Exposure at default in plain language

How to study this topic

Exposure at default, or EAD, is the bank's estimate of how much exposure the bank will have to a borrower or counterparty at the moment default occurs. The phrase may sound technical, but the banking meaning is practical. A bank uses this concept to decide how much risk it is taking, how much loss it may face, how to price or limit exposure, when to refer a case, how to monitor portfolios and how to explain credit decisions.

Keep the scope banking-first. This chapter is about lending, borrowers, counterparties, credit obligations, portfolios, limits, collateral, repayment, arrears, provisions, capital, governance and customer treatment. It is not a payments chapter. Transaction behaviour may appear only as account conduct evidence where it genuinely supports credit risk.

A learner should finish this topic able to explain it to a credit analyst, business analyst, data engineer, model validator, underwriter, collections manager and auditor. Plain language matters because credit-risk models are not useful if only modellers understand the answer.

The plain language meaning

EAD answers the question: if default happens, how much will the bank be exposed to at that moment? For a normal term loan this may be close to outstanding balance. For revolving credit, overdrafts, credit cards, trade finance or undrawn commitments, the exposure at default may be higher than today's drawn amount.

The concept must always be tied to purpose. A number without purpose is dangerous. A PD used for application decisioning may differ from a PD used for IFRS 9, capital or portfolio monitoring. An LGD used for pricing may differ from a downturn LGD view. An EAD for stress testing may differ from day-to-day exposure monitoring. Scorecards also differ by product, portfolio and decision point.

Plain language does not mean weak explanation. It means the explanation should remain accurate without hiding behind formulas.

Source systems and evidence

Useful source areas include outstanding balance, undrawn commitment, credit limit, utilisation trend, overdraft usage, loan amortisation schedule, prepayment behaviour, credit conversion factor, trade finance exposure, treasury counterparty exposure, netting agreement, and collateral margin. A credit-risk chapter depends on controlled data because every input can change the risk view. Income evidence, arrears data, utilisation, collateral value, limit data, customer identity, bureau information and case outcomes all need clear source ownership and timing.

The bank must know which source is authoritative. A front-end application may show what the customer declared. A document-verification platform may show what was evidenced. A core loan system may show what was booked. A collections system may show what happened after stress started. A risk engine may show the rating and score. Finance may show impairment treatment. These views must be mapped, not casually mixed.

The source question becomes sharper when AI and ML are used. If a model learns from a field, the bank must know whether that field is complete, stable, corrected, overridden, refreshed and legally usable for the purpose.

Key inputs and calculations

Key inputs include outstanding balance, undrawn commitment, credit limit, utilisation trend, overdraft usage, loan amortisation schedule, prepayment behaviour, and credit conversion factor. The calculation must define the time horizon, observation period, product scope, customer population, exclusion rules and treatment of missing values. Without these, the same concept can be calculated in several different ways and produce conflicting results.

For EAD, the bank needs balance, limit, utilisation, undrawn amount, product type, amortisation, prepayment, drawdown behaviour and off-balance-sheet treatment. The challenge is to estimate what exposure will exist at default, not merely what exposure exists today.

The formula is only one part. Banking interpretation, data quality and governance decide whether the output is safe to use.

Controls before use

Controls should include exposure definition, limit data quality, utilisation snapshot, drawdown behaviour, credit conversion factor governance, and amortisation logic. These controls must run before the concept feeds a model, decision engine, risk report, provisioning process or strategy review. Technical checks are not enough. The bank must check whether the value makes sense in credit terms.

Important checks include population coverage, source reconciliation, date alignment, treatment of overrides, treatment of outliers, missingness analysis, validation against realised outcomes and monitoring of drift. Credit-risk models can fail slowly. They may not throw a system error. They may simply become less aligned with borrower behaviour or economic conditions.

The bank should also define who can override the model, how overrides are recorded, and how override performance is reviewed. Human judgement can improve credit decisions, but undocumented override culture can damage validation.

