14. Accounting and Financial Reporting for Treasury

Every trade, hedge and funding decision that treasury makes ends up as numbers in the bank's ledgers, financial statements and management reports. This chapter covers how those numbers are produced: how IFRS 9 decides the measurement of treasury positions, how lifecycle events become journal entries and flow to the general ledger, how hedge accounting is implemented in systems, and how management reports are built and controlled. For an engineer, accounting is where small modelling choices become reported profit, equity and capital, so it rewards precision more than almost any other part of the treasury stack.

<a id="141-treasury-accounting"></a>

14.1 Treasury Accounting

This subchapter explains how treasury positions are classified and measured, how interest and fair values are calculated for accounting, and how business events are turned into postings that reach the general ledger.

IFRS 9 Classification and Measurement

Why it matters. Three banks can buy the same government bond at the same price on the same day and, a year later, report three different results from it: one shows a loss in profit or loss, one shows a loss only in equity, and one shows no loss at all. The difference is not in the bond but in its accounting classification, which is decided on day one and is very hard to change afterwards. For treasury, classification decides how much volatility the liquidity buffer brings into earnings and capital, and which positions can be sold without consequences. For engineers, it decides which valuation and posting logic every position flows through.

The two tests. IFRS 9 classifies debt-type financial assets by answering two questions:

1. Business model test: how does the bank manage this group of assets to generate cash flows? It is assessed at the level of a portfolio (in practice, a book), not of a single instrument, and must reflect how the portfolio is actually managed, reported to management and remunerated.
2. SPPI test (solely payments of principal and interest): are the contractual cash flows only repayments of principal and interest on the principal outstanding? "Interest" here means compensation for the time value of money, credit risk, other basic lending risks and a profit margin. It is assessed instrument by instrument.

                        Financial asset (debt instrument)
                                     |
                      Do the cash flows pass the SPPI test?
                         |                          |
                         No                         Yes
                         |                          |
                       FVTPL                What is the business model?
                                  +-----------------+-----------------+
                                  |                 |                 |
                          Hold to collect    Hold to collect     Other (trading,
                                  |            and sell          managed on a
                                  |                 |            fair value basis)
                                  v                 v                 v
                           Amortized cost         FVOCI             FVTPL

   Option: designate at FVTPL (fair value option) if this removes an accounting mismatch.

Business models.

Business model	Description	Treasury example	Sales
Hold to collect (HTC)	Assets held to collect contractual cash flows	Bonds bought to match long-term fixed-rate deposits; loans	Allowed if infrequent, insignificant in value, close to maturity or caused by credit deterioration
Hold to collect and sell (HTC&S)	Collecting cash flows and selling are both integral	Actively managed liquidity buffer, sold to meet liquidity needs or rebalance duration	Expected and frequent
Other	Trading, or managed and evaluated on a fair value basis	Trading book, short-term positions	Integral to the model

SPPI in practice.

Contractual feature	SPPI?	Reason
Fixed coupon, or floating coupon such as Euribor + spread	Pass	Compensation for time value and credit
Prepayment at par plus accrued interest	Pass	Prepayment amount reflects principal and interest
Coupon linked to an equity index or commodity price	Fail	Exposure to risks unrelated to lending
Convertible bond	Fail	Payoff depends on the issuer's share price
Leveraged coupon, such as 3 x Euribor	Fail	Amplifies rate variability beyond time value
Rate tenor mismatch, such as a 1-year rate reset monthly	Depends	Requires a "modified time value" benchmark test

Equity instruments do not have principal and interest. They are measured at FVTPL unless the bank makes an irrevocable election to present changes in OCI.

Measurement categories.

Category	Balance sheet	Interest income	Fair value changes	On sale	Expected credit loss (ECL)
Amortized cost	Amortized cost less loss allowance	Effective interest rate (EIR) in P&L	Not recognized	Gain or loss to P&L	Yes, allowance reduces the asset
FVOCI (debt)	Fair value	EIR in P&L	OCI	Cumulative OCI recycled to P&L	Yes, expense in P&L, allowance held in OCI
FVOCI (equity election)	Fair value	Dividends in P&L	OCI	Never recycled; transferred within equity	No
FVTPL	Fair value	Usually in P&L (presentation choice)	P&L	Already in P&L	No

Financial liabilities and derivatives. Deposits and issued debt are measured at amortized cost by default. Trading liabilities and derivatives are at FVTPL. If a liability is designated under the fair value option, the part of its fair value change caused by the bank's own credit risk goes to OCI, so the bank does not report a profit when its own credit quality worsens. Derivatives are always at FVTPL unless they are designated as hedging instruments (Section 14.2).

Expected credit loss. IFRS 9 also replaced the "incurred loss" model with an expected loss model for assets at amortized cost and debt at FVOCI.

Stage	Trigger	Loss allowance	Interest revenue
1	Performing, no significant increase in credit risk (SICR) since initial recognition	12-month ECL	EIR on gross carrying amount
2	SICR since initial recognition	Lifetime ECL	EIR on gross carrying amount
3	Credit-impaired	Lifetime ECL	EIR on net carrying amount (after allowance)

Treasury buffer bonds are usually high-grade, so they sit in Stage 1 with small allowances. Allowances can still matter for large sovereign and bank exposures.

Example: the same loss in three books. A bank buys EUR 500m of 10-year government bonds at par in each of three books. Rates rise and the price falls to 92.

Book	Balance sheet value	P&L	OCI	CET1 impact
Amortized cost (HTC)	500m	0	0	0 (fair value of 460m is only disclosed)
FVOCI (HTC&S)	460m	0	-40m	-40m
FVTPL (trading)	460m	-40m	0	-40m

The economic loss is the same in all three books. Classification only changes when and where the loss is reported. If the bank has to sell the amortized-cost bonds to raise cash, the 40m loss goes to P&L immediately. If sales out of the HTC book become more than infrequent, auditors may also question whether the remaining HTC portfolio still meets its business model. This combination of hidden losses and a forced sale was central to the failure of Silicon Valley Bank in 2023 (Chapter 16.4).

US GAAP comparison.

IFRS 9	US GAAP	Main difference
Amortized cost	Held to maturity (HTM)	HTM requires positive intent and ability to hold to maturity; sales "taint" the whole category
FVOCI (debt)	Available for sale (AFS)	Similar mechanics; US capital rules let some smaller banks exclude AOCI from capital
FVTPL	Trading	Similar
Three-stage ECL	CECL	CECL requires lifetime expected losses from day one, with no staging

Reclassification. Financial assets are reclassified only when the business model itself changes, which should be very rare. The change applies prospectively from the first day of the next reporting period. Financial liabilities are never reclassified.

Engineering view.

• Store the inputs of classification, not only the result: the business model is configured per book with effective dates, and the SPPI assessment is stored on the instrument's reference data with evidence and an approver. The category of each position is derived from both.
• Classification is fixed at initial recognition for each lot. Moving a bond between books with different business models must not be a silent "book transfer": it is either a sale and repurchase or a reclassification event that needs accounting policy approval.
• Monitor sales out of HTC books against frequency and value thresholds, and alert accounting policy before a sale, not after it.
• A position can have different categories under different frameworks (IFRS, local GAAP, US GAAP). Model classification per framework from the start.

Analogy. Classification is like the wrapper you put an investment in: the same shares can be held in a pension account, a tax-free savings account or an ordinary brokerage account. The market value is identical, but gains are taxed, reported and withdrawn differently, and you choose the wrapper when you put the money in.

Key takeaways

• IFRS 9 classifies debt assets with the business model test (per portfolio) and the SPPI test (per instrument).
• The result is amortized cost, FVOCI or FVTPL; derivatives are FVTPL unless they are designated in a hedge relationship.
• Classification changes when and where gains and losses are reported, not their economic size.
• Expected credit losses apply to amortized cost and FVOCI debt, in three stages.
• Systems should store classification inputs per book and instrument, per framework, and prevent silent transfers between business models.

🧪 Practice

1. Beginner: Classify each position: (a) German government bonds in the liquidity buffer that are regularly sold to manage liquidity; (b) a corporate loan held until maturity; (c) a bond whose coupon is linked to the EURO STOXX 50; (d) an interest rate swap that is not designated in a hedge.
2. Intermediate: Treasury holds EUR 2bn in an HTC portfolio. In one quarter it sells EUR 300m, five years before maturity, to fund deposit outflows. Explain the accounting consequences for the bonds sold and the risk for the remaining portfolio. Which data would you prepare for the auditors?
3. Interview: "Design the data model for IFRS 9 classification in a treasury system that also reports under local GAAP." Hint: separate the inputs (business model per book with effective dates, SPPI assessment per instrument with evidence and approver) from the derived output (category per position per framework), keep history, make derivations reproducible, and block book transfers that would change the category without an approved event.

Amortized Cost and Effective Interest Rate

Why it matters. Suppose treasury buys a bond with a 3% coupon at 97.22. It will receive the coupons and also 100 at maturity, so the 2.78 discount is part of its return. Recognizing the discount only at maturity would understate income for years and then show a jump. Recognizing only the coupons would ignore the true yield. The effective interest method spreads the total return evenly in yield terms: each period's interest income is the same percentage of the carrying amount.

Definitions.

• Effective interest rate (EIR): the rate that exactly discounts the estimated future cash flows over the expected life of the instrument to its initial carrying amount. It includes fees and transaction costs that are integral to the instrument (arrangement fees, transaction costs, premiums and discounts) and excludes expected credit losses.
• Amortized cost: the initial amount, minus principal repayments, plus or minus the cumulative amortization (calculated with the EIR) of the difference between the initial amount and the maturity amount, minus the loss allowance.
• Gross carrying amount: amortized cost before deducting the loss allowance.

  Initial carrying amount = sum over t of  CF_t / (1 + EIR)^t

  Each period:
    Interest income   = Opening carrying amount x EIR
    Amortization      = Interest income - Cash coupon
    Closing carrying  = Opening carrying amount + Interest income - Cash received

Example. Treasury buys a 3-year bond with face value 1,000,000 and an annual coupon of 3% for 972,249.09. The EIR that equates the price with the cash flows is 4.00%.

