What is investment bank collateral call reconciliation testing?

Investment bank collateral call reconciliation testing checks that collateral calls calculated, sent or received by a bank agree with the relevant counterparty records and the underlying agreement terms. It validates the amounts and data used to produce each call, then checks that breaks are identified and routed correctly through dispute and settlement workflows. The precise rules depend on the portfolio, agreement and margin process in scope.

What investment bank collateral call reconciliation testing covers

A test usually compares records from systems such as trade capture, valuation, risk, collateral management and payment processing. It may also compare internal records with a counterparty statement or an external collateral platform. The comparison needs a defined scope: legal entity, counterparty, netting set, agreement, call date and valuation cut-off.

The expected call depends on the applicable Credit Support Annex (CSA) or other collateral agreement. Relevant inputs can include trade population, mark-to-market exposure, collateral already held or posted, thresholds, minimum transfer amounts, eligible collateral, haircuts, currency conversion, rounding rules and pending movements. Some processes also handle initial margin separately from variation margin. These rules are agreement-specific, so test expectations should come from approved requirements and configured terms rather than a single assumed formula.

Reconciliation testing can cover both the call calculation and the resulting operational state. Examples include checking that a call has the correct direction and amount, that a message contains the right counterparty and currency, and that an accepted, disputed, partially met or settled call is reflected consistently across connected systems.

How to test the reconciliation

Build test cases around known inputs and expected outcomes. Include ordinary calls as well as boundary conditions, such as an amount at the minimum transfer threshold, a threshold change, a currency conversion requiring rounding, an ineligible asset, or collateral with a haircut. Use controlled trade, market data and agreement records so that a discrepancy can be traced to its source.

Check that the systems use a consistent as-of time and trade population. A call may appear incorrect if one side includes a late trade or uses a different valuation snapshot. Other common failure modes include stale foreign exchange rates, a missing collateral movement, a duplicated message, incorrect sign conventions, mismatched agreement parameters and an incorrect settlement status.

Tests should verify break handling as well as matching. A genuine difference should be recorded with enough information to investigate it, such as the affected netting set, call reference, amount, currency and source records. A disputed or unmatched call should not be silently marked as reconciled. Reprocessing the same input should also be tested for duplicate calls or movements, where the system supports replay or retry.

Practical example

Assume an agreement-specific test case has a calculated exposure of £1,250,000 and £900,000 of collateral already held. For this simplified case, the applicable threshold is £100,000, there are no other adjustments, and the call direction is to receive collateral. The expected call is £250,000. The test should also confirm that the amount exceeds the configured minimum transfer amount.

If the counterparty record shows £240,000, the reconciliation should report a £10,000 break rather than mark the call as matched. The test then checks whether the difference can be traced to an input or rule, such as a missing trade or a different valuation snapshot, and whether the call enters the configured investigation or dispute process. The example is illustrative; actual calculation rules must reflect the applicable agreement and bank configuration.

Verifying test outcomes

Compare source and target records at field level, not just by total value. Confirm amount, currency, direction, agreement and netting-set identifiers, valuation timestamp, call status and relevant movement references. For automated runs, retain the input data, expected results, actual results and break details so that failures can be reproduced and reviewed.

Run the relevant scenarios across system integrations and regression tests after changes to agreement configuration, valuation feeds, collateral eligibility rules or message processing. This connects reconciliation results to the systems that create and settle calls, and helps distinguish a calculation defect from a data, timing or workflow issue.