How it supports bank decisions

Common uses include expected loss calculation, capital calculation, limit management, portfolio monitoring, stress testing, counterparty credit risk, credit approval, and early warning. Each use has a different control requirement. A model used for internal portfolio insight may need one level of governance. A model used for credit approval, regulatory capital, provisioning or customer treatment needs much stronger evidence, validation and monitoring.

EAD is central because it influences the exposure side of credit loss. A low-risk borrower with a large undrawn commitment can still create material exposure if the facility is drawn before default. Banks need to measure possible exposure, not only current balance.

The output should be explainable to the decision owner. If a credit manager cannot understand why the model changed a view, the bank has a governance problem even if the statistical result is strong.

Provisioning, capital and portfolio view

Credit-risk concepts matter beyond individual decisions. They feed portfolio monitoring, risk appetite, concentration review, stress testing, capital planning and expected credit loss. The same borrower-level evidence can roll up into management and regulatory views.

IFRS 9 expected credit loss uses credit-risk movement to decide whether a financial asset remains in a 12-month expected loss view or moves into lifetime expected loss because credit risk has increased significantly. That makes data timing, default definition and risk movement important. A poor risk signal can affect provisions and reported performance.

Capital frameworks also care about credit-risk parameters. PD, LGD and EAD are not classroom abbreviations. They are part of how banks understand expected and unexpected loss, risk-weighted assets, portfolio quality and risk-adjusted return.

Customer and conduct impact

Credit-risk models affect customers. They can approve credit, restrict credit, increase price, reduce limit, refer for manual review, trigger collections activity or identify customers needing support. A wrong model can harm a customer by granting unaffordable credit or by denying reasonable credit without a fair basis.

Fairness is not solved by removing obvious sensitive fields. Proxies can appear through income, geography, employment type, channel usage, account history, product access or missing data. The bank should review whether the model treats customer groups fairly and whether limitations are visible to decision-makers.

Human review is important where impact is high. A score can support judgement, but the bank remains accountable for the decision.

Validation and monitoring

Validation should review concept, data, methodology, calibration, discrimination, stability, limitations and use. It should not only ask whether the model predicts. It should ask whether the model is suitable for the bank's portfolio, product, policy, customer population and decision purpose.

Monitoring should compare predicted outcomes with actual outcomes. For PD, defaults should be compared with predicted default rates. For LGD, realised loss and recovery experience should be reviewed. For EAD, drawdown and exposure at default should be tested. For scorecards, approval quality, referral rates, override outcomes and portfolio performance should be monitored.

If performance deteriorates, the bank may need recalibration, redevelopment, policy adjustment, overlays, tighter controls or temporary restrictions. Model monitoring must have a response path.

Common mistakes

The first mistake is confusing the parameter with the whole credit decision. PD, LGD, EAD and scorecards are decision inputs, not the entire decision. The second mistake is using weak labels. The third mistake is ignoring data timing. The fourth mistake is assuming a score remains valid after products, policy or economic conditions change.

Another mistake is treating a model as neutral because it is quantitative. Credit models reflect historical data, policy decisions, source availability, customer access and economic cycles. The bank must challenge whether the model learned credit risk or learned old operating patterns.

The final mistake is failing to explain limitations to users. A model can be useful with limitations if users understand them. A model becomes dangerous when limitations are hidden.

Practical banking example

A corporate borrower has a partly drawn revolving facility. Today's outstanding amount is not the full risk. If the borrower draws more before default, the bank's exposure at default increases. EAD tries to capture that future exposure possibility.

The practical point is that credit-risk models support judgement. They should make banking decisions more consistent and evidence-based, but they should not remove responsibility from the bank.

Bank-ready checklist

Before using this concept in production, ask: is the definition clear, is the population right, are sources authoritative, are time windows correct, are labels validated, are overrides governed, are missing values explained, are customer impacts reviewed, and can the result be reproduced later?

Then ask: is the model calibrated, independently reviewed, monitored, documented and approved for this exact purpose? Is it being used for approval, pricing, provisioning, capital, collections or internal insight? Has the bank matched controls to impact?

If the answers are strong, the concept is bank-ready. If not, the model may still produce numbers, but the bank should not treat those numbers as trusted credit-risk evidence.