Year	Opening amortized cost	Interest income (4.00%)	Cash received	Discount amortization	Closing amortized cost
1	972,249.09	38,889.96	30,000.00	8,889.96	981,139.05
2	981,139.05	39,245.56	30,000.00	9,245.56	990,384.62
3	990,384.62	39,615.38	1,030,000.00	9,615.38	0.00
Total		117,750.91	1,090,000.00	27,750.91

Total interest income equals the coupons (90,000) plus the discount (27,750.91). Straight-line amortization would recognize 9,250.30 of discount every year. IFRS allows straight-line only where the result is not materially different. Here the difference is small, but it grows with the size of the discount or premium and the length of the instrument.

Changes in estimated cash flows. For instruments with uncertain cash flows (prepayable mortgages, callable bonds), the EIR is based on expected cash flows. When the estimate changes, the gross carrying amount is recalculated as the present value of the revised cash flows at the original EIR, and the difference goes to P&L immediately as a catch-up adjustment. Continuing the example, suppose at the end of year 1 the issuer announces that it will call the bond at par at the end of year 2:

  Revised carrying amount = 1,030,000 / 1.04 = 990,384.62
  Current carrying amount =                    981,139.05
  Catch-up gain in P&L                     =     9,245.57

For floating-rate instruments, the regular reset of the coupon to market rates simply changes the EIR going forward, with no catch-up adjustment.

Solving for the EIR. The EIR is the internal rate of return of the instrument's cash flows. A robust approach is bisection, because the present value falls as the rate rises:

from decimal import Decimal, getcontext

getcontext().prec = 28

def solve_eir(price: Decimal, cash_flows: list[Decimal]) -> Decimal:
    """Annual EIR such that the PV of cash_flows (one per year) equals price."""
    lo, hi = Decimal("-0.99"), Decimal("1")
    for _ in range(200):
        mid = (lo + hi) / 2
        pv = sum(cf / (1 + mid) ** (t + 1) for t, cf in enumerate(cash_flows))
        if pv > price:      # PV too high -> rate too low
            lo = mid
        else:
            hi = mid
    return (lo + hi) / 2

solve_eir(Decimal("972249.09"), [Decimal(30000), Decimal(30000), Decimal(1030000)])
# Decimal('0.0399999998...')  -> 4.00%

Real instruments have irregular periods and day count conventions (Chapter 11.1), so production code solves on actual dates, in the same way as an XIRR calculation.

Engineering view.

• Store the EIR per lot, computed at initial recognition including transaction costs and integral fees. Version it when estimated cash flows are revised.
• Generate EIR schedules with the same cash flow generator used for pricing, so that schedule dates, day counts and amounts are consistent with the rest of the system.
• Systems accrue daily. Post rounded daily amounts but carry the rounding residual forward, so the total over the life of the instrument equals the unrounded schedule exactly (Chapter 11.4). At maturity, the carrying amount must equal the redemption amount. Alert if the final residual exceeds a tolerance.
• Partial sales need lot tracking (FIFO or weighted average, set by policy), because each lot has its own EIR and amortized cost.
• The same method applies to liabilities: issued debt is carried at amortized cost, with issuance costs and discounts amortized into interest expense.

Analogy. The EIR is like spreading a gym's joining fee over the months of membership to find the true monthly cost. The joining fee is a cost of the whole contract, not of the first month.

Key takeaways

• The EIR discounts expected cash flows to the initial carrying amount, including integral fees and transaction costs.
• Interest income is the carrying amount times the EIR; the difference from the coupon amortizes the premium or discount.
• Revised cash flow estimates create a catch-up adjustment at the original EIR; floating-rate resets change the EIR prospectively.
• Production systems accrue daily, track lots and must manage rounding so the schedule closes exactly at maturity.

🧪 Practice

1. Beginner: A bond with a 5% coupon is bought at 102. Will annual interest income under the EIR method be higher or lower than the coupon? Explain why.
2. Intermediate: A 2-year bond with face value 500,000 and a 2% annual coupon is bought for 490,000 plus 1,000 of transaction costs. Use solve_eir (or a spreadsheet) to find the EIR and build the amortization schedule. Check that the carrying amount reaches zero after the final payment.
3. Interview: "The expected prepayment rate on a mortgage portfolio is revised from 5% to 10% per year. What happens to amortized cost, and how should the system process the change?" Hint: re-project cash flows with the new assumption, discount at the original EIR, post the catch-up to P&L, note that deferred fees and premiums now amortize faster, version the schedules and keep an audit trail of the assumption change and its approval.

Fair Value Hierarchy

Why it matters. Everything measured at FVTPL or FVOCI, including every derivative, needs a fair value at each reporting date. For a liquid government bond, that value is a screen price. For a 15-year cross-currency swap with an amortizing notional, there is no quote: the value comes from a model whose inputs may themselves not be observable. Readers of financial statements need to know how much of the reported value rests on observed prices and how much on management judgment. IFRS 13 (and ASC 820 in US GAAP) defines fair value and requires every fair value measurement to be classified into one of three levels.

What fair value is. Fair value is the price that would be received to sell an asset, or paid to transfer a liability, in an orderly transaction between market participants at the measurement date. It is an exit price in the principal (or most advantageous) market. It is not the price the bank paid, not a forced-sale price, and it uses the assumptions of market participants rather than the bank's own.

The three levels.

Level	Inputs	Treasury examples	Judgment
1	Unadjusted quoted prices in active markets for identical instruments	Actively traded government bonds, exchange-traded futures	Low
2	Observable inputs other than Level 1 prices, direct or indirect	Vanilla interest rate swaps and FX forwards valued from observable curves; less liquid bonds priced from quotes of similar bonds	Medium
3	At least one significant unobservable input	Derivatives beyond observable tenors, illiquid structured notes, private placements	High

The level of a measurement is set by the lowest-level input that is significant to the whole value:

  Is there a quoted price for the identical instrument in an active market?
      Yes --> Level 1 (use it, unadjusted)
      No  --> Valuation technique. Are all significant inputs observable?
                  Yes --> Level 2
                  No  --> Level 3

  "Significant": the unobservable input moves the value by more than a policy
  threshold (for example 5-10% of the position's value) under reasonable alternatives.

Example: a treasury portfolio.

Position	Valuation approach	Key inputs	Level
EUR 2bn German government bonds	Dealer and exchange quotes, actively traded	Quoted price	1
10-year EUR interest rate swaps (hedges)	Discounted cash flows	€STR and Euribor curves, observable	2
Covered bond that rarely trades	Matrix pricing from similar bonds	Spreads of comparable issues	2
40-year GBP inflation swap	Model	Inflation curve beyond observable tenor	3 if significant
Issued callable structured note	Model	Long-dated volatility, mean reversion	3

Valuation adjustments and price verification. The reported fair value is the model or mid value plus adjustments that a market participant would price in: bid-offer (close-out cost), counterparty and own credit (CVA and DVA, Chapter 8.3), funding (FVA) and model uncertainty reserves. Product control performs independent price verification (IPV): it compares front-office marks with independent sources and adjusts differences above tolerance. In the EU, regulators additionally require prudent valuation adjustments (AVAs), which are deducted from CET1 but do not change the accounting value.

Day-one P&L. If a trade's fair value at inception relies on Level 3 inputs and differs from the transaction price, the difference cannot be recognized immediately. It is deferred and released as the inputs become observable, or over the life of the trade. Example: the bank sells a structured swap whose model value implies a gain of EUR 2.0m, but the model uses an unobservable correlation. The 2.0m is deferred and, under a straight-line policy over the 5-year life, released at 0.4m per year.

Disclosures. Banks disclose fair values by level, transfers between levels and their reasons, a Level 3 roll-forward (opening balance, gains and losses, purchases, sales, settlements, transfers in and out, closing balance) and the sensitivity of Level 3 values to reasonably possible alternative assumptions.

Engineering view.

• Store per position and date: fair value, price source, valuation adjustments, level, and the evidence behind the level (number of quotes, staleness, bid-ask width).
• Maintain an observability matrix for market data (for example, EUR swap rates observable to 50 years, a volatility surface to 10 years). Derive levels by rule from the inputs each trade uses and its sensitivity to them. Log manual overrides with approver and reason.
• Keep level history: the Level 3 roll-forward needs the date and value of every transfer.
• Track deferred day-one P&L per trade with its release schedule and release trigger.

Analogy. Level 1 is reading the price tag in a busy shop. Level 2 is valuing your car from listings of similar cars. Level 3 is valuing a one-off custom car that nobody has ever sold: you need a model and assumptions, and anyone relying on the number should know that.

Key takeaways

• Fair value is an exit price under market participants' assumptions, not the transaction price or the bank's own view.
• Level 1 uses quoted prices for identical instruments, Level 2 uses observable inputs, Level 3 depends on significant unobservable inputs.
• The level is set by the lowest-level significant input; significance is tested against a policy threshold.
• Day-one P&L on Level 3 trades is deferred; Level 3 positions require a roll-forward and sensitivity disclosures.
• Levelling should be rule-based, evidence-backed and historized.

🧪 Practice

1. Beginner: Assign a level to each: an exchange-traded Bund future; a 5-year EUR/USD cross-currency swap; an unlisted bond of a small issuer priced with a model that uses an internally estimated credit spread.
2. Intermediate: A position is valued at EUR 100m. Replacing an unobservable correlation input with reasonable alternatives changes the value by up to EUR 3m, and the significance threshold is 5% of value. Which level applies? What changes if the impact is EUR 8m? Sketch how you would implement this test.
3. Interview: "How would you produce the Level 3 roll-forward disclosure from system data?" Hint: level per position per date, lifecycle events (purchases, sales, settlements), the P&L and OCI split, transfers recorded on a consistent convention (beginning or end of period values), and a reconciliation of the closing balance to the balance sheet.

Trade-Date vs Settlement-Date Accounting

Why it matters. A bond bought on 29 September settles two business days later, on 1 October. The quarter ends on 30 September. Is the bond on the quarter-end balance sheet? The answer changes the reported size of the balance sheet, which day's price movements hit P&L, and how pending trades must be tracked. At a quarter end with billions of euros of pending purchases and sales, the difference is material.

The rule. IFRS 9 allows a regular-way purchase or sale of a financial asset (one delivered within the time frame set by market convention, such as T+2 for bonds) to be recognized either on trade date or on settlement date. The choice must be applied consistently to all purchases and sales of assets in the same category. Trades that settle later than market convention, and all derivatives, are recognized on trade date as derivatives until settlement.

