Skip to content

Toolkit — System description

Gate: G1 Observe. Category: framing artefact.

What problem it solves

A framed problem names what is wrong; a stakeholder map names who is touched. Neither says how the thing actually works right now. Without a system description, G2's decomposition invents its own picture of the system, and G3 routes pieces against that invented picture. The description forces one accurate diagram of the current state — inputs, flows, decision points, hand-offs, substrates — before anyone proposes to change it.

How it is used

A 60–90 minute G1 walk-through, with one practitioner narrating the current process while the chair draws. The output is a single-page diagram and a one-page written companion. The description is as-is, not to-be. Revisited at G3 to identify the decision points that routing will assign to, and at G4 to identify the substrates the piece's rollback path reverts to.

Inputs

  • Framed problem paragraph and stakeholder map.
  • A practitioner who performs the process end-to-end (or two who together cover it).
  • Access to the systems, dashboards, and artefacts the process touches.

Outputs

  • A single-page diagram of the current process: inputs, steps, decision points, hand-offs, outputs.
  • A written companion naming each decision point, who makes it, and on what information.
  • A substrate inventory: the rule-sets, dashboards, spreadsheets, and scripts the process depends on today. These become candidate rollback substrates at G4.
  • Flagged items where the practitioner's description and the documented process disagree — the work-as-done versus work-as-imagined gap.

Visualisation

System description — inputs flowing through steps and a decision diamond to outputs, with substrate callouts beneath Input truck arrives Step radio check-in Decision slot free? Step assign slot Step hold + radio Output truck to bay carrier feed substrate yard map (paper) substrate radio log substrate

Solid boxes: steps and decisions in the current process. Dashed boxes: substrates each step depends on. The lines trace material flow; the verticals trace substrate dependencies.

Anatomy

Inputs and outputs. What enters the process (a truck, a carrier feed, an operator request) and what leaves it (a slot assignment, a dispatched allocation). Named in the substrate's own language.

Steps. Atomic actions, each performed by a named role or system. "Dispatcher checks the yard map and radios the driver" is a step; "truck arrives" is an input event.

Decision points. Places where the process branches based on the input's state. Drawn as diamonds. "Is the slot free?" "Does the carrier have priority status?" Each decision point is a candidate for routing at G3.

Hand-offs. Transitions from one role or system to another. Hand-offs are where information is lost, re-formatted, or delayed; they are the reliable residence of slowness and error.

Substrates. The artefacts the process depends on: rule-sets, spreadsheets, scripts, dashboards, paper forms. Inventoried because they will appear at G4 as candidate rollback substrates. "The slot-assignment step uses a paper yard map; the map is updated on the fly by the dispatcher."

Work-as-done vs. work-as-imagined. Practitioners describing the process often name steps that are not in the documented procedure, and skip steps that are. The gap is flagged; it is not a bug — it is evidence of where the documented process has already fallen out of alignment.

Example

Paper trail — describing the freight-yard system

G1 walk-through, 75 minutes. Practitioner: Raj. Chair: Ada. Observer: Priya.

T+0 — inputs and outputs. Ada: "what comes into this process?" Raj: "trucks. The carrier feed tells us roughly when each truck will arrive." Ada draws carrier feed + truck arrival as inputs. Outputs: truck to bay + radio confirmation.

T+15 — steps. Raj walks through a typical truck: radio check-in → dispatcher looks at yard map → dispatcher decides slot → radio back the assignment. Ada draws four steps and one decision diamond.

T+30 — the paper yard map. Raj: "the yard map is a paper board with magnets. The dispatcher moves magnets as trucks come and go." Ada: "that is a substrate. Is the map authoritative, or is there a digital copy?" Raj: "the paper is authoritative. There is a digital copy in the yard management system, but it lags." Ada notes the substrate; flags the lag as a potential rollback issue.

T+45 — work-as-done gap. Ada: "how often do dispatchers go off-map?" Raj: "all the time. A carrier calls for priority, a driver is short on hours, someone negotiates a swap. The map shows the plan; the radio carries the reality." Ada flags the gap: work-as-imagined is the map; work-as-done is the radio. The allocator piece, if it relies on the map alone, will be out of step with the radio-level reality.

T+60 — substrate inventory. Three substrates: carrier feed, yard map (paper), radio log. Each annotated with who owns it, how often it updates, and whether it is a candidate rollback substrate (paper map: candidate; radio log: not — no structured read).

T+75 — close. Diagram photographed, written companion drafted. The dispatcher-radio workaround becomes a G3 note about routing: the allocator cannot assume the yard map alone is ground truth.

Pitfalls

As-is drifts to to-be. The practitioner describes the process as it should be. The chair has to keep asking is that what actually happens on a Tuesday? The value of the description is its fidelity to current state.

Missing substrates. The most load-bearing substrates are the ones nobody mentions — the paper form at the end of the shift, the spreadsheet on one person's laptop, the Slack channel where decisions get confirmed. Ask what does this step write to? what does the next step read from? and the substrate appears.

Box-and-arrow illusion. A clean diagram suggests a clean process. If the practitioner's narration is messier than the diagram, the diagram is lying. Annotate the messy parts.

Decision points as implementation details. Decision points are the engagement's most valuable features — they are candidates for routing at G3. Skipping them because they seem like internal mechanics is the routing's loss.

Diagrams without the written companion. The diagram compresses; the companion names who makes each decision, on what. Without the companion, later readers reinterpret the diagram to match their own mental model.

When not to use

  • Engagements where the process is already well-documented in a current, trusted source. Read the document; confirm its accuracy with one practitioner; do not redraw.
  • Pure greenfield engagements where no current process exists. The system description is thin; the engagement's first output is the system.

Provenance

The discipline of as-is mapping before to-be design is foundational in process engineering, documented in the IDEF0 standard [1] for functional modelling. The distinction between work-as-done and work-as-imagined is Hollnagel's contribution to resilience engineering [2]. SIPOC, covered separately in this toolkit, is a reduced form of the system description used at G3 when the G1 walk-through has already been done.

  • Framed problem paragraph. Names what the system is failing to do.
  • Stakeholder map. Names who touches which step of the system.
  • SIPOC (G3). A compressed as-is view when a routing-stage description is needed.
  • Value-stream mapping (G3). The G3 analogue for pieces whose routing is operational.

Verification

[1] Integration DEFinition (IDEF) 0 Function Modeling. FIPS Publication 183. National Institute of Standards and Technology; 1993. [verified] The IDEF0 standard for as-is and to-be functional modelling.

[2] Hollnagel E. Safety-I and Safety-II: The Past and Future of Safety Management. Ashgate; 2014. [verified] Names the work-as-done / work-as-imagined distinction and its consequences for system design.