The exposure question at a future default

Exposure at default (EAD) estimates the amount owed or otherwise exposed when a borrower defaults. It is measured at a future event, not necessarily at today's balance. A term loan that amortises may have less outstanding at that event. A credit card, overdraft or committed line may have more drawn than it does now. The bank needs to specify the facility, observation date, default horizon, product terms and intended use before an EAD number can be interpreted.

An EAD estimate is not a loss amount. Probability of default (PD) addresses whether a defined default occurs; loss given default (LGD) addresses the proportion or amount not recovered under the chosen method. EAD addresses the exposure at that point. Collateral, guarantees and expected recoveries usually belong in the loss-severity or credit-risk-mitigation analysis, subject to the governing methodology. They should not be subtracted casually from the drawn balance and called EAD. The exact treatment differs between accounting, capital, pricing and internal risk uses.

Consider a fictional small-business revolving facility with a limit of 100,000 currency units and 40,000 drawn today. The undrawn portion is 60,000. If an illustrative internal forecast assumes that 30% of the remaining commitment would be drawn before a possible default, its simple projected exposure would be 40,000 plus 18,000, or 58,000. That 30% is only a teaching assumption, not a Basel credit conversion factor, a regulatory minimum or a recommended model parameter. A real estimate would consider the timing of drawdowns, repayments, limit changes, interest, fees and the facility's legal availability.

Drawn, undrawn and legally available amounts

A loan system may show a credit limit and an outstanding balance, but the difference is not always a usable undrawn commitment. An uncommitted line, a cancelled facility, a conditional drawdown, a pending limit change and an already authorised but unposted transaction can each affect what the bank is exposed to. The analyst should identify the contractual commitment, operational availability and accounting balance separately. A screen that shows a customer-facing available amount may not equal the exposure measure required for a model or regulatory calculation.

For a credit card, purchases, cash advances, repayments, interest and fees can move the balance between observation and default. For an amortising term loan, scheduled instalments and prepayments may reduce it. A mortgage with a committed but not fully advanced construction amount has a different drawdown pattern from a revolving working-capital line. Trade-finance instruments and guarantees can create contingent exposure whose conversion follows their own terms and applicable method. One universal percentage for every undrawn product would ignore these distinctions.

The source contract should identify outstanding principal, accrued amounts if relevant, approved limit, amount already drawn, remaining commitment, expiry date, utilisation restrictions and effective dates. The bank must also decide how to handle negative available amounts, over-limit balances, multiple sub-limits and linked facilities. A product can have an overall customer limit with several drawings; adding each sub-limit without recognising the shared cap could double-count the exposure. Conversely, using only the master limit could hide facility-level utilisation needed for validation.

A timeline for a revolving facility

At the observation date in our example, 40,000 is drawn and 60,000 remains under the limit. Over the next months, the borrower draws another 25,000, repays 5,000 and incurs 1,000 of amounts included under the bank's defined exposure measure. Immediately before the recorded default event, the relevant exposure would be 61,000 under this simplified arithmetic. The bank should compare the observed value with the earlier estimate only after checking the methodology's treatment of repayments, fees and timing. It should not assume that the example's earlier 58,000 forecast was intended to predict one customer's exact future balance.

A development dataset needs many such dated observations, not just one. The bank selects an observation date for each facility, the then-current drawn and undrawn amounts, product and limit terms, and a later default event where one occurs. It calculates the exposure at that event under a consistent definition. Accounts without a default still matter to the modelling population and selection process, even though they have no observed default-date balance. The analyst should document how the method estimates exposure conditional on a possible default and how it treats facilities that expire, close or refinance.

Changing a limit shortly before default creates a difficult interpretation. A bank may reduce availability because it has detected deterioration, while a borrower may draw more before restrictions take effect. The model cannot use the later limit action as a feature at an earlier observation date. For validation, the owner should retain the dated history of limit decisions and distinguish contractual availability from actual drawings. A back-test that substitutes the limit visible today would rewrite the exposure path and can make the original forecast seem more accurate than it was.