Aspect	Trade-date accounting	Settlement-date accounting
Asset recognized	Trade date	Settlement date
Liability to pay	Recognized on trade date (settlement payable)	Not recognized before settlement
Fair value change between trade and settlement	Recognized as usual for the category	Recognized for FVTPL (P&L) and FVOCI (OCI); not for amortized cost
Interest accrual	Generally from settlement date, when title passes	From settlement date
Derecognition on sale	Trade date	Settlement date
Period-end balance sheet	Shows pending purchases and their payables	Smaller; pending trades not shown gross

Example. Treasury buys EUR 10m nominal of an FVTPL bond at 99.50 on Tuesday 29 September 2026 for settlement on Thursday 1 October. At the quarter end on 30 September, the price is 99.80. Accrued interest is ignored for simplicity.

  29 Sep (trade)          30 Sep (quarter end)           1 Oct (settlement)
       |------------------------|------------------------------|
   price 99.50              price 99.80                 cash 9,950,000 paid,
   cost 9,950,000           value 9,980,000             bond delivered

Balance at 30 September	Trade-date	Settlement-date
Bond (asset)	9,980,000	not recognized
Fair value change on pending purchase (asset)	not applicable	30,000
Settlement payable (liability)	9,950,000	not recognized
P&L for the quarter	+30,000	+30,000

The P&L is the same. The balance sheet under trade-date accounting is EUR 9.95m larger. If the bond had been at amortized cost, trade-date accounting would show the bond and payable at 9,950,000 with no P&L, and settlement-date accounting would show nothing until 1 October.

Engineering view.

• Every event needs at least three dates: trade date, settlement (value) date and accounting (posting) date. The recognition date is derived by rule from the classification and the accounting policy.
• Settlement payable and receivable accounts must reconcile to the back office's list of unsettled trades. Failed settlements stay in these accounts and need ageing reports.
• Trades booked late with a trade date in a closed period must follow a defined cut-off policy: post in the open period with the original effective date, or reopen the period under approval.
• Liquidity reporting (Chapter 5.2) uses settlement-date cash flows whatever the accounting policy. Do not let one feed serve both purposes without making the date basis explicit.

Analogy. Buying a house has a signing date (trade date) and a completion date when you get the keys (settlement date). Trade-date accounting treats the house as yours when you sign. Settlement-date accounting waits for the keys, but if the house's value changes in between, you are already committed, so that change is still yours.

Key takeaways

• Regular-way purchases and sales can be recognized on trade date or settlement date, consistently per category.
• Under settlement-date accounting, value changes between trade and settlement are still recognized for FVTPL and FVOCI assets.
• Derivatives and non-regular-way trades are always recognized on trade date.
• Systems must carry trade, settlement and accounting dates, and reconcile pending settlement accounts to back-office data.

🧪 Practice

1. Beginner: Explain why settlement-date accounting still recognizes the fair value change between trade date and settlement date for an FVTPL bond.
2. Intermediate: Treasury sells EUR 50m of FVOCI bonds on 30 June at 101.20, for settlement on 2 July. The bonds have an amortized cost of EUR 50.2m and a fair value adjustment of EUR 0.3m held in OCI. Describe the postings on 30 June and 2 July under trade-date accounting, including the recycling of OCI.
3. Interview: "At quarter end, the balance sheet shows EUR 400m of securities settlement payables, but the back office reports only EUR 250m of unsettled purchases. How do you investigate?" Hint: reconcile at trade level; look for cancelled or amended trades whose postings were not reversed, settlement events that were not posted, failed settlements, different cut-off times, trade-date versus value-date mismatches and trades booked to the wrong entity.

Accounting Events and Postings

Why it matters. A treasury system thinks in trades and lifecycle events: booked, settled, fixed, matured. Finance thinks in debits and credits on general ledger accounts. The bridge is the accounting event model: every business event that changes the financial position produces an accounting event, and an accounting rules engine turns it into balanced journal entries (Chapter 1.4). If this bridge is designed well, the ledger reconciles automatically. If it is not, month-end becomes days of manual fixes.

Typical treasury accounting events.

Event	Trigger	Typical postings
Trade booking	Trade captured (trade-date accounting)	Asset vs settlement payable
Settlement	Cash and securities move	Settlement payable vs nostro
Interest accrual	Daily end of day	Accrued interest and amortization vs interest income
Coupon or interest payment	Payment date	Nostro vs accrued interest receivable
Revaluation	Daily (FVTPL) or period end (FVOCI)	Fair value adjustment vs P&L or OCI
FX revaluation	Daily or period end	Foreign-currency monetary items revalued to functional currency, difference to P&L
Impairment (ECL)	Period end or stage change	ECL expense vs loss allowance (amortized cost) or OCI (FVOCI)
Amendment	Trade changed	Reverse the original postings, post the new ones
Cancellation	Trade removed	Reverse all postings
Sale or maturity	Disposal	Derecognize carrying amount, realized gain or loss, recycle OCI

Worked example: the life of an FVOCI bond. Treasury buys EUR 10m nominal of a 5-year bond with a 3% annual coupon at 99.50, on a coupon date (so there is no accrued interest to pay). The EIR is 3.1095%. The bank uses trade-date accounting and, for simplicity, ACT/365 daily accruals.

Account	Name
110200	Nostro: EUR
120100	Debt securities FVOCI: amortized cost
120190	Debt securities FVOCI: fair value adjustment
120900	Accrued interest receivable: securities
190100	Securities settlement payable
330100	OCI reserve: debt instruments at FVOCI
410200	Interest income: securities
450100	Realized gains and losses: securities

  Event                          Dr/Cr  Account                                          EUR
  ---------------------------------------------------------------------------------------------
  1. Trade booked (T)            Dr     120100 Debt securities: amortized cost    9,950,000.00
                                 Cr     190100 Securities settlement payable      9,950,000.00

  2. Settlement (T+2)            Dr     190100 Securities settlement payable      9,950,000.00
                                 Cr     110200 Nostro: EUR                        9,950,000.00

  3. Daily interest accrual      Dr     120900 Accrued interest receivable              821.92
     (9,950,000 x 3.1095%/365    Dr     120100 Debt securities: amortized cost           25.74
      = 847.66)                  Cr     410200 Interest income: securities              847.66

  4. Month-end revaluation       Dr     120190 Debt securities: FV adjustment        29,227.63
     (fair value 9,980,000 vs    Cr     330100 OCI reserve: debt at FVOCI            29,227.63
      amortized cost 9,950,772.37 after 30 days of accruals)

  5. Sale at 99.90, 60 days after settlement (amortized cost 9,951,544.73)
     a. Revalue to sale price    Dr     120190 Debt securities: FV adjustment         9,227.64
        (38,455.27 - 29,227.63)  Cr     330100 OCI reserve: debt at FVOCI             9,227.64
     b. Derecognize              Dr     110200 Nostro: EUR                        9,990,000.00
                                 Cr     120100 Debt securities: amortized cost    9,951,544.73
                                 Cr     120190 Debt securities: FV adjustment        38,455.27
     c. Recycle OCI to P&L       Dr     330100 OCI reserve: debt at FVOCI            38,455.27
                                 Cr     450100 Realized gains: securities            38,455.27

Notes on the example:

• In step 3, interest income is the rounded EIR amount (847.66), and the coupon accrual is rounded separately (821.92), so amortization is posted as the difference (25.74). Daily rounding residuals are carried forward, which is why the 30-day amortization totals 772.37 rather than 30 x 25.74 (Chapter 11.4).
• The accrued interest receivable is settled by the dirty sale proceeds; those postings are omitted for brevity.
• After step 5, accounts 120100, 120190 and 330100 are all back to zero for this position, which is a useful automated check.

Design choices.

Choice	Option A	Option B
Revaluation postings	Reverse and replace: reverse yesterday's fair value adjustment and post today's full amount. Simple and self-correcting, but doubles posting volume	Incremental: post only the change. Fewer postings, but an error persists until someone corrects it
Amendments	Reverse and rebook all postings of the trade. Transparent and easy to audit	Delta postings for the changed amounts only. Fewer lines, harder to follow
Events in closed periods	Post in the current open period with the original effective date recorded	Reopen the period (rare, needs approval)

A minimal rules engine. Posting rules map an event type and classification to debit and credit accounts. A negative amount swaps the sides:

from dataclasses import dataclass
from decimal import Decimal

@dataclass(frozen=True)
class Posting:
    account: str
    debit: Decimal
    credit: Decimal

# (event type, classification) -> [(debit account, credit account, amount field)]
RULES = {
    ("TRADE_BOOKED", "FVOCI"):     [("120100", "190100", "consideration")],
    ("SETTLED", "FVOCI"):          [("190100", "110200", "consideration")],
    ("INTEREST_ACCRUED", "FVOCI"): [("120900", "410200", "coupon_accrual"),
                                    ("120100", "410200", "amortization")],
    ("REVALUED", "FVOCI"):         [("120190", "330100", "fv_change")],
}

def postings_for(event: dict) -> list[Posting]:
    rule = RULES.get((event["type"], event["classification"]))
    if rule is None:
        raise LookupError(f"no posting rule for {event['type']}/{event['classification']}")
    result = []
    for debit_account, credit_account, field in rule:
        amount = Decimal(event[field])
        if amount < 0:  # a fair value loss reverses the sides
            debit_account, credit_account, amount = credit_account, debit_account, -amount
        result.append(Posting(debit_account, amount, Decimal(0)))
        result.append(Posting(credit_account, Decimal(0), amount))
    assert sum(p.debit for p in result) == sum(p.credit for p in result)
    return result

In production, an event without a rule is routed to a suspense queue rather than failing the whole batch (Chapter 1.4), and rules are effective-dated configuration rather than code.

Engineering view.

• Give every accounting event a unique ID and make posting idempotent: replaying the event feed must not create duplicate entries (Chapter 12.4).
• Never update or delete posted entries. Corrections are reversals and new postings, so the ledger remains an append-only history.
• Version posting rules with effective dates, so that any historical posting can be explained by the rule in force at the time.
• One event can produce postings in several ledgers (IFRS, local GAAP, tax). Keep the event independent of the framework and apply framework-specific rule sets.
• Check every journal for balance per entity and currency before it leaves the subledger.

