Qualitative Discovery Method Selection — Decision Matrix

Reference matrix for choosing among the qualitative process discovery methods catalogued in sources/2018-dumas-fundamentals-of-bpm §5.2. Designed to feed a decision flowchart: each row/column is a candidate node or branch condition.

Five methods (Dumas's evidence-based class is split into three sub-methods that behave differently for selection purposes):

Document analysis (§5.2.1)
Observation — passive (§5.2.1)
Observation — active ("play the customer", §5.2.1)
Automated process discovery (§5.2.1; covered fully in methods/process-mining-basics)
Interview (§5.2.2; see also syntheses/interview-structuring-for-process-models)
Workshop (§5.2.3)

For comparison with quantitative/automated alternatives see methods/process-discovery-methods.

1. Dumas's four headline criteria (Table 5.1)

Method	Objectivity	Richness	Time consumption	Immediacy of feedback
Document analysis	High	Medium	Low–medium	Low
Observation (passive)	High	Medium-high	Medium-high	Low
Observation (active)	Medium-high	Medium	Low	Medium
Automated discovery	High	Medium	Medium-high	Low
Interview	Medium-high	High	Medium	High
Workshop	Medium-high	High	Medium	High

(Active observation is not in Dumas's table but is described separately in §5.2.1 and behaves distinctly.)

2. Strengths and weaknesses (Table 5.2)

Method	Strengths	Weaknesses
Document analysis	Structured information; independent of stakeholder availability	Outdated material; wrong abstraction level; normative ≠ actual
Observation	Context-rich insights	Potentially intrusive; Hawthorne effect (stakeholders behave differently); only few cases observable; not all processes can be observed
Automated discovery	Extensive case set; objective data	Data-quality / abstraction issues; data may be unavailable or partial; data extraction is time-consuming
Interview	Context-rich insights	Requires sparse stakeholder time; several iterations before sign-off
Workshop	Context-rich; direct resolution of conflicting views	Requires simultaneous availability of multiple stakeholders; multiple sessions typically required

3. Decision-relevant characteristics (extended — for flowchart nodes)

Each row is a candidate flowchart branch condition. "Required" means the method cannot be used without it; "–" means not applicable / does not constrain.

Branch condition	Document analysis	Observation (passive)	Observation (active)	Automated discovery	Interview	Workshop
Documentation exists?	Required	–	–	–	–	–
Process supported by an IT system with a clean event log?	–	–	–	Required	–	–
Analyst can play the customer/initiator?	–	–	Required	–	–	–
Site / observation access permitted?	–	Required	Required	–	–	–
Access to domain experts?	Not required	–	–	Not required	Required	Required
Simultaneous availability of multiple roles?	–	–	–	–	Not required	Required
Typical iterations to converge	1	1–2	1–2	1–2	≥2 per expert	3–5 sessions
Sensitivity to politics / culture	Low	Low–medium	Low	Low	Medium (1-on-1 dampens politics)	High (collapses in hierarchical/closed cultures)
Risk of sunny-day bias	Low (but normative bias)	Low	Low	Low	High without rainy-day questions	Medium
Hawthorne effect	No	Yes	No	No	No	Low
End-to-end visibility	Partial	Partial (observable parts only)	Customer-facing steps only	If the log covers it	Limited by expert's scope	Good (multiple roles)
Output granularity	Activity level (often too fine)	Activity + step level	Customer-touchpoints only	Activity level	Activity level	Activity level (start coarse)
Output recency	Backward-looking (docs lag reality)	Now	Now	Now (log-recency)	Now	Now
Suitable for ill-defined process?	No	No	No	Only if logged	Yes	Yes
Suitable for process with no IT support?	If documented	Yes	Maybe	No	Yes	Yes
Suitable for remote / dangerous environment?	Yes	No	No	If logged	Phone/video possible	No
Surfaces undocumented workarounds?	No	Yes	Limited	Maybe (anomalies)	Yes (with rainy-day)	Yes
Resolves conflicting perceptions across roles?	No	No	No	No	Indirect (cross-check across interviews)	Direct and fast

4. Dumas's explicit recommendation (§5.2.4 final paragraph)

"Start with document analysis (low friction, accessible at short notice), then layer interviews/workshops on top depending on context and budget."

In other words: documents are the default first node, not an alternative — the other methods are chosen on top of document analysis.

5. Hard "cannot be applied" conditions (§5.2.4 question box)

These are explicit terminator nodes for a flowchart:

No IT support / partial coverage → rules out automated discovery.
Remote / dangerous environment (e.g. offshore oil platform) → rules out observation (both modes).
Strict-hierarchical, non-open culture → rules out workshop (subordinates won't speak freely).
Rapid-growth start-up without documents → rules out document analysis.
Strong internal politics + hidden agendas → avoid workshop, prefer parallel interviews.