Conversion factors and capital use

A credit conversion factor (CCF) is a way of turning an undrawn commitment into an exposure equivalent under a specified framework. The illustrative 30% in the worked example is not an actual prescribed factor. The Basel Framework's current CRE32 risk-components chapter sets out how EAD and off-balance-sheet exposures are treated under internal ratings-based approaches, including conditions for using internal estimates of undrawn revolving commitments. A bank must apply the method available to it under its jurisdiction's implementation and supervisory permissions.

The analyst should distinguish the foundation internal ratings-based approach, an advanced approach where permitted, and a standardised approach. Their use of factors and model estimates differs. A facility that is on balance sheet also raises questions about which amount is already recognised, while an undrawn component may need conversion. Treating all commitments as identical or applying an internal estimate where the bank is not permitted to do so can produce a wrong capital calculation. A training chapter should explain the mechanics without inventing a factor for a product.

A capital EAD can also differ from the amount used in a credit decision, liquidity forecast or accounting impairment estimate. The underlying source ledger may be common, but cut-off dates, eligible amounts and risk parameters can be different. The bank should label every EAD with its purpose and calculation version. A report that shows "EAD 58,000" without product, approach, date and included components is not decision-ready.

Expected credit losses and revolving facilities

For expected-credit-loss accounting, the lender estimates cash shortfalls using a probability-weighted, forward-looking method under its applicable accounting standard. The exposure component may evolve through the life of a facility. Under IFRS 9, expected credit losses are not obtained by multiplying today's drawn balance by a PD and an LGD without considering timing, contractual cash flows and the instrument's expected life. Revolving credit facilities can require special attention to the period over which the bank remains exposed to credit risk and the effect of normal risk-management actions. The IFRS Foundation has specific implementation material for revolving facilities.

A 12-month expected-credit-loss calculation refers to defaults possible within the next 12 months, not necessarily losses collected or paid within that same period. For a credit card, additional drawings before a possible default may affect the exposure even when the reporting-date balance is low. Lifetime measurement after a significant increase in credit risk introduces a longer and changing horizon. The bank needs an approved method for forecasting usage, payment and closure. The model should not turn every undrawn pound or rupee into certain future debt, nor should it assume that no borrower will draw more after signs of stress.

The accounting team may group facilities by shared risk characteristics, use scenarios and apply overlays where modelled information is incomplete. Risk and finance should reconcile what each EAD field means. An exposure measure created for prudential capital is not automatically the right accounting input. If two teams produce different numbers for the same portfolio, the discrepancy may be legitimate, but the definition and bridge need to be documented. This chapter does not set accounting policy for a particular bank.

Building a point-in-time EAD dataset

The data pipeline should preserve the facility identifier, customer identifier, source system, observation timestamp, balance, limit, unused portion, product code, currency, status and applicable contract version. For a facility with several linked accounts, the bank needs an aggregation rule that avoids double counting. For a multi-currency line, amounts should be translated under the method appropriate to the use and date; a later exchange rate cannot be silently substituted into an earlier observation. The source of an accrued-fee amount and its inclusion rule should be explicit.

Events do not always arrive in business order. A repayment may post after a card authorisation; a limit change may be approved before it is effective; a correction may be backdated. The feature builder should use effective-time semantics suited to the scoring purpose, while preserving ingestion time for audit. If a batch contains a duplicate transaction, it must not inflate balance. If a reversal arrives later, historical replay should show what the system knew then and also retain the corrected ledger state for investigation. A one-column "latest balance" extract is insufficient for dated EAD validation.

The default event itself must be linked to the right facility and time. A customer with two facilities may default on one while the other remains current, depending on the bank's and framework's default definition. The analyst should not assume that every account balance at a later month-end is the balance at default. Data owners should reconcile the source event, default date, exposure calculation and any subsequent collections posting. A field mapping that passes a schema test can still be economically wrong if it uses the wrong balance or facility hierarchy.

Validation beyond an average error

An EAD method should be challenged on cases with high and low utilisation, unused commitments, recent limit changes, cancellations, redraws and early repayments. The validator compares estimated exposure with observed default-date exposure on a cohort whose outcomes have matured. Aggregate error can cancel out: overestimating one product and underestimating another may produce a reassuring total. Segment results should include counts and the size of exposures, because a handful of large lines can dominate the monetary impact.