Analogy. Accounting events work like a translator between two departments. The kitchen (the treasury system) says what it cooked; the translator (the rules engine) turns this into the entries that the accountant needs, using a fixed dictionary that both sides have agreed on.

Key takeaways

• Business events become accounting events, and a rules engine turns them into balanced postings.
• A complete event model covers booking, settlement, accrual, payment, revaluation, FX revaluation, impairment, amendment, cancellation and disposal.
• On sale of an FVOCI asset, the cumulative OCI is recycled to P&L.
• Posting must be idempotent and append-only, with effective-dated rules and suspense handling for unmapped events.

🧪 Practice

1. Beginner: Using the chart of accounts above, write the postings for the receipt of the first annual coupon of 300,000 EUR, assuming the full amount has already been accrued.
2. Intermediate: A bond purchase booked yesterday with a nominal of EUR 10m is amended today to EUR 12m. List the postings under reverse-and-rebook. What changes if yesterday was the last day of a period that is now closed?
3. Interview: "The finance team configures the accounting rules engine. How would you make rule changes safe?" Hint: effective-dated, versioned rules; maker-checker approval; a test harness of golden events with expected postings; coverage checks that every event, product and classification combination has a rule; a parallel run before activation; an impact report of postings that would change.

Subledger to General Ledger Flow

Why it matters. Chapter 1.4 introduced the chart of accounts, the general ledger (GL) and subledgers. A treasury subledger can produce millions of postings per day: daily accruals and revaluations for every trade. The GL is not designed to hold this detail, and should not. The subledger-to-GL flow aggregates postings, transfers them reliably and keeps the link so that any GL balance can be explained down to individual trades. At month end, this flow is on the critical path of the financial close.

The flow.

  Treasury system,                 Accounting subledger                              General ledger
  core banking, payments
  -----------------------          ----------------------------------------         ---------------------
  business events  -- feed -->  [1] Validate and enrich: static data,
                                    classification, GL segments
                                [2] Rules engine -> postings per trade
                                [3] Balances per trade and account
                                [4] Aggregate by GL key and accounting date
                                [5] Journal batch with control totals    ------>  [6] Import, validate,
                                                                                      post or reject
                                [7] Drill-back: GL line -> journal -> postings -> events -> trades

                                Unmapped or invalid events -> suspense account + exception queue

Aggregation. Postings are summarized by the GL key: entity, account, cost centre, product, currency and accounting date (the segments of Chapter 1.4).

Trade	Entity	Account	Cost centre	Product	Currency	Debit	Credit
T-1001	IT01	120900	CC7100	GOVBOND	EUR	821.92
T-1002	IT01	120900	CC7100	GOVBOND	EUR	1,369.86
T-1003	IT01	120900	CC7100	GOVBOND	EUR	273.97
T-1001 to T-1003 (3 postings)	IT01	410200	CC7100	GOVBOND	EUR		2,465.75

These six postings become two GL lines: IT01-120900-CC7100-GOVBOND-EUR debit 2,465.75 and IT01-410200-CC7100-GOVBOND-EUR credit 2,465.75. Across a full day, a treasury subledger might reduce several million postings to tens of thousands of GL lines.

Control totals. Each journal batch carries a header with a batch ID, the number of lines, total debits and total credits per currency, and often a hash. The GL confirms receipt with the same figures. A batch that does not balance or does not match its header is rejected as a whole, never posted partially.

Multi-currency and multi-GAAP. Postings are made in transaction currency and converted to the entity's functional currency. At period end, foreign-currency monetary items (bonds, deposits, accrued interest) are revalued at the closing rate, and the difference goes to P&L. For FVOCI debt, the FX difference on the amortized cost goes to P&L and the rest of the fair value change to OCI. Banks reporting under several frameworks keep parallel ledgers, each fed by its own rule set from the same events.

The month-end close.

  Working day   Activity
  -----------   ---------------------------------------------------------------------
  WD-2          Freeze static data changes; confirm market data cut-off times
  WD0           Last business day: end-of-day batch, final accruals, month-end revaluation
  WD+1          Subledger close; journal batches to GL; subledger-to-GL reconciliation
  WD+2          Manual adjustments with approval; intercompany matching
  WD+3          GL close per entity; consolidation and eliminations
  WD+4 to +6    Financial statements, FINREP/COREP (Chapter 9.3), management reports (14.3)

Adjustments. Manual "top-side" journals entered directly in the GL break the link to trades and are the most common cause of subledger-to-GL differences. Prefer adjustments in the subledger with a reason code. Where top-side entries are unavoidable, tag them, approve them and set them to reverse automatically in the next period.

Engineering view.

• Derive GL segments during enrichment. A missing mapping must send the event to suspense with an alert, never default silently to a catch-all account.
• Make the transfer idempotent: the GL should reject a batch ID it has already posted.
• Lock the period after subledger close. Late events post into the next open period and keep their original effective date.
• Reconcile subledger balances to the GL by GL key every day, not only at month end (Chapter 13.4).
• Provide a drill-back API from any GL line to the underlying postings and events. Auditors will ask for it.
• Convert to functional currency consistently (per posting, then aggregate, or the reverse) so that converted totals still balance; inconsistent conversion is a classic source of one-cent breaks.

Analogy. The subledger is the shoebox of receipts and the GL is the household budget sheet with one total per category. The budget sheet is what you read, but you must always be able to go back to the shoebox to explain a total.

Key takeaways

• The subledger holds trade-level postings; the GL holds balances by GL key.
• Postings are aggregated by entity, account, cost centre, product, currency and date, and transferred in batches with control totals.
• The month-end close runs on a fixed working-day calendar; the subledger close and reconciliation are on the critical path.
• Drill-back, idempotent transfers, period locks and daily reconciliation keep the GL explainable.

🧪 Practice

1. Beginner: Give three reasons why a bank does not send every subledger posting to the GL.
2. Intermediate: A journal batch of 18,412 lines is rejected because total debits and total credits in the functional currency differ by EUR 0.01. List likely causes and describe how to prevent this in the aggregation and conversion steps.
3. Interview: "Design the month-end interface between a treasury subledger and the GL for five legal entities reporting under both IFRS and local GAAP." Hint: event-driven daily postings, aggregation keys per framework, batch control totals and idempotent batch IDs, period locks and late-event handling, reconciliation reports per entity and framework, drill-back, and monitoring against the close calendar.

<a id="142-hedge-accounting-implementation"></a>

14.2 Hedge Accounting Implementation

Chapter 6.3 explained why hedge accounting exists and how effectiveness is tested. This subchapter covers how a bank implements hedge accounting in its systems: designation records, the mechanics of each hedge type, the calculation pipeline and the journal entries it produces.

Hedge Designation and Documentation

Why it matters. Hedge accounting is optional, and it is an exception to the normal measurement rules. Standards therefore require the bank to prove that a relationship was set up as a hedge from the start, not chosen after the results were known. Without contemporaneous documentation, auditors will disallow hedge accounting, and the full fair value change of the hedging derivative will hit P&L. A hedge accounting system is first of all a record-keeping system.

Qualifying criteria (IFRS 9).

1. The relationship consists only of eligible hedging instruments and eligible hedged items.
2. At inception, there is formal designation and documentation of the relationship and of the risk management objective and strategy.
3. The effectiveness requirements are met: an economic relationship exists, credit risk does not dominate the value changes, and the hedge ratio is consistent with how risk is actually managed.

What the documentation contains.

Element	Description	Example
Risk management objective and strategy	Why the hedge exists	Convert a fixed-rate bond to floating to stay within IRRBB limits
Hedging instrument	Specific trade(s) and proportion	Swap SWP-55871, 100% of notional
Hedged item	Asset, liability, firm commitment, highly probable forecast transaction or portfolio	Bond XS0000000000, EUR 100m nominal
Hedged risk	The risk component being hedged	Fair value changes due to the EUR benchmark rate
Hedge type	Fair value, cash flow or net investment	Fair value hedge
Hedge ratio	Quantity of instrument per quantity of item	1:1
Effectiveness assessment	Prospective method and ineffectiveness measurement	Critical terms and regression; cumulative dollar offset for measurement
Sources of ineffectiveness	Expected differences	Discount curve basis, floating leg, counterparty credit
Designation date	On or before the start of hedge accounting	2026-10-19

Example: a designation record.

hedge_relationship:
  id: HR-2026-0412
  version: 1
  type: FAIR_VALUE
  framework: IFRS9
  designation_timestamp: 2026-10-19T14:32:05Z
  designated_by: alm_desk_user_014
  approved_by: accounting_policy_user_007
  objective: >
    Convert the fixed-rate exposure on a EUR 100m 5-year bond held at amortized cost
    to floating, in line with the IRRBB hedging strategy approved by ALCO.
  hedged_risk: EUR benchmark interest rate (ESTR-discounted swap curve)
  hedged_items:
    - instrument: XS0000000000
      nominal: 100000000
      proportion: 1.0
  hedging_instruments:
    - trade_id: SWP-55871
      proportion: 1.0
  hedge_ratio: 1.0
  effectiveness:
    prospective: critical_terms_and_regression
    measurement: dollar_offset_cumulative
  expected_ineffectiveness: [discount_curve_basis, floating_leg, counterparty_credit]
  status: DESIGNATED

Lifecycle.

  DRAFT --approve--> DESIGNATED --period-end assessment--> passes --> stays DESIGNATED
                        |    ^                                |
                        |    |                                | criteria no longer met
                        |    +-- REBALANCED (new hedge ratio, |
                        |        same objective; IFRS 9)      v
                        |                                DISCONTINUED
                        +--- instrument expires, is sold or terminated;
                             hedged item derecognized; forecast no longer expected ---^

• Rebalancing (IFRS 9): if the hedge ratio drifts but the risk management objective is unchanged, the ratio is adjusted and the relationship continues.
• Discontinuation: under IFRS 9, a bank cannot voluntarily de-designate a relationship that still meets the risk management objective. Under IAS 39 it can.
• Macro hedges: banks using the IAS 39 portfolio fair value hedge (Chapter 6.3) designate an amount of a portfolio per repricing time bucket rather than specific items. Because the portfolio changes all the time, they de-designate and re-designate it every period, often monthly, and track the basis adjustments of each closed designation separately.

