BIM Information Manager — Expert Program | Module 2: Interface Management & ICDs (Lead
Appointed Party PoV)
CHAPTER 2.3
Interface Register & Matrix (RASCI)
General-first teaching · Applied-to-AMIA segment · Verdict: Expert (all 5 Gauntlet layers cleared)
How to read this chapter
Body is general (any sector), with an Applied-to-AMIA box. Assessment = Expert
Gauntlet: MCQ → Scenario → Rapid-Fire → Viva → Red-Team (pass ≥8).
● correction · ● learner enrichment · ▲ tracked pattern · Deferrals point to the future
chapter where a topic completes.
Part A — Body
2.1 gave what/why an interface is; 2.2 gave the ICD (the DEPTH artefact). 2.3 covers the
two instruments that let the LAP see and steer the WHOLE interface landscape — the
Register (tracker) and the Matrix (map) — plus RASCI, the model that pins
accountability to every interface.
2.3.1 Register vs Matrix: tracker vs map
Instrument Question it answers Form
Interface Register What's the status of each A list — rows, status, owner,
interface? dates
Interface Matrix Which parties/systems interface A grid/map — systems on
with which, and in which both axes
direction?
A register alone tells you the trees; the matrix shows you the forest.
2.3.2 The Interface Matrix / N² (N-squared) chart
Place systems/parties on BOTH axes, in the SAME order, on the diagonal. Off-diagonal
cells are the interfaces.
• Diagonal = the systems themselves (no self-interface).
• A filled off-diagonal cell = an interface exists, WITH DIRECTION — cell (row i, col j) is
an output from i that is an input to j.
• Cells filled on BOTH sides of the diagonal for the same pair (i→j AND j→i) = a TWO-
WAY / circular dependency — visible on the map instead of discovered on site.
• An empty row/column = a system with no interfaces — should trigger: ‘is that
correct, or have we missed one?’ (gap-spotting).
The N² chart is how you make the n(n−1)/2 explosion from 2.1 legible — it turns an
abstract count into a picture you can govern.
2.3.3 RACI vs RASCI (the responsibility model)
Letter Role Meaning
R Responsible Does the work
A Accountable Answers for the outcome — exactly ONE per interface
S Supportive (RASCI addition) provides resources/help to the
Responsible
C Consulted Two-way — input sought BEFORE action
I Informed One-way — kept updated AFTER
Two rules to lock
The ‘one A’ rule — exactly one Accountable per interface. This IS the ownership rule
from 2.1, formally expressed. The single resolution owner = the single A. Two A's
recreates the ‘shared equally’ defect; zero A's is the unowned-interface leak.
RASCI adds S (Support) to RACI — used on complex interfaces where a party helps
deliver but is neither the core doer (R) nor answerable (A). Common on multi-system
integration.
Why single-A survives genuinely joint work (from the Viva)
R can be shared/plural — many parties can genuinely do the work together. A
cannot, precisely BECAUSE the work is joint: without exactly one answerable party, a
failure becomes a diffused blame game with no one accountable. Single-A isn’t
dishonest about collaboration — it's the safeguard against collaboration collapsing
into finger-pointing when it fails.
2.3.4 Mapping RASCI onto an interface
• A (Accountable) = the resolution owner (the single owner from 2.1)
• R (Responsible) = the party doing the resolution work — often includes the first
mover
• S (Supportive) = a party feeding data/resources to enable it
• C (Consulted) = the other side of the interface
• I (Informed) = downstream stakeholders who must know once resolved
2.3.5 Dependency direction, completed (deferred from 2.1)
• One-way = a single off-diagonal cell (i→j only).
• Two-way = both i→j and j→i filled — iterate, or pick a first mover.
• Circular (3+) = a loop across three or more (A→B, B→C, C→A) — visible as a CYCLE in
the matrix. No single pair looks like a deadlock; only the map reveals the loop. Break
with first-mover selection + freeze points — treat as ONE connected dependency,
never split into separate pairwise ICDs.
