When events within a case have identical timestamps (either same date or same datetime), the actual sequence in which these activities occurred becomes ambiguous. This enrichment automatically detects and categorizes these uncertainty patterns, creating attributes that allow you to assess the reliability of your process discovery results and identify cases where ordering assumptions may be incorrect.

The enrichment requires no configuration and performs comprehensive timestamp analysis across both date-level and time-level precision, categorizing uncertainty patterns as "SameDay" (date recorded but no time component), "SameTime" (identical datetime values), or "SameDayAndTime" (case contains both patterns). This enables you to understand the scope and nature of timestamp uncertainty in your event log and make informed decisions about data quality requirements and process analysis reliability.

Common Uses

  • Assess data quality before performing process discovery or conformance checking
  • Identify cases where activity sequences are ambiguous due to timestamp limitations
  • Detect bulk-loaded or batch-processed events that share identical timestamps
  • Evaluate whether source system timestamp granularity is sufficient for process analysis
  • Flag cases where manual ordering assumptions may be required for accurate analysis
  • Measure the prevalence of timestamp uncertainty across your entire event log
  • Filter out low-quality cases where uncertain ordering would compromise analysis results

Settings

This enrichment requires no configuration. It automatically analyzes all timestamps in your event log and creates comprehensive attributes that categorize timestamp uncertainty patterns at both the event level and case level. Simply add this enrichment to your workflow to begin analyzing timestamp quality.

Examples

Example 1: Healthcare Patient Journey Analysis

Scenario: A hospital is analyzing patient flow through their emergency department but discovers that many activities on the same day have no recorded time component, making it impossible to determine the actual sequence of treatments and examinations.

Settings: No configuration required - the enrichment automatically detects timestamp uncertainty.

Output: The enrichment creates the following attributes:

Event-level attributes:

  • OrderUncertainty: TRUE for events where ordering cannot be determined with certainty
  • OrderUncertaintyCategory: "SameDay" for events that share a date with other events but have no time component

Case-level attributes:

  • UncertainEventOrder: TRUE (this case has uncertain ordering)
  • UncertainEventOrderCount: 8 (eight events in this case have uncertain ordering)
  • UncertainEventOrderCategory: "SameDay"

For a patient case with events recorded as:

  • 2024-03-15 00:00:00 - Patient Registration
  • 2024-03-15 00:00:00 - Triage Assessment
  • 2024-03-15 00:00:00 - Vital Signs Check
  • 2024-03-15 00:00:00 - Physician Consultation
  • 2024-03-15 14:30:00 - Lab Results Received
  • 2024-03-15 14:30:00 - Treatment Decision
  • 2024-03-15 14:30:00 - Medication Administered
  • 2024-03-15 18:00:00 - Patient Discharge

The first four events (all at 00:00:00) are marked as "SameDay" uncertainty because they share a date but the time component is missing. The three events at 14:30:00 would be marked as "SameTime" uncertainty because they share an identical datetime. This case would be categorized as "SameDayAndTime" because it exhibits both patterns.

Insights: The hospital discovers that 67% of emergency department cases have uncertain event ordering due to missing time components in their registration system. This reveals a critical data quality issue that must be addressed before accurate process discovery can be performed. They can now filter cases to analyze only those with complete timestamps or work with IT to enhance timestamp granularity in their source systems.

Example 2: Financial Transaction Processing

Scenario: A bank is analyzing credit card transaction approval processes but notices that batch-processed transactions often share identical timestamps, making it impossible to determine the true sequence of fraud checks, authorization steps, and approval decisions.

Settings: No configuration required.