Engineering view.

• The designation timestamp must be on or before the start of hedge accounting. The system should block backdated designations or route them to accounting policy as exceptions.
• Hedge relationships are immutable once approved. Rebalancing creates a new version that links to the previous one, with maker-checker approval.
• Subscribe to trade lifecycle events from the treasury system for every trade used in a hedge. An amendment or termination of a hedging swap that the hedge accounting team does not see is one of the most common hedge accounting failures.
• A trade can be partly designated in several relationships. Check that the designated proportions of each trade never sum to more than 100%.
• Generate the documentation from the record itself, so the document and the data cannot diverge.

Analogy. Designating a hedge is like filing a patent: you must file before showing the invention to the world, and the filing defines exactly what is protected. A filing made after the fact protects nothing.

Key takeaways

• Hedge accounting requires formal designation and documentation at inception, before the results are known.
• Documentation covers objective, instrument, item, hedged risk, type, ratio, effectiveness method and expected ineffectiveness.
• Relationships go through designation, assessment, rebalancing and discontinuation; macro hedges are re-designated periodically.
• Systems must prevent backdating, version relationships, watch hedging trades for lifecycle events and control designated proportions.

🧪 Practice

1. Beginner: List five items that hedge documentation must contain under IFRS 9.
2. Intermediate: A trader asks to designate a swap traded three weeks ago as a hedge of a bond, effective from the swap's trade date, because the swap has since lost value. Explain the problem and what the system should allow instead.
3. Interview: "How would you detect that a hedging swap has been amended or terminated in the treasury system without the hedge accounting team being informed?" Hint: subscribe to lifecycle events filtered on designated trade IDs, block or alert on changes, run a daily reconciliation between the hedge relationship inventory and live trades, and check designated proportions against current notionals.

Fair Value vs Cash Flow Hedges

Why it matters. Chapter 6.3 introduced the hedge types. Choosing between them is a design decision with long-lasting consequences. The same pay-fixed swap could be documented as a fair value hedge of a fixed-rate asset or as a cash flow hedge of floating-rate funding, depending on how risk management describes the exposure. The choice decides where volatility appears (P&L, OCI or capital), which balances the system must track, and what happens when the hedge ends.

Mechanics compared.

Aspect	Fair value hedge (FVH)	Cash flow hedge (CFH)
Hedged exposure	Fair value changes of a recognized fixed-rate item or firm commitment	Variability of cash flows of floating-rate items or highly probable forecast transactions
Hedging instrument	Fair value change to P&L	Effective part to OCI (cash flow hedge reserve), ineffective part to P&L
Hedged item	Carrying amount adjusted for the hedged risk (basis adjustment), through P&L	Measurement unchanged
Ineffectiveness	Net of the derivative and the basis adjustment, both in P&L	Over-hedge to P&L; under-hedge not recognized (lower-of test)
Effect on equity	Through P&L only	Through the OCI reserve; the reserve is filtered out of CET1 (CRR Article 33)
Balances the system tracks	Cumulative basis adjustment per hedged item or portfolio	Cumulative changes of the hedge and the hypothetical derivative; reserve balance
On discontinuation	Basis adjustment amortized to P&L over the item's remaining life	Reserve kept and recycled when the hedged cash flows affect P&L; to P&L at once if the forecast transaction is no longer expected

Where a value change ends up.

  A pay-fixed swap gains 10.0 when rates rise.

  FAIR VALUE HEDGE                                  CASH FLOW HEDGE
  (hedged item: fixed-rate bond)                    (hedged item: floating-rate funding)

  Swap                  +10.0  --> P&L              Swap                  +10.0 --> OCI reserve
  Bond basis adjustment  -9.6  --> P&L                                            (effective part)
  ----------------------------------------          Funding               no entry today
  Net P&L                +0.4  (ineffectiveness)    Later: the reserve is recycled to P&L as the
                                                    higher floating interest is paid, offsetting it

Choosing the hedge type.

Balance sheet exposure	Usual hedge	Swap
Fixed-rate assets (bonds, mortgages)	Fair value hedge: convert fixed to floating	Pay fixed, receive floating
Issued fixed-rate debt	Fair value hedge: convert fixed to floating	Receive fixed, pay floating
Floating-rate loans	Cash flow hedge: lock in income	Receive fixed, pay floating
Floating-rate funding, rolled over	Cash flow hedge: lock in cost	Pay fixed, receive floating
Forecast bond issuance	Cash flow hedge: lock in the future coupon	Forward-starting pay-fixed swap

Example: forecast issuance (cash flow hedge). In January, treasury plans to issue a EUR 500m 5-year fixed-rate bond in June. To lock in today's rates, it enters a forward-starting pay-fixed swap and designates it as a CFH of the forecast coupons.

1. January to June: rates rise. The swap gains EUR 12.0m and the hypothetical derivative EUR 12.3m. Under the lower-of test, EUR 12.0m goes to the CFH reserve and nothing to P&L.
2. June: the bond is issued at a higher coupon, and the swap is closed for EUR 12.0m in cash. The reserve stays in equity.
3. The next 5 years: the reserve is recycled to interest expense as the coupons are paid, about EUR 2.4m per year (following the EIR pattern in practice). This offsets the higher coupon, so the effective cost of the bond is close to the January rate.
4. If the issuance is cancelled and no longer expected to happen, the whole EUR 12.0m is reclassified to P&L immediately. If it is only delayed but still expected, the reserve remains.

Example: discontinuing a fair value hedge. A fixed-rate bond has a cumulative basis adjustment of EUR -1.35m when its hedging swap is terminated, three years before the bond matures. The bond's carrying amount stays 1.35m lower, and the adjustment is amortized into interest income over the remaining three years: about EUR 0.45m per year, following the EIR method.

Engineering view.

• A fair value hedge needs a valuation of the hedged item for the hedged risk only, which the bank does not otherwise compute for an amortized-cost item. The hedge system needs its own pricing of hedged items on the benchmark curve.
• A cash flow hedge needs hypothetical derivatives: synthetic swaps with terms frozen at designation, valued by the pricing library but never booked as real trades.
• Basis adjustments and reserves outlive the hedges that created them, sometimes by decades for long mortgages. Design for years of tracking after the derivative has gone.
• Capital reporting needs the CFH reserve broken down, because it is removed from CET1.

Analogy. A fair value hedge is like putting a protective case on a phone you own: the case and the phone are valued together. A cash flow hedge is like fixing the price of next year's holiday today: nothing changes in your possessions now, but the saving is kept aside and released when you actually pay for the trip.

Key takeaways

• Fair value hedges adjust the hedged item's carrying amount through P&L; cash flow hedges store the effective part of the derivative's change in OCI.
• Under the CFH lower-of test, over-hedging goes to P&L but under-hedging does not.
• On discontinuation, FVH basis adjustments are amortized; CFH reserves are recycled when the hedged cash flows affect P&L, or immediately if the forecast transaction is no longer expected.
• The hedge type determines which long-lived balances and synthetic valuations the system must maintain.

🧪 Practice

1. Beginner: A bank issues a 7-year fixed-rate bond and swaps it to floating. Which hedge type fits, and where do the swap's fair value changes go?
2. Intermediate: In the forecast issuance example, the issuance is first delayed by three months but remains highly probable, and is later cancelled. Describe the accounting at each step.
3. Interview: "Why might a bank prefer cash flow hedge accounting for some exposures even when a fair value hedge could also work?" Hint: consider which balance sheet items are available as hedged items, P&L volatility versus OCI volatility, the CET1 filter, the operational effort of basis adjustments on large portfolios compared with hypothetical derivatives, and what happens on discontinuation.

Effectiveness Calculations

Why it matters. Chapter 6.3 described the effectiveness tests and their thresholds. In a system, effectiveness is a calculation pipeline that runs every reporting period for every relationship: obtain the right valuations, compute cumulative changes, assess, measure ineffectiveness and produce postings. Most failures come from the inputs, not from the test formula: the wrong curve, dirty instead of clean values, or a hedged item valued with risks that were never hedged.

The pipeline.

  Hedge relationship inventory (designation records)
         |
         v
  [1] Period-end valuations per relationship
        - hedging instrument: clean fair value, full market curves
        - FVH: hedged item valued for the hedged risk only
        - CFH: hypothetical derivative, clean value
         |
         v
  [2] Changes since designation (cumulative) and since the last period
         |
         v
  [3] Assessment: prospective (regression on rate scenarios); retrospective where required
         |                                          fail --> discontinuation workflow
         v
  [4] Ineffectiveness measurement and split: P&L / OCI / basis adjustment
         |
         v
  [5] Accounting events --> rules engine (14.1) --> postings
      All inputs and results stored per run for audit

Getting the inputs right.

• Clean values: use values excluding accrued interest. Otherwise accruals and coupon payments appear as value changes and create artificial ineffectiveness. Accruals are booked separately in net interest income.
• Isolating the hedged risk (FVH): value the hedged item's cash flows on the benchmark curve with the credit spread frozen at its designation level. Many banks hedge only a benchmark component: the cash flows at the swap rate prevailing at designation. Changes in the issuer's credit spread are not part of the hedged risk and must not enter the calculation.
• Hypothetical derivative (CFH): a swap whose terms match the hedged cash flows, priced at market at designation (zero value at inception) and without counterparty credit adjustments. It models a perfect hedge, so its value moves in the same direction as the actual hedge.
• Cumulative measurement: measure ineffectiveness on a cumulative basis since designation. The period amount is the cumulative amount now minus the cumulative amount already booked.

Worked example: fair value hedge (EUR m).

	Q1	Q2	Cumulative to Q2
Swap clean fair value change	+2.40	-0.90	+1.50
Hedged item change for hedged risk	-2.30	+0.95	-1.35
Dollar offset (cumulative)	104.3%		111.1%
Ineffectiveness to P&L	+0.10	+0.05	+0.15

Both cumulative ratios are inside 80-125%, and the cumulative ineffectiveness of +0.15 equals the sum of the two quarters.

