Find what the averages hide.

Coming soon

Find the slow steps. Inspect the handovers. Challenge a workload comparison or an approval that breaks the rule. Follow each finding to the records and definitions behind it.

The dark Reader process map connects order items, deliveries, orders and invoices through recorded activities and transitions.
Object-centric process map · Development demo

01 / Where you can improve

Nine questions you can investigate.

Nine ways to understand performance and focus your next improvement. Each connects a business question to the Reader view that answers it.

Each question, its view and the data it needs

02 / Reliable measures

Know what each measure includes.

A number without a definition is an invitation to get it wrong. Ghost Observatory keeps recorded effort, elapsed time and missing data distinct.

  1. 01

    Keep gaps in activity visible.

    Gaps in telemetry are shown as unavailable. An absent record never becomes a claim that someone was idle.

  2. 02

    Distinguish time from effort.

    Processing intervals and a person’s recorded work time measure different things. Four hours in a queue represents delay; effort needs its own evidence.

  3. 03

    Respect what each source measures.

    System-recorded and self-reported work remain separately inspectable. Their admitted work intervals also form a total clock coverage, with overlapping time counted once. Screen observations remain separate.

  4. 04

    Keep every selection grounded.

    The Reader uses the figures published by the Engine. It declines filters with no published results, keeping each selection tied to a measured scope.

See how the safeguards work.

See how the product handles missing activity, waiting time and recorded work so you can interpret each measure correctly.

How the safeguards work

See a rejected row.

The sample records three rejected rows, each with its source file, row number and reason.

Inspect the rejected rows

03 / Details

Calendars, privacy and configuration.

One SQLite file holds every result; optional pseudonymisation replaces mapped personal identifiers with stable pseudonyms when you enable ANONYMIZE: true and provide a 32-byte key; your calendars and a pinned time-zone database set the clock; and the AI, if any, is the one you connect.

How each rule works

04 / Next

Start with the published sample.

Explore 300 synthetic claims in the published sample. Follow a finding from the Reader to its source row, then see which exports can answer the same questions in your operation.

Follow a finding Explore data sources Process mining explained Tour the Reader

Choose a measure by what it answers.

Elapsed time, recorded work and event-count concentration describe different aspects of an operation. Their definitions explain which comparisons the records support.