---
title: Process Discovery
type: concept
tags: [bpm, discovery, modelling]
sources: ["[[sources/2018-dumas-fundamentals-of-bpm]]"]
created: 2026-04-13
updated: 2026-04-13
---

# Process Discovery

The act of eliciting a process model from reality — translating tacit organisational knowledge into an explicit representation ([[sources/2018-dumas-fundamentals-of-bpm]], ch. 5).

## Three intrinsic challenges (§5.1.2)
1. **Fragmented process knowledge** — no single person holds the full process.
2. **Thinking-in-cases** — domain experts describe the common path but not variants/exceptions.
3. **Analyst vs domain-expert gap** — analysts know modelling conventions; experts know the work.

## Two flavours
- **Human-driven discovery** — [[methods/process-discovery-methods]] (evidence / interview / workshop-based).
- **Automated discovery** — [[methods/process-mining-basics|process mining]] from event logs (α-algorithm and successors; ch. 11).

## Imperative vs declarative discovery
Automated discovery bifurcates by target language: **imperative** (Petri nets / BPMN via α, Inductive, Heuristics, Split Miner) vs **declarative** (DECLARE / ConDec models via Declare Miner, MINERful, DisCoveR). See [[frameworks/declare]] and [[concepts/declarative-process-modelling]]; the original tool is documented in [[sources/2008-pesic-declare-manual]].

## Method selection
Choice depends on process complexity, access to domain experts, existence of an event log, and required granularity.

## Output quality
Governed by [[concepts/process-model-quality]] (syntactic / semantic / pragmatic).

## Related
[[concepts/bpm-lifecycle]] · [[methods/process-discovery-methods]] · [[methods/process-mining-basics]] · [[concepts/process-model-quality]]