Redesign heuristics vs. PrPM interventions — side-by-side mapping

This page maps the 29 design-time BPR heuristics of Reijers & Liman Mansar (2005) onto the runtime prescriptive interventions characterised by the PrPM SLR (Kubrak et al. 2022). The aim is to surface where the two literatures address the same class of intervention and where they do not — a question Kubrak's SLR conspicuously declines to answer (see also bpm-phases-and-bpr-legacy §Interlude for the narrative version).

Why the mapping matters

The 2005 catalogue is design-time, expert-driven, qualitative, scoped to the technical challenge of redesign. Trade-offs are evaluated against the devil's quadrangle (cost / time / quality / flexibility).
The 2022 SLR characterises runtime, data-driven, per-case prescriptive methods. Trade-offs are not evaluated against the devil's quadrangle; most methods optimise a single KPI.
The two literatures share an intervention vocabulary (resource (re)allocation, task re-ordering, next-task recommendation, escalation) but Kubrak does not cite Reijers; the design-time predecessor is invisible to the PrPM corpus.
Mapping them exposes which heuristics have a runtime analogue, which do not, and which PrPM intervention types are missing a design-time grounding.

PrPM intervention vocabulary (Kubrak §RQ2)

Kubrak collapses the 37-method corpus's interventions to two process perspectives — control-flow (next task / next sequence) and resource (allocation, reassignment) — both hand-specified by analysts. Five policy forms exist (similarity-based, rule-based, exceeding-metric-limits, maximum-metric-improvement, probability-of-negative-outcome-above-threshold), characterised orthogonally as continuous vs. discrete and as optimizing vs. guiding.

The mapping table

Reading: ✓✓ = direct runtime analogue exists in PrPM; ✓ = partial / adjacent analogue; — = no PrPM analogue in Kubrak's corpus.

#	Reijers 2005 heuristic	Element	Devil's-quadrangle dimensions	PrPM analogue (Kubrak 2022)	Mapping
1	Control relocation	Customers	quality, cost	— (no method moves controls to the customer)	—
2	Contact reduction	Customers	time, quality, cost	—	—
3	Integration	Customers	time, flexibility, cost	—	—
4	Order types	Operation	time, quality, cost, flexibility	Routing recommendation: "for this case, switch to variant X" — closest match is next-task / next-sequence guidance	✓
5	Task elimination	Operation	time, flexibility, cost	Skip-this-task recommendation under deadline pressure (degenerate next-task)	✓
6	Order-based work	Operation	time, cost	— (batching is a design-time choice; PrPM acts per-case)	—
7	Triage	Operation	quality, time, cost, flexibility	Continuous routing of cases to specialist queues based on predicted complexity	✓✓
8	Task composition	Operation	time, quality, cost, flexibility	— (composition is a design-time choice)	—
9	Resequencing	Behaviour	time, cost	Next-task recommendation = runtime resequencing of the remaining suffix	✓✓
10	Parallelism	Behaviour	time, cost, flexibility, quality	— (introducing parallelism mid-case is rare in PrPM)	—
11	Knock-out	Behaviour	time, cost	Maximum-metric-improvement policy that reorders remaining checks; alarm-based stop-this-case is closest	✓
12	Exception	Behaviour	time, quality, flexibility	Probability-of-negative-outcome-above-threshold policy that diverts cases to exception flow	✓
13	Order assignment	Org. structure	time, quality, flexibility	Resource-allocation methods that bind a case to one resource end-to-end	✓
14	Flexible assignment	Org. structure	queue time, quality, flexibility	Continuous resource (re)allocation per step (Conforti et al.; de Leoni et al.)	✓✓
15	Centralization	Org. structure	flexibility, time, cost	— (a structural choice, not a per-case action)	—
16	Split responsibilities	Org. structure	time, quality, flexibility	—	—
17	Customer teams	Org. structure	cost, time, flexibility, quality	—	—
18	Numerical involvement	Org. structure	time, cost, quality	—	—
19	Case manager	Org. structure	quality, satisfaction, cost	—	—
20	Extra resources	Org. population	time, flexibility, cost	"Hire / borrow capacity" alarm is implicit in some maximum-metric-improvement methods; Sindhgatta et al. (2016) prescribes adding resources to bottleneck step	✓
21	Specialist–generalist	Org. population	time, flexibility	Resource-recommendation methods using experience features (Arias et al.; Yang et al.) effectively prescribe specialist over generalist per step	✓✓
22	Empower	Org. population	time, quality, cost	— (PrPM does not prescribe organisational roles)	—
23	Control addition	Org. population	time, quality, cost	Probability-above-threshold alarms that add a check step on suspicious cases	✓
24	Buffering	Information	time, cost	—	—
25	Task automation	Technology	time, quality, flexibility	— (a design-time decision; the act of automating is not a prescription)	—
26	Integral BP technology	Technology	quality, cost, time	—	—
27	Trusted party	External	cost, time	—	—
28	Outsourcing	External	cost, quality	—	—
29	Interfacing	External	cost, quality, time	—	—