2.3.6 The LAP lens
The LAP (via the IM) owns both instruments: the register drives ACCOUNTABILITY
(status, TAT, escalation); the matrix drives FORESIGHT (gaps, loops, clusters needing a
dedicated workshop). The IM maintains the RASCI so every interface has exactly one A.
Currency — why the matrix never becomes wallpaper (from the Viva)
The matrix is not a standalone artefact redrawn from memory — it is a colour-coded,
AUTO-DERIVED VIEW of the Interface Register, refreshed on the same weekly ICM
cadence the register is already governed on, and specifically re-examined at every
change-control event (since that's when a new interface can silently appear or
vanish). Staleness is structurally prevented, not promised.
Applied to AMIA APM
Rows/columns = signalling, traction power, comms, PSD, rolling stock, trackwork, and
the civil/Fixed-Facility interface. The off-diagonal cells expose every system-to-system
seam; two-way cells flag the signalling↔power and similar coupled pairs for first-
mover + freeze-point treatment.
Interview line: ‘I'd build an N² interface matrix across the APM subsystems and run a
RASCI so each interface carries a single accountable owner, with circular
dependencies flagged and freeze-pointed to commissioning.’ Directly answers the
JD's #1 responsibility.
2.3.7 Sample N² charts across sectors
Three worked examples — same tool, different domains — showing one-way, two-way
and gap patterns.
Rail/APM-MRT: signalling shows two-way (circular) cells with traction power, comms
and rolling stock — the classic systems-integration coupling knot; PSD and trackwork sit
as clean one-way recipients.
Marine infrastructure: dredging↔civil is two-way (sequencing — bed levels vs
reclamation programme); marine structures↔mechanical equipment is two-way
(mounting positions vs loads back).
Water infrastructure: treatment↔electrical/SCADA and electrical↔pumping are two-
way (control signals vs status/load feedback) — the process-control coupling pattern
common to treatment plants.
Part B — Doubts
None raised before assessment for this chapter.
Part C — MCQ (Recall) — 30/30
Q1. Correct distinction between Register and Matrix (N² chart)?
A) The Register maps structure; the Matrix tracks status
B) The Register tracks status; the Matrix maps structure/direction ✓
C) They are two names for the same instrument
D) The Matrix replaces the Register once interfaces are closed
Feedback: 10/10.
Q2. Which RASCI role must there be EXACTLY ONE of — the formal expression of the
ownership rule?
E) R — Responsible
F) A — Accountable ✓
G) C — Consulted
H) S — Supportive
Feedback: 10/10. Connected A to the 2.1 ownership rule without hesitation.
Q3. What pattern on an N² chart reveals a two-way/circular dependency?
I) A single off-diagonal cell (i→j only)
J) Cells filled on both sides of the diagonal for the same pair ✓
K) An entirely empty row and column
L) A filled cell on the diagonal itself
Feedback: 10/10.
Part D — Scenario (Application) — 8 · 8
Scenario 1 — RASCI across four parties (fire alarm → comms → control
room)
Learner answer (summary): Initially R=Comms, A=FLS, S=LAP IL, C=LAP design lead,
I=control room — but justified Comms as ‘owner’ while naming FLS as A.
Evaluation — 6/10
C and I assignments defensible; correctly invoked ownership + first-mover rules for
the 'what goes wrong' question.
● Contradiction: named FLS as A but called Comms the 'owner' — A and owner are
the SAME role and cannot sit on different parties.
Re-do 1 — 8/10 CLEARS
R=FLS, A=Comms (‘they integrate every party, i.e. owner’) — A and owner now
consistently the same party. S=LAP IL, C=LAP design lead, I=control room all line up
behind that choice.
Scenario 2 — three-party loop (Signalling → PSD → Rolling Stock →
Signalling)
Learner answer (summary): Correctly named the safety-critical self-locking danger, but
initially dissolved the loop into 'signalling↔rolling stock bidirectional,' losing the PSD
leg.
Evaluation — 7/10
Excellent danger diagnosis (self-locking deadlock, safety risk); good freeze-point
reflex before commissioning.