Worked example: cash flow hedge lower-of test (EUR m). A pay-fixed swap hedges floating-rate funding. The reserve holds the smaller of the cumulative change in the swap and the cumulative change in the hypothetical derivative.

Period	Cumulative swap	Cumulative hypothetical	Reserve balance	OCI movement	Swap change in period	P&L ineffectiveness
Q1	+3.0	+3.2	+3.0	+3.0	+3.0	0.0 (under-hedge)
Q2	+4.5	+4.2	+4.2	+1.2	+1.5	+0.3 (over-hedge)
Q3	-1.0	-0.8	-0.8	-5.0	-5.5	-0.5 (over-hedge)

After Q3, cumulative P&L ineffectiveness is -0.2, which equals the cumulative swap change minus the reserve: -1.0 - (-0.8).

from dataclasses import dataclass
from decimal import Decimal

@dataclass(frozen=True)
class CfhResult:
    reserve: Decimal          # cumulative balance of the cash flow hedge reserve
    oci_movement: Decimal     # posting to OCI this period
    ineffectiveness: Decimal  # posting to P&L this period

def cfh_lower_of(cum_hedge: Decimal, cum_hypothetical: Decimal,
                 prev_reserve: Decimal, period_hedge_change: Decimal) -> CfhResult:
    if cum_hedge * cum_hypothetical <= 0:   # no offset: nothing qualifies for the reserve
        reserve = Decimal(0)
    else:
        sign = 1 if cum_hedge > 0 else -1
        reserve = sign * min(abs(cum_hedge), abs(cum_hypothetical))
    oci = reserve - prev_reserve
    return CfhResult(reserve, oci, period_hedge_change - oci)

cfh_lower_of(Decimal("4.5"), Decimal("4.2"), Decimal("3.0"), Decimal("1.5"))
# CfhResult(reserve=Decimal('4.2'), oci_movement=Decimal('1.2'), ineffectiveness=Decimal('0.3'))

Prospective assessment. A regression test revalues the hedging instrument and the hedged item under a set of rate scenarios (parallel shifts, twists, or historical changes) and regresses one series of value changes on the other. Store the scenario set and results with the relationship, because auditors will ask how the test was performed at each date. Under IFRS 9 there is no 80-125% bright line, so a weak result leads to a review of the hedge ratio (rebalancing) rather than automatic discontinuation.

Engineering view.

• Use the same curves, cut-off times and model versions as the official valuations, and store the inputs (curve snapshots, model versions, trade versions) so any period can be reproduced exactly.
• Keep hypothetical derivatives and hedged-risk valuations inside the hedge accounting module as synthetic instruments. Booking them as trades in the treasury system lets them leak into risk and P&L.
• Run a dry run a few days before period end so failing relationships are found while there is still time to act.
• Macro hedges can involve thousands of relationships and time buckets. Batch the valuations and parallelize per relationship.
• Guard ratio tests against very small denominators, and report ratios together with absolute amounts (the "small numbers problem" of Chapter 6.3).

Analogy. An effectiveness run is like a monthly bank reconciliation. The arithmetic is trivial; the work lies in making sure both sides are measured on the same basis and at the same moment.

Key takeaways

• Effectiveness is a periodic pipeline: valuations, cumulative changes, assessment, ineffectiveness split and postings.
• Use clean values, isolate the hedged risk with frozen credit spreads, and model CFH hedged items with hypothetical derivatives.
• Fair value hedge ineffectiveness is the net of the derivative and the basis adjustment; CFH ineffectiveness follows the lower-of test.
• Reproducibility requires storing all valuation inputs per run.

🧪 Practice

1. Beginner: Explain why effectiveness calculations use clean rather than dirty values.
2. Intermediate: Extend the cash flow hedge table with Q4: the cumulative swap change is +0.5 and the cumulative hypothetical change is +0.6. Compute the reserve balance, the OCI movement and the P&L ineffectiveness, then check your answer with cfh_lower_of.
3. Interview: "After a change in curve methodology, 40 hedge relationships fail the effectiveness test in March. How do you investigate?" Hint: reproduce the results with stored inputs before and after the change; check whether the hedged item and the hedging instrument were both moved to the new curve set; compare values for each relationship one at a time; and consider whether the change went through model governance.

Accounting Entries for Hedges

Why it matters. The final output of hedge accounting is journal entries. They use accounts that do not exist elsewhere (basis adjustments, the CFH reserve, hedge results), they must be traceable to a relationship and a period, and some of them continue long after the hedge has ended. Errors here appear directly in reported profit and equity.

Accounts used.

Account	Purpose
Derivative assets and liabilities: hedging	Fair value of hedging derivatives, shown separately from trading derivatives
Hedge basis adjustment (per hedged item)	FVH adjustment to the carrying amount of the hedged item
Fair value changes of hedged items in portfolio hedges	Separate balance sheet line for IAS 39 macro fair value hedges
Cash flow hedge reserve (OCI)	Effective portion of CFH derivatives
Hedge result (P&L)	Fair value changes of hedging instruments and hedged items, and ineffectiveness
Interest income and expense: hedging derivatives	Net swap accruals, presented within net interest income

Fair value hedge entries (EUR m). Using the effectiveness example: a EUR 100m fixed-rate bond at amortized cost hedged with a pay-fixed swap.

  Q1: rates rise
  Dr  Derivative assets: hedging                         2.40
      Cr  Hedge result: hedging instrument (P&L)                    2.40
  Dr  Hedge result: hedged item (P&L)                    2.30
      Cr  Debt securities at AC: hedge basis adjustment             2.30

  Q2: rates fall partially
  Dr  Hedge result: hedging instrument (P&L)             0.90
      Cr  Derivative assets: hedging                                0.90
  Dr  Debt securities at AC: hedge basis adjustment      0.95
      Cr  Hedge result: hedged item (P&L)                           0.95

  Q3: swap terminated for its fair value of 1.50; hedge discontinued
  Dr  Nostro                                             1.50
      Cr  Derivative assets: hedging                                1.50
  Each quarter for the remaining 3 years (straight-line shown; EIR in practice):
  Dr  Debt securities at AC: hedge basis adjustment      0.1125
      Cr  Interest income: securities                               0.1125

Interest on the hedging swap. The swap's accruals are booked in net interest income next to the interest on the hedged item, so NII shows the hedged result. Suppose the bond pays 3.00%, the swap pays 2.60% fixed and receives 3-month Euribor fixed at 2.10%:

  Quarterly accruals on EUR 100m
  Dr  Accrued interest receivable: bond          0.750
      Cr  Interest income: securities                       0.750
  Dr  Interest expense: hedging derivatives      0.650
      Cr  Accrued interest payable: swap                    0.650
  Dr  Accrued interest receivable: swap          0.525
      Cr  Interest income: hedging derivatives              0.525

  Net interest income = 0.750 - 0.650 + 0.525 = 0.625
                      = (Euribor 2.10% + 0.40%) x 100m / 4  -> the bond now earns floating

Cash flow hedge entries (EUR m). Using the lower-of example:

  Q1: Dr  Derivative assets: hedging       3.00
          Cr  Cash flow hedge reserve (OCI)               3.00
  Q2: Dr  Derivative assets: hedging       1.50
          Cr  Cash flow hedge reserve (OCI)               1.20
          Cr  Hedge result: ineffectiveness (P&L)         0.30
  Q3: Dr  Cash flow hedge reserve (OCI)    5.00
      Dr  Hedge result: ineffectiveness    0.50
          Cr  Derivative assets / liabilities: hedging    5.50

Recycling. For a forecast transaction such as the bond issuance in the previous topic, recycling is an explicit entry each period:

  Swap closed at issuance:   Dr  Nostro                        12.0
                                 Cr  Derivative assets: hedging          12.0
  Each year, for 5 years:    Dr  Cash flow hedge reserve (OCI)  2.4
                                 Cr  Interest expense: issued debt        2.4
  If the issuance is cancelled and no longer expected:
                             Dr  Cash flow hedge reserve (OCI) 12.0
                                 Cr  Hedge result (P&L)                  12.0

For a cash flow hedge of an ongoing floating-rate exposure, the swap's periodic net settlements are booked in net interest income in the same period as the hedged interest. The reserve, which holds the effective part of the swap's clean value, unwinds as the swap approaches maturity. Banks implement this either with explicit reclassification entries each period or through their accrual booking design. The method must be documented in the accounting policy and applied consistently.

Engineering view.

• Post hedge entries through the same rules engine as other events, with the relationship ID as an attribute. This allows reconciliation and audit per relationship.
• By default, a treasury system posts every swap revaluation to trading P&L. For hedging swaps, the hedge accounting module must redirect these postings, either through a hedge clearing account that it allocates or through reclassification entries. If the treasury system and the hedge module both post the same fair value change, P&L is double counted: a classic defect.
• Reconcile basis adjustment and reserve balances in the GL to the hedge module every period, per relationship.
• Generate amortization schedules for basis adjustments and reserves at discontinuation, and keep running them after the derivative has gone.

Analogy. Hedge entries are like keeping two halves of a matched receipt in the same folder. Each half alone looks like a gain or a loss; filed together, they show the protection that was bought.

Key takeaways

• FVH entries post the derivative's change and the basis adjustment to P&L; the net is ineffectiveness.
• CFH entries split the derivative's change between the OCI reserve and P&L under the lower-of test.
• Swap accruals are booked in net interest income so that NII reflects the hedged position.
• Reserves and basis adjustments must be recycled or amortized after discontinuation, and hedge postings must not duplicate the treasury system's default derivative postings.

🧪 Practice

1. Beginner: Write the fair value hedge entries for a period in which the swap loses 0.80 and the hedged item gains 0.75 for the hedged risk. What is the ineffectiveness?
2. Intermediate: The treasury system automatically posts every swap revaluation to "Trading P&L". Explain what happens to reported results if the hedge accounting module also posts the change of the same swap to "Hedge result", and design a posting flow that prevents it.
3. Interview: "Walk me through the journal entries over the life of a forward-starting swap used to hedge a forecast bond issuance." Hint: cover the periods before issuance (OCI versus ineffectiveness), the settlement of the swap at issuance, recycling over the life of the bond, and what happens if the issuance is delayed or cancelled.