6. Reference flowchart spine

Dumas's logic condenses naturally to:

Documentation available? → yes: start with document analysis (always, if available). No: skip.
End-to-end IT support with a clean event log? → yes: run automated discovery in parallel.
Stakeholder conflict expected OR time budget tight? → yes: workshop. No: continue.
Political / hierarchical environment OR geographically distributed? → yes: serial interviews. No: workshop is still cheaper long-term.
Residual uncertainty about actual vs. normative behaviour? → add observation (passive if access; active as proxy otherwise).

7. Sharp's 90-minute facilitated scoping session — a fast alternative

Alec Sharp (sources/2014-sharp-using-scope-models) describes a distinct discovery pattern that Dumas does not catalogue separately: the 90-minute facilitated scoping session, which produces a Process Summary Chart (TRAC scope + organisational functions on a one-pager) plus an issues/goals table in a single sitting.

Attribute	Sharp's 90-min scoping session
Output	TRAC scope model + Process Summary Chart + initial assessment (issues, to-be goals)
Participants	8–12 with significant domain background
Pre-work	Sketch initial TRAC; gather documentation
Facilitator	Required (experienced); single-person
Iterations	1 session to first draft; often 1 follow-up to clean up swimlane "offline"
Granularity	Subprocess level (5±2), not activity level
Best for	Time-pressured executives; when swimlane mapping would be overkill; when the process is emergent/adaptive (no defined flow to map)
Worst for	Deep ill-defined investigations; hostile/political groups (facilitator cannot hold the room)

Sharp's position: swimlane mapping is often over-applied — it remains irreplaceable for illustrating ugly reality, but a Process Summary Chart + enabler assessment can replace it when time-pressure dominates. This is an explicit down-shift from Dumas's 3–5 session workshop cadence.

When to prefer Sharp's 90-min approach over Dumas's multi-session workshop

Executive briefing or sponsor buy-in is the deliverable, not a full as-is BPMN.
The process is emergent, adaptive, or knowledge-intensive — no stable flow to map.
Time budget <1 week to first usable artefact.
There are strong facilitator skills available.

When to prefer Dumas's multi-session workshop

Full as-is swimlane with exceptions and data objects is the deliverable.
Process is well-defined and repetitive.
≥3 sessions over 2–4 weeks is available.

8. Workshop logistics constraints (§5.2.3, for flowchart capacity nodes)

If workshop is selected:

3–5 sessions of ~3 hours each.
≤10–12 participants per session (above this, parole budget collapses).
Session 1: lightweight participatory modelling (sticky notes, no gateways yet).
Session 2: brief BPMN primer + validation of consolidated model.
Subsequent sessions: control flow → exceptions → data.
Requires facilitator, process modeler, scribe roles in addition to the analyst.

9. Interview logistics constraints (§5.2.2)

If interview is selected:

~1 hour per interview, balanced 45 min structured + 15 min free-form.
≥2 iterations per interviewee (Interview → Modeling → Validation cycle).
See syntheses/interview-structuring-for-process-models for the full operational guide.

10. CAMAS-aware extension

For organisations operating CAMAS-style (sources/2021-vombrocke-context-aware-bpm-camas) method selection, the discovery-method choice is itself a context-conditional decision. Map the project's frameworks/bpm-context-framework profile to the matrix above:

High knowledge-intensity + high creativity + non-repetitive → workshop dominant; document analysis weak (little to document); automated discovery weak (low log volume).
Repetitive + low-knowledge + IT-supported + large-org → automated discovery dominant; document analysis strong; workshop optional for cross-role validation.
Start-up + low resources + supportive culture → workshop + active observation; document analysis weak; automated discovery often impossible.
Strict-hierarchy + non-supportive culture → serial interviews + document analysis; workshop ruled out.

11. Out of scope: JTBD as a customer-lens alternative

All six methods above take a performer / process-owner lens. When the deliverable is a product/service redesign or a customer-facing process, consider Jobs-to-Be-Done as a seventh method positioned alongside — not inside — this matrix. JTBD interviews the customer, reconstructs a purchase timeline, and probes the "struggling moment" (sources/2016-christensen-jobs-to-be-done); Ulwick's ODI adds quantitative importance × satisfaction scoring.

JTBD is not in the flowchart spine above because it answers a different question (what job? what unmet outcome?) rather than (what does the as-is process look like?). But it should be offered in the inception phase whenever the process is customer-facing and the goal is redesign — not just documentation.

methods/process-discovery-methods — the parent method page.
syntheses/interview-structuring-for-process-models — operational guide if the interview branch is chosen.
concepts/process-discovery · concepts/context-aware-bpm · frameworks/bpm-context-framework · concepts/six-enablers-framework
frameworks/jtbd · concepts/outcome-driven-innovation — customer-lens alternative for redesign inception.
sources/2018-dumas-fundamentals-of-bpm-ch5-discovery — primary chapter source.
sources/2014-sharp-using-scope-models · sources/2014-sharp-whats-wrong-with-this-process — practitioner 90-min scoping pattern + assessment framework.