● Reduced a genuine 3-node cycle to a 2-party pair — the matrix must be read as a
CYCLE, not simplified.
Re-do 2 — 8/10 CLEARS
Kept all three nodes; matrix read as a cycle ('filtering shows three of them
interconnected'); broke it at one arc — signalling issues its standard data sheet first,
PSD interprets and prepares outputs across all variants, remaining legs follow.
(Refinement: name the data as PROVISIONAL/assumed under agreed LOIN, iterating
to firm.)
Part E — Rapid-Fire (Terminology)
# Definition Answer Verdict
1 Instrument that tracks interface STATUS Interface Register ✓
# Definition Answer Verdict
2 Grid/map showing direction of dependency N2 chart ✓
3 Systems-engineering name for the diagonal N2 ✓
grid
4 RASCI role — exactly one, equals the owner Accountable ✓
5 RASCI role that DOES the work “A — Accountable” ✗R
6 Letter RASCI adds to RACI S — Supportive ✓
7 What an empty row/col should prompt Gap-spotting / ✓
recheck
8 Loop of 3+ systems, no single pair looks like a Circular ✓
deadlock dependency
7/10 first pass — #5 collapsed A into R (A given twice for two different questions).
Isolated re-run + scenario test (junior=R/lead=A; technician=R/manager=A) → all
correct. Rapid-Fire CLEARED.
Anchor phrase (memorise)
“R rolls up their sleeves; A answers to the boss.” Apply this to any worked example
before assigning letters — seniority does not automatically equal A; check who
ANSWERS FOR the outcome.
Part F — Viva Cross-Examination — 8 · 8 · 8
Challenge 1 — “Your N2 chart is wallpaper by month six.”
Panel: Is the matrix a live governance tool or a one-time tender artefact that dies on
contact with a live project?
Answer (summary): First pass repeated 'I'll keep it updated' in different words (6/10).
Re-run named the mechanism: the matrix is a colour-coded, auto-derived view of the
Interface Register — 8/10 CLEARS.
Evaluation
It can't go stale because it isn't a standalone document — it's derived from the
register, refreshed on the same weekly ICM cadence, and re-examined at every
change-control event (since that's when a new interface can silently appear).
Challenge 2 — “RASCI is just a spreadsheet with no teeth.”
Panel: What makes naming someone 'A' actually change behaviour?
Answer (summary): 8/10 CLEARS on first pass. Correctly disclaimed direct contractual
force, then named the real mechanism: RASCI is agreed by all stakeholders in the ICM,
and non-fulfilment surfaces at the stage gate, gated against payment.
Evaluation
The RASCI chart itself has no legal force — its teeth come from incorporation, same
as the ICD. Each appointment requires IMSOP compliance; RASCI is how that process
assigns the single Accountable party. Failure surfaces as an NCR at the stage gate,
gated against payment.
Challenge 3 — stress-test of the single-A rule on genuinely joint design work.
Panel: Two systems must jointly, simultaneously redesign a shared control cabinet —
isn't forcing one 'A' dishonest here?
Answer (summary): First pass resolved the practical case (identify principal supplier as
A) but didn't name the theoretical defence (6/10 → 7/10). Re-run named it directly:
many parties can work together, but a single A prevents a 'generic blame game' on
failure — 8/10 CLEARS.
Evaluation
R can be shared/plural for genuinely joint work; A cannot, precisely because the work
is joint — without one answerable party, failure becomes unattributable. Single-A
isn't denying the collaboration; it's the safeguard against it collapsing into a blame
game when teamwork isn't enough.
Part G — Red-Team Audit — 9/10 (5 clean, 1 corrected)
Exhibit D: a flawed Interface Matrix & RASCI procedure. Five genuine defects plus one
CONTROL ITEM (a correct statement, included to test over-flagging).
