Creating a clear decision structure for Expected Credit Loss so that ownership is visible, judgment is controlled and no critical impairment decision sits in ambiguity
Expected Credit Loss is not only a model. It is a governance framework. That fact becomes more obvious as the institution matures. Once methodology, data and controls exist, another question quickly becomes central: who actually decides what? Who owns the policy? Who approves overlays? Who challenges scenario assumptions? Who decides whether a stage override is justified? Who signs off on the final estimate? Who explains the reserve to the board? This is why governance architecture deserves a pillar article of its own.
Governance Architecture for ECL Committees, Roles and Decision Rights explains how institutions should structure ownership, review forums, approval authorities and escalation pathways so that expected credit loss remains clearly controlled and fully governable.
Expected Credit Loss is not only a model. It is a governance framework. That fact becomes more obvious as the institution matures. Once methodology, data and controls exist, another question quickly becomes central: who actually decides what? Who owns the policy? Who approves overlays? Who challenges scenario assumptions? Who decides whether a stage override is justified? Who signs off on the final estimate? Who explains the reserve to the board? This is why governance architecture deserves a pillar article of its own.
A weak ECL framework often does not fail because the underlying concepts are wrong. It fails because the decisions around those concepts are poorly structured. Too many responsibilities remain implied rather than defined. A committee exists, but its authority is vague. A model team believes it owns methodology, while finance believes it owns the booked number and credit believes it owns the qualitative view of the portfolio. Overlays are debated, but no one is fully clear on who can approve them, who can challenge them or when they must be released.
A mature institution solves this through governance architecture. It creates a clear structure of committees, roles and decision rights so that the ECL process is not only technically sound, but also institutionally governable.
ECL is one of the most judgment-sensitive estimates in many institutions. It involves choices and decisions around portfolio scope, segmentation, default definitions, SICR logic, stage overrides, macroeconomic scenarios, overlay use, model change, disclosures and final approval of the booked number. Whenever a framework involves this much judgment, governance architecture becomes critical. Without it, ambiguity and fragmentation arise quickly.
True governance architecture is not simply the existence of meetings. It is the structured allocation of decision rights, review responsibilities, challenge responsibilities, escalation pathways, approval authority and accountability for execution. A committee that receives a pack and nods to it is not yet performing strong governance.
One of the most important design principles in ECL governance is to distinguish between those who prepare or own an estimate component and those who challenge or review it. This is especially important for model changes, stage override frameworks, macroeconomic assumptions, material overlays and final reserve approval.
A good governance architecture aligns decision level with significance. A minor mapping update should not require board-level review. A major sector overlay should not be left to a working analyst alone. A broad change in SICR framework should not be approved casually in an operational forum.
Because ECL sits between risk and financial reporting, governance architecture that is dominated by only one of those perspectives is usually weak. A strong architecture ensures that both risk and finance are structurally present in the key decision chain.
Although institutions vary, many mature governance structures contain several recurring layers: an operational working layer; a technical methodology or model layer; a central ECL review or impairment committee; a senior management or executive risk/finance layer; and a board or audit committee layer.
Operational teams are often best placed to detect issues and prepare analysis, but they may not be the right forum to approve major overlays, major scenario shifts or significant stage-policy changes. A mature architecture keeps the operational layer focused on process completion, data quality, exceptions and routine run review.
Methodology forums should mainly focus on framework design and change, not on every period-end allowance debate. Typical subjects include default and cure definitions, segmentation revisions, SICR methodology, PD/LGD/EAD redevelopment, validation findings and recurring overlay-to-model migration decisions.
In many mature institutions, a central ECL or impairment review committee acts as the key integrative forum. It brings together finance, credit risk, portfolio insight, modelling and sometimes data or economics input, and reviews major movement, stage migration, portfolio deterioration signals, scenario effects, overlays and material exceptions.
A strong architecture should define who can propose an overlay, who must review it, who can challenge the rationale and amount, who can approve it, what documentation is required and when release or reapproval is required.
Macroeconomic scenarios and scenario weights are too important to live in informal discussion channels. The governance architecture should define which forum reviews macroeconomic assumptions, how scenario proposals are documented and who approves final scenario use for the reporting period.
The governance framework should define what qualifies as a material override, what evidence is needed, who approves the override, how long it remains valid, how it is reviewed in subsequent periods and when it must be escalated beyond routine operations.
The institution must know exactly which forum or authority converts analysis into the final approved estimate for booking.
Boards and audit committees are not expected to recalculate ECL, but they should understand the size and direction of reserve movement, major portfolio drivers, material judgment areas, forward-looking assumption themes, use of significant overlays and whether the governance and control environment remains strong.
A strong governance structure not only defines routine approvals. It also defines escalation for unresolved reconciliations, disagreements over overlays, validation findings, late macroeconomic updates and large exposure reclassifications.
Some institutions create too many forums with overlapping mandates. The same issue is discussed repeatedly in different committees, but actual ownership becomes weaker rather than stronger. Good governance is not the maximum number of meetings. It is the minimum structure necessary to make decisions clearly and with adequate challenge.
A mature committee design should rely on clear terms of reference, standard information packs, defined role presence, documented decision rights and standing reporting formats so the governance remains institutional even as individuals change.
A well-designed ECL governance framework is often supported by committee terms of reference, role and responsibility matrices, approval authority maps, overlay governance standards, model change approval protocols, scenario governance guidance, exception escalation procedures and standard pack structures for period-end review.
Recurring weaknesses include unclear final ownership, excessive reliance on one individual to bridge risk, finance and model logic, informal overlay governance, overlapping committee mandates, weak escalation pathways, board materials that report the number but not the judgment behind it, and model, scenario and reporting changes being handled in separate silos without one integrative forum.
One institution may have several committees but weak decision clarity, while another has the same broad governance layers yet clear mandates, scenario handoffs, explicit override rules and a defined approval route for the final reserve. Both have committees; only one has real governance architecture.
A strong institutional governance architecture usually includes clear committee hierarchy, defined role ownership across risk, finance, modelling and business functions, explicit decision rights for overlays, scenarios, model changes and final reserve approval, formal escalation paths, well-structured committee packs and board reporting with meaningful judgment visibility.
Governance architecture for ECL committees, roles and decision rights is what turns the impairment framework from a technically assembled estimate into an institutionally controlled decision system. It ensures that no material choice is left floating in ambiguity, no major judgment sits without a home, and no final reserve reaches the books without a clear chain of challenge and approval behind it.