Output: For a transaction case processed in a batch system:

  • 2024-10-15 02:15:33 - Transaction Received
  • 2024-10-15 02:15:33 - Fraud Risk Assessment
  • 2024-10-15 02:15:33 - Credit Limit Check
  • 2024-10-15 02:15:33 - Merchant Verification
  • 2024-10-15 02:15:33 - Transaction Approved
  • 2024-10-15 02:15:34 - Confirmation Sent

Event attributes:

  • First five events: OrderUncertainty = TRUE, OrderUncertaintyCategory = "SameTime"
  • Last event: OrderUncertainty = FALSE

Case attributes:

  • UncertainEventOrder: TRUE
  • UncertainEventOrderCount: 5
  • UncertainEventOrderCategory: "SameTime"

Insights: The bank identifies that all batch-processed transactions (approximately 40% of their daily volume) have uncertain ordering for critical fraud and credit checks. This revelation prompts them to investigate whether their batch processing system maintains an internal sequence number that could be used to establish true ordering, or whether they need to enhance timestamp precision in their transaction logging system.

Example 3: Manufacturing Production Line Analysis

Scenario: A manufacturing company is analyzing production workflows but discovers that quality control checkpoints are recorded with date-only timestamps, while machine operations have precise timestamps, creating mixed uncertainty patterns.

Settings: No configuration required.

Output: For a production case:

  • 2024-10-20 08:15:22 - Raw Material Loaded
  • 2024-10-20 08:18:45 - Machining Started
  • 2024-10-20 00:00:00 - Visual Inspection
  • 2024-10-20 08:45:12 - Machining Completed
  • 2024-10-20 00:00:00 - Dimension Check
  • 2024-10-20 00:00:00 - Quality Approval
  • 2024-10-20 09:10:30 - Packaging Started

Event attributes:

  • Visual Inspection, Dimension Check, Quality Approval: OrderUncertainty = TRUE, OrderUncertaintyCategory = "SameDay"
  • Other events: OrderUncertainty = FALSE

Case attributes:

  • UncertainEventOrder: TRUE
  • UncertainEventOrderCount: 3
  • UncertainEventOrderCategory: "SameDay"

Insights: The company discovers that their manual quality control system records only dates while automated machine operations capture precise timestamps. This mixed precision means they cannot determine whether quality checks occurred in the documented sequence or whether dimension checks sometimes happened before visual inspections. They can now prioritize upgrading their quality control logging system or adjust their process analysis to account for this uncertainty.

Example 4: E-commerce Order Fulfillment

Scenario: An online retailer is analyzing order processing workflows but notices that warehouse management system events often have identical timestamps due to rapid scanning operations that exceed the system's one-second timestamp precision.

Settings: No configuration required.

Output: For an order with rapid fulfillment:

  • 2024-10-21 10:23:45 - Order Received
  • 2024-10-21 10:24:18 - Inventory Allocated
  • 2024-10-21 10:24:18 - Pick List Generated
  • 2024-10-21 10:24:18 - Items Picked
  • 2024-10-21 10:24:18 - Quality Verified
  • 2024-10-21 10:24:18 - Packing Completed
  • 2024-10-21 10:25:03 - Shipping Label Created

Event attributes:

  • Five events at 10:24:18: OrderUncertainty = TRUE, OrderUncertaintyCategory = "SameTime"

Case attributes:

  • UncertainEventOrder: TRUE
  • UncertainEventOrderCount: 5
  • UncertainEventOrderCategory: "SameTime"

Insights: The retailer discovers that their warehouse operations are so efficient that multiple steps occur within the same one-second window, but their system timestamp precision is insufficient to capture the true sequence. They find that 25% of orders have uncertain ordering for warehouse activities. This prompts them to consider adding sub-second timestamp precision to their warehouse management system or implementing sequence numbers for same-second events.

Example 5: IT Service Desk Ticket Resolution

Scenario: An IT department is analyzing support ticket resolution processes but discovers that bulk status updates and automated system actions often share timestamps, creating uncertainty about the actual sequence of troubleshooting steps.

Settings: No configuration required.