# Item Correct position
1 N2 produced once at tender, fixed for DEFECT — must be derived from the
the project register and updated after each ICM
2 Both parties marked 'A' for joint work DEFECT — single-A rule; no joint
accountability, ever
3 Empty row/col 'confirms no interfaces, DEFECT — triggers a mandatory gap-
no review needed' spotting review with technical heads
4 R and A may sit with the same single NOT A DEFECT — valid RASCI; control item
party
5 3-party loop split into 3 separate 2-party DEFECT — a genuine cycle stays ONE
ICDs connected dependency with one
coordinated process
6 Matrix retired once construction starts DEFECT — matrix spans design through
construction as a live register-derived
view
● The most important finding of this chapter — confidence to certify correctness
Item 4 was a CONTROL ITEM: a factually correct statement. First reaction
manufactured an unrelated 'fix' rather than saying 'this is correct.' On reflection, the
learner named it accurately: 'I thought this is correct, but felt underconfident to write
that' — then gave a strong substantive answer (valid if genuinely internal; may not
need a full register entry; but must still be visible to any downstream party sitting as
'I').
Principle for the job, not just the exercise: an auditor must be equally willing to say
'this is correct, no action needed' as to flag a fault. Manufacturing defects in sound
material is itself an audit failure.
Part H — Record: Corrections, Pattern, Key Terms, Glossary
● Corrections logged
• Scenario 1: split Accountable and 'owner' across two different parties — they are
the same role, always.
• Scenario 2 (first pass): reduced a genuine 3-node cycle to a 2-party 'bidirectional'
pair — read cycles as cycles.
• Rapid-Fire #5: answered 'Accountable' for both 'who owns' and 'who does the work'
— the A/R collapse.
● Learner enrichments captured
• Audit-confidence principle: certify correct material as correct; don't manufacture
defects to appear thorough.
• If a same-party R=A interface is genuinely internal, it may skip a full register entry —
but must still be surfaced to any downstream party sitting as 'I'.
▲ Tracked pattern (Modules 1–2)
The naming-under-load thread continues, now sharpened to one precise, named
confusion: A (Accountable) vs R (Responsible) collapsing under speed. Fixed instantly
with the anchor ‘R rolls up sleeves; A answers to the boss,’ applied correctly across
three follow-up tests. A second, more valuable pattern also emerged: readiness to
certify correctness, not just find fault — corrected in real time without a re-teach.
“Use This Term When…”
When you mean… Say this term
who DOES the interface work R — Responsible (‘rolls up sleeves’)
who ANSWERS for the outcome / the owner A — Accountable (‘answers to the boss’)
why joint work still needs one A R can be plural; A cannot — prevents a
blame game
why the matrix never goes stale Auto-derived view of the register, weekly
cadence + change-control trigger
a 3+ node loop — don't split it Circular dependency — one connected
loop, one process
reviewing something that might be correct Certify it — don't manufacture a fault
Terms introduced in this chapter (feeds the Module 2 glossary)
Term Definition
Interface Matrix / N2 Grid with systems on both axes (diagonal blank) showing
interface existence and direction
RACI / RASCI Responsibility model: Responsible, Accountable,
Term Definition
(Supportive,) Consulted, Informed
One-A rule Exactly one Accountable per interface — the formal
expression of the 2.1 ownership rule
R vs A anchor ‘R rolls up sleeves (does); A answers to the boss (owns)’
Two-way / circular cell Both i→j and j→i filled on the N2 — a mutual dependency
pattern needing a first mover
Gap-spotting Reviewing an empty N2 row/column to confirm no missed
interface
Matrix currency Matrix as an auto-derived, colour-coded view of the register,
mechanism refreshed on the ICM cadence + change-control triggers
Diffused accountability The failure mode single-A prevents — a joint failure with no
answerable party
Score summary
Layer Score Status
MCQ 30/30 ✓ Cleared
Scenario 8·8 ✓ Cleared (after re-do)
Rapid-Fire 7 → re-cleared ✓ Cleared
Viva 8·8·8 ✓ Cleared
Red-Team 7.5 → 9 ✓ Cleared
Verdict: Chapter 2.3 cleared at expert standard.