<a id="143-management-reporting"></a>

14.3 Management Reporting

Financial statements serve outside readers and arrive monthly or quarterly; management needs faster, more granular and forward-looking views. This subchapter covers the main treasury management reports and the data quality controls that make them trustworthy.

Liquidity and ALM Dashboards

Why it matters. At 08:30 the treasurer needs to know two things: whether the bank is safe today, and whether it is moving in the right direction. Liquidity can change within hours (Chapter 5.3; Silicon Valley Bank faced USD 42bn of withdrawal requests in a single day), while interest rate risk changes more slowly. A dashboard brings the key metrics, limits and trends together, and a good one makes a breach or an adverse trend impossible to miss.

Audiences.

Audience	Question	Frequency	Typical content
Treasury desk	Can we settle everything today? What must we fund?	Intraday or real time	Central bank balance, intraday usage, payment queue, today's maturities
Treasurer, head of ALM	Are we within limits? Where are we heading?	Daily	LCR, survival horizon, buffer, funding concentration, deposit flows
ALCO, CFO	Is the strategy working?	Monthly	Trends, forecasts, plan versus actual (ALCO packs, below)
Board risk committee	Are we within risk appetite?	Quarterly	Risk appetite metrics with status

Metric catalogue.

Metric	Defined in	Typical frequency
LCR	Chapter 9.2	Daily (internal), monthly (regulatory)
NSFR	Chapter 9.2	Monthly or quarterly
Survival horizon under stress	Chapter 5.1	Daily
Liquidity buffer size and composition	Chapter 5.4	Daily
Intraday liquidity usage	Chapter 5.3	Intraday
Maturity ladder and cash flow gaps	Chapter 5.2	Daily
NII and EVE sensitivity	Chapter 6.2	Monthly, some banks daily
Funding concentration and deposit flows	Chapter 7.1	Daily or weekly
Asset encumbrance	Chapter 5.4	Monthly

Example: a daily dashboard.

  LIQUIDITY AND ALM DASHBOARD - GROUP            As of 2026-10-16 EOD | Run 20261016-03 | Data completeness 99.8%
  ------------------------------------------------------------------------------------------------------
  Metric                       Value      Appetite / early warning   Status   1-day     20-day trend
  LCR (internal)               148%       >= 120% / 130%             GREEN    -3 pp     falling
  Survival horizon (stress)    67 days    >= 60 / 70 days            AMBER    -5 days   falling
  NSFR (last month end)        126%       >= 110% / 115%             GREEN    n/a       stable
  HQLA buffer (EUR bn)         21.4       >= 18.0 / 19.5             GREEN    -0.6      falling
  Peak intraday usage          38%        <= 60% / 50%               GREEN    +4 pp     stable
  Top-10 depositor share       14.2%      <= 20% / 17%               GREEN    +0.1 pp   rising
  NII sensitivity, -200bp      -6.1%      >= -10% / -8%              GREEN    n/a       stable
  EVE sensitivity (SOT)        -11.3%     >= -15% / -13%             GREEN    n/a       stable
  ------------------------------------------------------------------------------------------------------
  LCR drivers (1-day): retail deposit outflows -1.2 pp, settled bond purchases -1.0 pp,
                       wholesale maturities -0.8 pp
  Alert: survival horizon below early warning level (70 days) since 2026-10-15

Design principles.

• One metric, one definition. Metrics are calculated in a governed calculation layer, never in the dashboard tool.
• Context for every number. Show the as-of date, run ID and data completeness, and place the limit and early warning level next to the value. Status is derived by rule, not set by hand.
• Trends and drivers, not only levels. A falling LCR at 148% may matter more than a stable one at 135%.
• Drill-down by entity, currency, product and counterparty.
• Consistency with regulatory figures. On regulatory reporting dates, the internal LCR must reconcile to the reported LCR, with documented differences.

Architecture.

  Liquidity engine ---+
  ALM engine ---------+--> results store -----------> metric layer -------------> dashboard
  Payments / RTGS ----+    (versioned snapshot        (definitions, limits,        (presentation,
  Limits system ------+     per run, run ID)           status rules, drivers)       drill-down)

Engineering view.

• Tie every published number to a run ID. A restatement creates a new version; it never overwrites the original.
• Monitor delivery against the agreed deadline (for example, the daily dashboard available by 07:30) and show data freshness on the screen.
• Support a stress mode: during a crisis, management will ask for refreshes every hour or more often. BCBS 239 expects banks to produce risk data quickly in stress (see the last topic of this subchapter).
• Restrict access: liquidity data can be highly sensitive, and for listed banks it may be inside information.

Analogy. A good dashboard is like a car's instrument panel: a few gauges that matter, each with a red zone, and warning lights that come on before the engine fails, not after.

Key takeaways

• Different audiences need different metrics at different frequencies, from intraday to quarterly.
• Core metrics include LCR, NSFR, survival horizon, buffer, intraday usage, gaps, NII and EVE sensitivity, and concentration.
• Show limits, early warning levels, trends, drivers, as-of dates and data completeness next to each metric.
• Calculate metrics in a governed layer, version every run, and be ready for frequent refreshes in stress.

🧪 Practice

1. Beginner: Choose five metrics for a treasurer's daily liquidity dashboard and justify each choice in one sentence.
2. Intermediate: The dashboard shows an LCR of 148%, while the regulatory report for the same date shows 141%. List possible reasons and describe how you would design the system so that the difference can be explained.
3. Interview: "Design the data architecture for a liquidity dashboard that must refresh every 15 minutes during a stress event." Hint: incremental position updates from payment and trading events (Chapter 13.3), cached HQLA valuations, a stress mode with reduced scope, completeness indicators per source, versioned snapshots, and graceful degradation when a source is late.

P&L Explain Reports

Why it matters. Chapter 8.1 described P&L attribution for trading desks. Treasury also owns the banking book, whose income is mostly accrual-based net interest income, and management asks a different question there: why is NII different from last month, from budget, or from forecast? P&L explain reports answer both questions. Without a credible explanation, reported numbers are not trusted and decisions are delayed.

Two kinds of explain.

Report	Scope	Measure	Frequency	Main components
Daily P&L explain	Fair-valued treasury positions, derivatives, hedge results	Fair value P&L and OCI	Daily	Carry, rates, spreads, FX, new trades, amendments, unexplained (Chapter 8.1)
NII variance analysis	Banking book: loans, deposits, funding, hedges	Net interest income	Monthly, versus prior period, budget or forecast	Volume, rate, mix, day count, hedges, one-off items

Volume and rate effects.

  Income = average balance (V) x rate (r)

  Change = V1 x r1 - V0 x r0
         = (V1 - V0) x r0      volume effect: change in balance at the old rate
         + (r1 - r0) x V1      rate effect:   change in rate on the new balance

The order matters. Some banks use a three-way split with a separate joint term (V1 - V0) x (r1 - r0). Choose one convention, document it and apply it everywhere.

Example (annualized, EUR m).

Line	V0 (bn)	r0	V1 (bn)	r1	NII before	NII after	Volume effect	Rate effect	Total
Mortgages (income)	10.0	3.00%	10.5	3.20%	300.0	336.0	+15.0	+21.0	+36.0
Retail deposits (expense)	8.0	1.00%	8.2	1.25%	-80.0	-102.5	-2.0	-20.5	-22.5
Net					220.0	233.5	+13.0	+0.5	+13.5

NII grew by EUR 13.5m, almost entirely because of mortgage volume. The higher mortgage rates (+21.0m) were almost fully absorbed by higher deposit rates (-20.5m), so the margin gain from higher rates has nearly disappeared. This is a typical late-cycle pattern, in which deposit rates catch up with market rates (deposit beta, Chapter 6.1).

Other components of an NII explain.

• Day count: February has 28 days and March 31, so monthly NII changes even when balances and rates do not.
• Hedges: net swap accruals from Section 14.2, shown separately so that management sees what hedging contributed.
• One-off items: EIR catch-up adjustments (Section 14.1), prepayment fees, corrections of prior periods.
• Business split: through FTP (Chapter 6.4), the NII change is allocated to lending, deposits and treasury.

Workflow.

  Flash P&L        Official P&L            Explain                Commentary            Sign-off
  (T+0 evening, -> (T+1, product control -> (components +      -> (items above        -> (product control,
   front office)    after IPV and           unexplained)           threshold,             desk head)
                    adjustments)                                   by desk)

Thresholds decide what needs commentary, for example unexplained amounts above 5% of the day's P&L or above EUR 100k.

Engineering view.

• The explain must add up: explained plus unexplained must equal the official P&L or NII exactly, using the same positions and cut-off.
• Store the inputs for both dates (positions, market data, rates, balances) so any explain can be re-run.
• Make the separation of effects deterministic, with a documented order of steps.
• Capture commentary as structured data attached to the report version, so it becomes part of the audit trail rather than an email.
• Keep the explain additive across levels: the sum of desk explains must equal the total explain.

Analogy. A P&L explain is like a household comparing this month's spending with last month's: more people at dinner (volume), higher grocery prices (rate), and one extra day in the month (day count). If a large difference remains unexplained, something was missed or wrongly recorded.

Key takeaways

• Daily explains cover fair-valued positions; NII variance analysis covers the accrual banking book.
• Volume and rate effects split NII changes; the decomposition convention must be documented and consistent.
• Day count, hedges, one-off items and FTP allocation complete the NII explain.
• Explains must reconcile exactly to official figures and be reproducible from stored inputs.

🧪 Practice

1. Beginner: A deposit balance falls from EUR 5.0bn at 0.50% to EUR 4.6bn at 0.90%. Compute the volume and rate effects on annual interest expense.
2. Intermediate: NII in February is 4% lower than in January, although balances and rates did not change. What explains this, and how should the report present it?
3. Interview: "Product control complains that the daily explain for treasury's FVOCI portfolio shows 15% unexplained P&L every day. How would you debug it?" Hint: compare position sets and cut-off times between the explain and the official figures; look for missing risk factors (basis spreads, issuer-specific credit curves); check how new trades and amendments are classified, clean versus dirty value treatment, FX translation and sign conventions.

ALCO Reporting Packs

Why it matters. Chapter 4.1 described ALCO's mandate and the metrics it reviews. The ALCO pack is the evidence of how the bank governs its balance sheet, and supervisors read packs and minutes during inspections. Producing it is a monthly process with hard deadlines, many contributors and high stakes. Traditionally it was a large spreadsheet assembled by hand, which is an operational risk in itself.

