Glossary

The words you meet on the Process Intelligence screens, in alphabetical order, each as this product uses it. Where a page covers the term in full, the entry links to it.

Activity - what happened in an event, such as "Application received" or "Contract sent". When you load a flat event log, you say which column holds it. See Loading Data.

AI job - a request the AI assistant works out from the project's data, kept with its brief and every run it has made, so you can run it again when the data changes. Every report is made by a job. See AI Jobs.

API key - a key that lets a tool or an agent outside mindzieStudio reach Process Intelligence. A tenant key reaches every project you can open; a project key reaches one project only. See Driving a Project from Your Own Agent.

Attribute - any other value carried in the log, such as an amount, a channel or a team. An attribute is a case attribute when its value never changes inside a case, and an event attribute otherwise; this is measured from the data, not set by you. See Loading Data.

Automation candidate - a step a machine could take over, scored from how often it runs, whether the same thing always follows it and whether it takes the same time each run. Each gets a verdict: strong, partial, judgement or blocked. See Automations.

Bottleneck - a step that cases wait a long time for. The Bottleneck register on Performance gathers the waiting by the step that was waited for. See Performance.

BPMN model - a process diagram in the BPMN 2.0 standard. Conformance checks cases against one: either the diagram drawn from your own log, or one you upload. See Conformance.

Case - the thing whose journey you follow through the process: an order, an invoice, a patient. Every figure that counts cases depends on what a case is. See Choosing What a Case Is.

Case notion - the choice of what a case is, when the data holds several kinds of objects. You make it, and you can change it later without loading the data again. See Choosing What a Case Is.

Cases in view - the cases a screen is counting right now: all of them, or only those a saved filter keeps. Every share on the screen is taken over them.

Clock time - a duration counted in every hour of the day, weekends included. The alternative is working time, set by the working calendar. See Working Calendar.

Cohort - the set of cases a figure is measured over. The screens show it as cases in view, and say which cases they are and how many, so two figures can be compared only when they share the same cohort. See Performance.

Conformance - whether cases run the way the process is meant to, checked against a process diagram and against your rules. See Conformance.

Conformance rate - the share of cases in view that follow the process diagram. See Conformance.

Cost - what one counted thing costs: a late day, a case that broke a rule, a repeated step. Costs are kept for a costing report to multiply; no analysis screen turns counts into money. See Costs.

Cycle time - the time from a case's first event to its last. A case with a single event, or with its first and last event at the same moment, has no cycle time and is left out of duration figures. See Performance.

Deviation - a way a case leaves the process diagram or breaks a rule, such as a step done out of order or not recorded at all. See Conformance.

Enrichment - a change to the event log that every screen reads: a new column, changed values or activity names, or cases and events removed or added. Enrichments run in order, each on what the one before it left. See Enrichments.

Event - one thing that happened, at one moment: an activity and a timestamp, and often a resource and other attributes. In a flat event log, one row is one event.

Event log - the events of a process, grouped into cases. A project can keep more than one event log over the same data, each with a different answer to what a case is; one of them is in use, and every screen reads it. See Files, OCEL and Event Logs.

Filter - a named set of cases, such as "only the applications that were completed". Saving a filter changes nothing in the data; a screen, a rule or an enrichment uses it when it asks for it. See Filters.

Fitness - see Trace fitness.

Flat event log - an event log with one row per event and a case id, an activity and a timestamp on each row, usually a CSV or Excel file. See Loading Data.

Handoff - the move from one step straight to the next, and the wait between them. Performance measures the wait on every handoff and shows which ones hold the most time. See Performance.

Held-out score - how a prediction did on cases it never saw while it was trained. It is the score to trust, read beside the score for doing nothing. See Predictions.

Needs an answer - a project state: the data holds several kinds of objects, and the project will not guess which one a case should follow. You choose, and the project finishes building. See The Projects Page.

Object - one thing the data is about - one order, one delivery, one invoice. Events involve objects, and a case follows one of them.

Object type - a kind of object, such as order, delivery or invoice. When the data holds more than one object type, you choose which one a case follows. See Choosing What a Case Is.

OCEL (object-centric event log) - one log of events, the objects they involve, and the relations between them, keeping every object type instead of one. A project builds its OCEL from the data you load; OCEL is also the name of the file format you can load it from. See Files, OCEL and Event Logs.

Prediction - a model trained on your finished cases that scores running cases, for example which are likely to end late, with a verdict on how far it can be trusted. See Predictions.

Process map - a drawing of the process as your data shows it: the activities, and how cases move between them. See Process Maps.

Project - one piece of work in Process Intelligence. A project holds one set of data and everything made from it: definitions, reports and jobs. You start it to work in it and stop it when you are done. See The Projects Page.

Published - a report that is done and can be read and shared. A published report is a frozen copy; running it again adds a new version beside it. See Reports.

Report - a write-up of what you asked about, made from the project's data by the AI assistant, which says how it was made. See Reports.

Resource - who or what did an event: a person, a team or a system.

Rework - a step done more than once in the same case. The rework rate is the share of cases that repeat at least one step. See Performance.

Rule - a statement of what the process is supposed to do, such as "identity is verified before the contract is sent". For every case a rule answers met, broke or does not apply. See Rules.

Simulation - an answer to a "what if?" question, worked out by applying a change to the cases the project already holds. Every result is the most a change could win, not a promise. See Simulation.

Task mining - looking at how people do their work on their own computers, from recordings made by the mindzie Task Mining Agent. See Task Mining.

Trace fitness - how close a case comes to the process diagram, from 0 to 1, where 1 is a perfect fit. On Conformance it is the median of the deviating cases. See Conformance.

Variant - the sequence of activities a case went through. Cases that took the same path share a variant, so the variants show how many different ways the process really runs. See Variants.

Version - a report has one version for each time it was published; an earlier version stays readable. The data has versions too: each build of the log is one, and a report says which it read.

Working calendar - when the business is open, the dates it is closed, and up to what moment an unfinished case is measured. A duration measured against it is in working time. See Working Calendar.

Working time - a duration counted only in the hours the working calendar says the business is open. See Working Calendar.