Quantitative summary of the mapping

Direct runtime analogues (✓✓): 4 of 29 — Triage, Resequencing, Flexible assignment, Specialist–generalist. All four sit in Operation, Behaviour, or Organization and concern how a single case is routed or resourced now.
Partial / adjacent analogues (✓): 6 of 29 — Order types, Task elimination, Knock-out, Exception, Order assignment, Extra resources, Control addition. Mostly behaviour-layer alarms or resource bindings.
No PrPM analogue (—): 19 of 29. Categorically: every Customers, Information, Technology, and External Environment heuristic, and most Org. structure heuristics. These are design-time structural choices, not per-case actions.

Inverse view — what PrPM does that the 2005 catalogue does not name:

Alarm-based prescription triggered by predicted probability of negative outcome (Teinemaa et al. 2018; Fahrenkrog-Petersen et al. 2022). No 2005 heuristic prescribes "act at runtime when probability of failure exceeds θ" — the entire alarm primitive is post-2005.
Causality-aware policies using CATE / uplift modelling (Bozorgi et al. 2021; Shoush & Dumas 2021). The 2005 catalogue evaluates trade-offs but never estimates them counterfactually.
Continuous resource reallocation at every step (Schonenberg et al. 2008; Barba et al. 2011) — the catalogue assumes assignment is a design-time choice.

What the comparison reveals

PrPM concentrates on a slim slice of the 2005 catalogue. Only ~10 of 29 heuristics have a runtime analogue, and they all cluster in three framework elements: Operation, Behaviour, Organization. The remaining 19 — covering customers, information, technology, external environment, and structural org. choices — remain design-time and are not yet addressed at runtime.
The "missing 19" are structural. Heuristics like centralisation, customer teams, outsourcing, integral BP technology, buffering cannot meaningfully be applied to one running case. They reshape the process, not its instances. PrPM is, by construction, an instance-level discipline.
PrPM lacks a devil's-quadrangle scaffold. No PrPM method in Kubrak's corpus evaluates its prescription against the four-dimensional trade-off. Most optimise a single temporal KPI; quality, flexibility, and cost are mentioned single-digit times in the 37-method corpus. This is a regression relative to 2005.
PrPM adds two primitives 2005 could not express. Alarm-based triggering on predicted-probability thresholds, and causal estimation of intervention effects (CATE / uplift). These are first appearances in the design-vs-runtime arc.
The two literatures should converge. Heuristics 7, 9, 14, 21 in the 2005 catalogue already have ML-backed runtime methods. The honest research agenda — implicit in Kubrak's Gap #2 ("no method discovers interventions from logs") — is to extend PrPM's intervention discovery so that it can learn runtime variants of the 19 structural heuristics, with explicit trade-off labelling per the devil's quadrangle.

Where to act on this mapping

For PrPM researchers: Heuristics 1–3, 6, 8, 10, 15–19, 22, 24–29 are runtime-unexplored. Several admit instance-level reinterpretations (e.g., Numerical involvement → a runtime alarm that flags cases where the number of distinct resources exceeds a threshold and reassigns to consolidate). The catalogue offers 19 candidate research directions.
For redesign practitioners: A PrPM dashboard is not a substitute for the Reijers–Liman Mansar catalogue. It addresses ~10 of 29 levers; the rest remain a design-time conversation tied to scope, structure, and external relationships — exactly the levers the TRAC scope model frames first.
For agentic-BPM thinkers: The Dumas et al. (2026) call for an "agentic extension of the Reijers & Liman-Mansar redesign heuristics" (see syntheses/bpm-phases-and-bpr-legacy) is the integration target: a single catalogue indexed by both design-time framework element and runtime intervention class, with trade-off profiles attached.

Concepts: concepts/bpr-heuristics · concepts/devils-quadrangle · concepts/prescriptive-process-monitoring · concepts/intervention-policy · concepts/bpm-lifecycle

Methods: methods/process-redesign-heuristics

Sources: sources/2005-reijers-limanmansar-best-practices-bpr · sources/2022-kubrak-prescriptive-ppm-slr · sources/2018-dumas-fundamentals-of-bpm · sources/2014-groger-prescriptive-analytics-bpo · sources/2026-dumas-agentic-bpms-pyramid

Sibling syntheses: syntheses/bpm-phases-and-bpr-legacy (narrative companion — Interlude on PrPM vs BPR) · syntheses/ppm-landscape (PPM/PrPM corpus map) · syntheses/prescriptive-process-monitoring-lineage