Output: For a support ticket case:

  • 2024-10-18 09:15:00 - Ticket Created
  • 2024-10-18 09:15:00 - Auto-Assigned to Team
  • 2024-10-18 09:15:00 - Priority Set
  • 2024-10-18 09:15:00 - SLA Timer Started
  • 2024-10-18 10:30:22 - Engineer Assigned
  • 2024-10-18 00:00:00 - Initial Investigation
  • 2024-10-18 00:00:00 - Root Cause Identified
  • 2024-10-18 00:00:00 - Resolution Applied
  • 2024-10-18 14:45:10 - Ticket Closed

Event attributes:

  • First four events: OrderUncertainty = TRUE, OrderUncertaintyCategory = "SameTime"
  • Middle three events: OrderUncertainty = TRUE, OrderUncertaintyCategory = "SameDay"

Case attributes:

  • UncertainEventOrder: TRUE
  • UncertainEventOrderCount: 7
  • UncertainEventOrderCategory: "SameDayAndTime"

Insights: The IT department discovers that automated ticket creation steps all share the same timestamp, and manual investigation activities are logged with date-only precision. This mixed uncertainty pattern affects 55% of tickets and reveals that their process mining results may show incorrect activity sequences. They can now work with their IT service management system vendor to improve timestamp granularity and establish more reliable process discovery results.

Output

The Activity Order Classification enrichment creates comprehensive attributes at both the event level and case level to enable detailed analysis of timestamp uncertainty in your process data.

Event-Level Attributes:

OrderUncertainty (Boolean): Indicates whether this specific event has uncertain ordering relative to other events in the same case. Set to TRUE when the event shares an identical timestamp (either date only or complete datetime) with at least one other event in the case, making the sequence ambiguous. Set to FALSE when the event has a unique timestamp within the case.

OrderUncertaintyCategory (Text): Categorizes the type of timestamp uncertainty for this event:

  • "SameDay": The event shares a date with other events but has no time component (timestamp ends with 00:00:00), indicating date-only precision in the source system
  • "SameTime": The event has an identical datetime (including time component) with other events, indicating either simultaneous execution or insufficient timestamp granularity
  • "SameDayAndTime": The event exhibits both patterns (initially flagged as SameDay, then also found to match SameTime criteria)

Case-Level Attributes:

UncertainEventOrder (Boolean): Indicates whether this case contains any events with uncertain ordering. Set to TRUE if at least one event in the case has ambiguous ordering due to timestamp duplication. Set to FALSE only when all events in the case have unique timestamps and ordering can be determined with certainty.

UncertainEventOrderCount (Integer): The total number of events within this case that have uncertain ordering. This count helps you assess the severity of timestamp uncertainty - a case with two uncertain events is less problematic than one with dozens of events sharing the same timestamp.

UncertainEventOrderCategory (Text): Summarizes the timestamp uncertainty pattern for the entire case:

  • "SameDay": Case contains only date-level uncertainty (some events share dates but have no time component)
  • "SameTime": Case contains only time-level uncertainty (some events share identical datetime values)
  • "SameDayAndTime": Case contains both patterns of uncertainty

Data Type Details:

  • Boolean attributes use TRUE/FALSE values and can be used in filters with "equals TRUE" or "equals FALSE" conditions
  • Integer attributes can be used in range filters and calculations to measure uncertainty prevalence
  • Text attributes can be grouped and filtered to analyze different uncertainty patterns separately

Usage in Analysis: These attributes enable you to filter your dataset to exclude cases with uncertain ordering, create metrics showing the percentage of cases affected by timestamp uncertainty, identify which source systems or processes have the worst timestamp quality, and prioritize data quality improvements based on the impact on your process mining results. The attributes integrate seamlessly with conformance checking, process discovery, and performance analysis features in mindzieStudio.

See Also

  • Allowed Case End Activities - Conformance enrichment that requires reliable activity ordering
  • Allowed Case Start Activities - Conformance enrichment affected by uncertain first-event timestamps
  • Duration Between Two Activities - Performance enrichment that produces unreliable results when activity order is uncertain
  • Freeze Log Time - Data cleanup enrichment that can normalize timestamps to improve consistency

This documentation is part of the mindzie Studio process mining platform.

An error has occurred. This application may no longer respond until reloaded. Reload ??