README.md

metadata

title: ProcessScope | Process Mining Intelligence Lab
emoji: 🔍
colorFrom: green
colorTo: blue
sdk: static
app_file: index.html
pinned: false
license: mit
short_description: Local-first process mining and SLA audits
tags:
  - process-mining
  - operations-research
  - business-intelligence
  - data-visualization
  - workflow-analysis
  - javascript

ProcessScope | Process Mining Intelligence Lab

Project 40 of Hugdrigo's research and engineering portfolio.

ProcessScope is a standalone, browser-based process-intelligence application for analyzing event logs, exploring the observed directly-follows process network, inspecting journey variants, identifying elapsed-time bottlenecks, auditing service-level agreement (SLA) performance, and exporting reproducible reports. No backend, external APIs, CDN, tokens, or paid compute are necessary.

Live application

After the Space is published, open the App tab. Start with the synthetic Order fulfillment scenario, then switch to Customer support or Permit review.

Main capabilities

Workspace	What it does
Overview	Cases, event counts, lead-time percentiles, SLA breach rates, completion rate and time-distribution chart
Process discovery	Directly-follows network with node selection, transition frequency, and minimum-frequency threshold
Journey variants	Ranked exact activity sequences, median lead times, completion and late cases by route
Bottleneck explorer	Mean, median and 90th-percentile elapsed time between consecutive events
SLA & segments	Editable SLA and audit steps; comparisons across teams, channels and daily cohorts
Data & audit exports	Strict CSV import, case explorer, synthetic CSV template, JSON / CSV / SVG / Markdown exports

Event-log CSV schema

Required columns: case_id, activity, timestamp. Optional columns: team, channel, cost.

case_id,activity,timestamp,team,channel,cost
A-001,Received,2026-04-01T08:00:00Z,Ops,Web,12
A-001,Validated,2026-04-01T10:00:00Z,Ops,Web,8
A-001,Completed,2026-04-01T11:00:00Z,Ops,Web,5
A-002,Received,2026-04-01T09:00:00Z,Ops,Partner,10
A-002,Validated,2026-04-01T13:00:00Z,Ops,Partner,6
A-002,Completed,2026-04-01T14:00:00Z,Ops,Partner,8
A-003,Received,2026-04-01T10:00:00Z,Ops,Web,7
A-003,Validated,2026-04-01T12:00:00Z,Ops,Web,7
A-003,Completed,2026-04-01T14:00:00Z,Ops,Web,7

Time zone is mandatory. Use ISO 8601, for example 2026-04-01T10:00:00Z or a numeric offset such as 2026-04-01T10:00:00-04:00. Events within a case are sorted by parsed timestamps; in a tie, their input-row order determines sequence. Each case's team and channel are attributed from its first recorded event. Cost units are user-defined and are not assumed to be a specific currency.

Import limits: at most 3 MB input size, 12,000 events, 5,000 unique cases, 45 unique activities. The application validates headers, quotations, date/time, nonnegative costs, and row structure before changing the currently loaded dataset. Three or more cases are required for user-imported logs.

Process mining methodology

• Lead time: hours between a case's first and last logged events, expressed as elapsed calendar time, not business hours.
• SLA breach: lead time is strictly greater than the selected hour threshold. Unfinished cases are included under this time-span definition.
• Completion: the last recorded activity equals the configured completion activity.
• Required-step audit: case lacks a named activity; it is only a single rule, not full conformance checking.
• Rework signal: adjacent repeated activities. It does not detect all types of rework or backtracking.
• Variant: exact ordered list of activity labels, with repetitions retained.
• Directly-follows transition: consecutive events in the same case; its wait is the observed timestamp difference.
• Bottleneck ranking: transition-specific mean elapsed waiting hours, calculated over observed transitions. High wait is not proof of high work effort or root cause.
• Temporal comparison: cases ordered by their first event; first and second halves are compared descriptively. It is not a causal estimate.
• Cohort reporting: cases grouped by first-event calendar date (UTC, after parsing timezone offsets).

Privacy and limitations

All imported data are processed locally in the browser by the static application; it does not send them to a server. However, never commit sensitive CSVs to a public Space, and take care when sharing exported reports. The provided examples are entirely synthetic and do not reflect real businesses or government services. Conclusions depend on log coverage, event definitions, timestamps, and data quality. This is an exploratory research demo, not a certified process-audit or compliance platform.

Reproducibility & tests

The separate source/test bundle includes:

• A dependency-free JavaScript analytics core and application UI.
• 202 core assertions for calculations, seeded scenarios, parser failures, import limits, exports and independently checked synthetic cases.
• 132 interactive Chromium checks for navigation, filtering, state changes, CSV imports, exports, user interface and responsive behavior.
• Desktop and mobile screenshots.

These tests were executed locally. A final deployment check should be performed in the Hugging Face App tab after publication.

Deployment

1. Create a new Hugging Face Space called processscope-process-mining-lab.
2. Choose Static, Blank, Public, and MIT.
3. Upload only the unpacked README.md, index.html, and LICENSE files, at the Space root.
4. Commit to main and open App.

Technology

HTML5, CSS3, vanilla JavaScript, SVG, browser FileReader and Blob APIs; no third-party runtime dependencies.

License

MIT. See LICENSE.