The bank should test sensitivity to economic conditions and borrower behaviour. During stress, drawdown patterns can change; a calibration drawn from stable periods may not describe a stressed portfolio. A changed product rule can also alter utilisation independently of the economy. The validator separates model assumptions from operational controls such as limit freezes or manual reviews. If the bank uses a fallback method for a thin or new product, the approval pack should record its limitation and monitoring trigger.

Monitoring should track utilisation, missing limits, over-limit cases, undrawn amounts, product mix and differences between predicted and realised exposure at default. Realised comparisons mature only after defaults occur, while data-quality indicators can be checked sooner. The owner should record whether a movement came from a source defect, policy change, product redesign or customer behaviour. Updating the factor after a single unusual month without enough evidence can add instability rather than improve the estimate.

A case that crosses system boundaries

Take a fictional merchant with a 200,000 working-capital line. A channel shows 90,000 drawn. The credit system has approved a reduction in the line to 150,000, effective tomorrow. A same-day payment of 10,000 is authorised but has not posted to the loan ledger. At today's observation cut-off, the model may need the legally effective 200,000 limit and the balance that its documented source and timing rule define. It cannot take tomorrow's reduced limit merely because the approval exists in a workflow. It also cannot count the payment twice because the authorisation and posting later represent the same economic event.

The analyst maps the four records: contract limit, pending change, authorisation and posted balance. The EAD feature contract states which status each record must have before inclusion. A replay test at today's timestamp should return one answer; a replay after tomorrow's effective time may return another. The result should be explainable without a modeller manually editing the input. If the scoring service lacks a dependable view of the effective limit, the bank may need a controlled referral rather than a fabricated "available" amount.

If this merchant later defaults, the observed exposure is calculated under the bank's defined treatment of principal, accrued amounts and any subsequent adjustments. Collections recoveries are analysed separately for loss severity. The bank preserves the original score and input, the limit-change evidence, the final default-date exposure and later recoveries as distinct events. Collapsing all of them into a single current balance prevents both validation and a fair audit.

Customer, risk and operational consequences

An underestimated EAD can lead the bank to understate potential exposure or misjudge a portfolio's risk, depending on the use. An overestimate can constrain lending or distort pricing. Neither direction can be judged from a single account alone. The owner needs to see distributional effects, product mix and how the estimate feeds an approved decision. If a model flags a revolving line as high exposure, a human reviewer should know whether that comes from current utilisation, undrawn availability, an assumed future drawdown or a data error.

A limit decision has customer impact. Reducing a line can affect a business's ability to pay suppliers, and a delayed update can confuse the customer about available credit. A model's EAD estimate should not silently change contractual availability. The bank's policy and authorised staff decide limit actions, notices and appeals under applicable rules. Analysts should trace model output to the actual action taken, including manual overrides and the reason for a different outcome. The probability of future drawing is a risk input, not evidence that the customer has already borrowed the undrawn amount.

The exposure view also affects operations and reporting. Finance may need impairment inputs, risk may need capital or stress-test numbers, and collections may need the default-date claim and recoveries. These figures can differ legitimately. A reconciliation should state their definitions and bridge the components rather than forcing equality. If a system migration moves limits and balances onto different cut-off schedules, the bank should test reconciliation before relying on the new EAD pipeline.

Boundary with counterparty credit risk

The word exposure also appears in derivatives and securities-financing activity, but a credit-card drawdown method cannot simply be reused there. A derivative can have a value that changes with market prices, collateral, netting and the possibility that a counterparty defaults before final settlement. The Basel Framework's counterparty credit risk chapters set out specific approaches for such exposures. The analyst should identify the instrument and approved methodology before comparing a derivative EAD with a loan EAD.

The same caution applies to guarantees and letters of credit. A contingent obligation is not today's cash advance, yet it can become an exposure under its contractual terms. The data model needs instrument type, commitment amount, maturity, conditions and any conversion treatment required by the relevant use. A banking dashboard may aggregate these figures for risk reporting, but its columns should preserve the source method. Otherwise an apparently precise portfolio total can conceal inconsistent definitions.