Typical structure.

Section	Content	Main source
Executive summary	Key messages, breaches, decisions requested	Treasury and ALM commentary
Balance sheet and plan	Balance sheet versus plan, loan and deposit growth	GL (finance), planning
Liquidity	LCR, NSFR, survival horizon, buffer, stress results, funding plan	Liquidity engine
Interest rate risk	NII and EVE sensitivities, repricing gaps, hedging activity, outlier test	ALM engine
NII and margins	NII versus plan, volume and rate explain, FTP results	Finance, FTP engine
Capital (in some banks)	CET1, leverage ratio, RWA	Regulatory reporting
Market outlook	Rate scenarios, competitor pricing	Economics, business lines
Limits and breaches	Utilization, breaches, remediation	Limits system (Chapter 8.4)
Decisions and actions	Proposals, open actions from previous meetings	ALCO secretary

Production timeline (working days after month end).

  M+1   Month-end data final; ALM and liquidity engines run on month-end positions
  M+3   GL close complete; balance sheet reconciled to finance figures
  M+4   Metrics frozen (data lock); tables and charts generated
  M+5   Commentary by treasury, ALM and risk; review by heads of function
  M+6   Pack signed off by the treasurer and a risk delegate; distributed to members
  M+8   ALCO meeting; decisions and actions minuted
  M+9   Pack, data snapshot and minutes archived (immutable, retained per policy)

Production pipeline.

  GL -----------+
  ALM engine ---+--> reporting data mart ---> metric layer ---------> pack generator ---> draft pack
  Liquidity ----+    (frozen snapshot per     (definitions, limits,   (templates,            |
  FTP engine ---+     month end)               status rules)           tables, charts)       v
  Limits -------+                                                                commentary and review
                                                                                          |
                                                                                sign-off --> archive

Quality principles.

• Consistency across sections. Total deposits in the balance sheet section and in the liquidity section should match, or the difference (for example, accounting versus behavioural views) should be explained in the pack.
• Reconciliation to finance. Balance sheet figures reconcile to the GL (Section 14.1), and remaining differences are listed.
• Versioning. The pack shown at the meeting is archived together with its data snapshot. Later corrections are documented as restatements, never applied silently.
• Less end-user computing. Replace manual spreadsheet steps with generated tables. Inventory and control any spreadsheets that remain.
• Action tracking. Decisions are recorded as structured items with an owner and a due date, and reported on at the next meeting.

Engineering view.

• Generate the pack from templates that read the frozen snapshot, and print the snapshot or run ID on every page.
• After the data lock, route late corrections through a controlled restatement process instead of regenerating the pack quietly.
• Keep chart scales and definitions stable from month to month so that members can compare packs.
• Use reconciliation checks as gates in the pipeline: if the balance sheet does not reconcile to the GL within tolerance, the pack is not generated without an explicit override.
• Restrict distribution. For listed banks, ALCO packs can contain inside information under market abuse rules.

Analogy. An ALCO pack is like a flight's pre-departure briefing pack: compiled to a fixed deadline from many sources, read by the people who decide, and kept afterwards as evidence of what they knew when they decided.

Key takeaways

• The ALCO pack documents balance sheet governance and is read by supervisors.
• It covers the balance sheet, liquidity, interest rate risk, NII, limits and decisions, and draws on several engines and the GL.
• Production follows a fixed timeline with a data lock, commentary, sign-off and archiving.
• Automation should generate figures from a frozen snapshot with reconciliation gates; commentary and judgment remain human.

🧪 Practice

1. Beginner: Name four sections of an ALCO pack and the source system for each.
2. Intermediate: Deposits are EUR 52.3bn in the balance sheet section and EUR 51.1bn in the liquidity section of the same pack. Give possible reasons and explain how the pack should present the difference.
3. Interview: "Today the ALCO pack is 60 linked spreadsheets that two analysts assemble over five days. How would you automate it, and what would you not automate?" Hint: a data mart with month-end snapshots, governed metric definitions, generation from templates, reconciliation checks as gates, human commentary, judgment and sign-off, incremental migration and a parallel run against the manual pack.

Data Quality Controls in Reporting

Why it matters. A report is only as good as its data. During the 2008 crisis, many banks could not aggregate their exposures quickly or accurately. In response, the Basel Committee published BCBS 239 (2013), its principles for risk data aggregation and risk reporting, and supervisors still list data quality among banks' most persistent weaknesses. For treasury, a wrong LCR or a missing deposit feed can lead to wrong funding decisions or to incorrect regulatory reports.

BCBS 239 in engineering terms.

Principles	Requirement	Engineering implication
1-2: Governance and infrastructure	Clear ownership, integrated architecture, common taxonomy	Data owners, data dictionary, shared reference data
3: Accuracy and integrity	Data reconciled and largely automated	Reconciliations, minimal manual steps
4: Completeness	All material risk data captured	Completeness controls per feed and entity
5: Timeliness	Data available when needed, including in stress	Deadlines, monitoring, alerting
6: Adaptability	Ad hoc and stress requests can be met	Flexible data model, self-service queries
7-11: Reporting	Reports accurate, comprehensive, clear, frequent enough and distributed properly	Report-level controls, distribution lists, sign-off

Data quality dimensions.

Dimension	Question	Example control in treasury reporting
Completeness	Is everything there?	Record counts and control totals from source match what was loaded; every entity and feed arrived
Accuracy	Is it correct?	Balances reconcile to the GL; prices pass IPV
Validity	Does it follow the rules?	Maturity date after start date; valid ISO currency code; valid HQLA level
Consistency	Do sources agree?	Liquidity and ALM engines show the same deposit totals
Timeliness	Is it current?	Feeds arrive before cut-off; market data date equals the reporting date
Uniqueness	Are there duplicates?	One record per trade ID and version
Plausibility	Does it look reasonable?	Day-on-day change in an LCR outflow category above a threshold is flagged

Where controls sit.

  Source systems
     | [C1] Ingestion: record counts, control totals, schema, arrival time
     v
  Staging
     | [C2] Validation: validity rules, referential integrity, duplicates
     v
  Calculation engines
     | [C3] Reconciliation: to the GL and between engines
     v
  Results
     | [C4] Plausibility: variance against the previous day and thresholds
     v
  Reports
       [C5] Sign-off: data owner attestation, data quality scorecard attached

  Failed control --> severity --> critical: block publication
                                  major:    publish with caveat and approval
                                  minor:    log and track

Example: a plausibility check. A day-on-day variance check flags items that moved by more than both a relative and an absolute threshold, and also reports items that appeared or disappeared:

from decimal import Decimal

def variance_breaches(today: dict[str, Decimal], yesterday: dict[str, Decimal],
                      rel_threshold: Decimal, abs_threshold: Decimal) -> list[str]:
    issues = []
    for key in sorted(today.keys() | yesterday.keys()):
        if key not in today:
            issues.append(f"{key}: missing today (was {yesterday[key]})")
        elif key not in yesterday:
            issues.append(f"{key}: new today ({today[key]})")
        else:
            change = today[key] - yesterday[key]
            base = abs(yesterday[key]) or Decimal(1)
            if abs(change) > abs_threshold and abs(change) / base > rel_threshold:
                issues.append(f"{key}: {yesterday[key]} -> {today[key]} ({change:+})")
    return issues

# LCR outflows by category, EUR m
today = {"retail_stable": Decimal("1210"), "retail_less_stable": Decimal("2950"),
         "wholesale_operational": Decimal("4100")}
yesterday = {"retail_stable": Decimal("1205"), "retail_less_stable": Decimal("2380"),
             "wholesale_operational": Decimal("4090"),
             "wholesale_non_operational": Decimal("3600")}
variance_breaches(today, yesterday, Decimal("0.10"), Decimal("100"))
# ['retail_less_stable: 2380 -> 2950 (+570)',
#  'wholesale_non_operational: missing today (was 3600)']

Using two thresholds avoids noise from small lines with large percentage moves and from large lines with small percentage moves. The second finding is the dangerous one: a missing feed of EUR 3.6bn of outflows would make the LCR look better. Missing data often flatters liquidity metrics, so completeness controls must block publication rather than merely warn.

Issue management. Every failed control creates an issue with a severity, an owner and a root cause. Fix issues at the source, because fixes applied in the report come back every month. Manual adjustments ("overlays") are recorded with a reason, an approver and an expiry, and are reversed in the next period unless renewed.

Engineering view.

• Write controls as versioned code that runs automatically in the pipeline. Store each result (control ID, run ID, status, measured values) as data that can be shown to auditors.
• Do not allow controls to be disabled or thresholds changed without approval and an audit record.
• Maintain lineage from each report cell back to source fields, so that an issue can be traced and its impact assessed.
• Attach a data quality scorecard to each report version, so readers know which caveats apply.

Analogy. Data quality controls are like an airline's pre-flight checklist: almost every item passes on almost every day, which is exactly why the list must be run in full every time.

Key takeaways

• BCBS 239 sets expectations for accurate, complete, timely and adaptable risk data aggregation and reporting.
• Controls cover completeness, accuracy, validity, consistency, timeliness, uniqueness and plausibility, at ingestion, validation, reconciliation, results and sign-off.
• Missing data often makes liquidity metrics look better, so completeness failures must block publication.
• Controls should be automated, versioned and auditable, with issues fixed at the source and overlays approved and time-limited.

🧪 Practice

1. Beginner: Assign each control to a data quality dimension: (a) record count from source equals record count loaded; (b) no trade ID appears twice; (c) the market data date equals the reporting date; (d) deposit totals agree between the liquidity and ALM engines.
2. Intermediate: Explain why missing data often makes the LCR look better, and design a completeness control that would catch a missing feed of wholesale deposits even if the feed file arrives empty.
3. Interview: "A supervisor's BCBS 239 review finds that your liquidity reports depend on 40 manual adjustments each month. What is your remediation plan?" Hint: inventory the adjustments and group them by root cause; fix them at the source in order of materiality; turn recurring ones into governed rules; put controls on the remaining overlays; track the reduction as a metric; document lineage.

---

<a id="15-engineering-reliable-and-secure-treasury-software"></a>