For learning purposes, the revolving-line example explains the basic future-exposure idea. It does not prescribe a capital treatment for derivative, trade-finance or off-balance-sheet products. A bank's capital, finance and credit-risk specialists must approve the method for each product and jurisdiction.

Analyst acceptance tests and evidence

An acceptance set should include a fully drawn term loan, a partially drawn revolving line, an expired commitment, a pending limit reduction, an over-limit balance, a linked sub-limit and a closed facility with late transactions. Each case specifies the observation timestamp, source records, expected exposure components and calculation purpose. A missing limit should produce an explicit exception or approved fallback. The system should not use zero merely because a field failed to arrive.

Tests also need temporal edges. Re-score immediately before and after a limit's effective time, a repayment posting, a currency conversion and a default event. Confirm that a later correction does not rewrite the original scored record. Confirm that a balance captured after default is not treated as the pre-default estimate. An engineer can verify arithmetic, while the business analyst checks whether the records represent the correct economic event and contractual exposure.

The final evidence pack should let a reviewer move from a reported EAD to the model version, calculation method, limit history, balances, conversion assumptions, default event and approval record. It should show what the bank knew at the observation time and what became known later. That distinction lets risk, finance and operations challenge the same number without conflating projected exposure, legal commitment, outstanding balance and realised loss.

Banking practice note on data definition

For exposure at default, data definition matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

The practical discipline is to keep source fact, credit-risk interpretation, model estimate and business action separate. Source facts come from areas such as outstanding balance, undrawn commitment, credit limit, utilisation trend, and overdraft usage. The credit-risk interpretation explains what those facts mean. The model estimate turns that interpretation into a score, probability, severity, exposure or ranking. The business action decides approval, referral, monitoring, support, pricing, limit or reporting treatment.

This separation helps the bank challenge the result properly. When a score changes, the bank can ask whether borrower behaviour changed, source data changed, policy changed, model calibration changed or economic conditions changed. That is the kind of clarity serious banking AI needs.

Banking practice note on model use

For exposure at default, model use matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

Banking practice note on customer impact

For exposure at default, customer impact matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

Banking practice note on portfolio control

For exposure at default, portfolio control matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

Banking practice note on validation

For exposure at default, validation matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

Banking practice note on monitoring

For exposure at default, monitoring matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

Banking practice note on governance

For exposure at default, governance matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

Banking practice note on audit evidence

For exposure at default, audit evidence matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

Banking practice note on economic conditions

For exposure at default, economic conditions matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

Banking practice note on human judgement

For exposure at default, human judgement matters because credit risk is not only a model output. It is a banking responsibility that connects borrower behaviour, product policy, capital, provisions, customer treatment, collections, finance reporting, risk appetite and audit evidence. If one part is weak, the model result can look precise while the decision remains fragile.

The balance can move before default

Exposure at default estimates the amount outstanding at the specified default event, not always today's booked balance. A term loan that amortizes and a revolving facility with undrawn capacity behave differently. For a simple illustration, a card has 2,000 currently drawn and 3,000 available; the customer may draw more before default. Treating EAD as 2,000 for every horizon ignores that possibility, while assuming all 5,000 will be drawn can overstate it. State the product, horizon and permitted conversion method.

Build a cohort with dated facility limits, balances, payments, drawdowns and default events. Check that limit changes after the observation cutoff do not leak into a historic estimate. A customer who closes the line early is not a default at zero exposure. Missing facility records should be investigated, not defaulted to zero. Reconcile balances to source ledgers and ensure one facility is not duplicated by an account migration.

The EAD estimate affects expected loss alongside PD and LGD, but the components may be correlated and use different data conventions. Compare a small term loan and a large revolving line at the same reported PD, and explain why potential exposure differs. Validate estimates against observed mature defaults, segment by product and document uncertainty where draws or limits change. A model output becomes useful only when its exposure definition matches the decision and reporting purpose.