Depositary Internal Controls: HKSI Paper 18
A depositary control system should assign accountable owners, preserve independence, obtain reliable information, reconcile assets, monitor delegates and escalate exceptions.
- The design must fit the scheme and applicable SFC requirements.
- A written policy matters only when staff follow it, evidence decisions and correct control failures.
On this page6 sections
Map the process before testing it
An internal control review begins with the workflow. Map who receives instructions, who checks them, who authorizes cash or asset movements, who reconciles records and who reviews exceptions. Include outsourced providers and the systems that transfer data. A policy that omits the actual service path can look complete while leaving a practical gap.
For every step, identify the risk, preventive check, detective evidence, owner and escalation path. A maker-checker control may prevent one user from releasing an unauthorized payment, while a reconciliation may detect an error afterward. Neither control works if access rights are stale or exceptions are routinely closed without evidence.
Governance and independence
Senior management should receive information that makes risks visible, not only operational totals. Report aged exceptions, recurring root causes, late data, unresolved valuation challenges and delegate weaknesses in a way that supports decisions. Escalation thresholds should distinguish routine corrections from events requiring urgent attention.
Independence can be tested through actual influence. If a manager can block the depositary’s access to valuation data, delay an issue from reaching the board or tie compensation to an outcome under review, the depositary should assess the conflict and apply safeguards. Document the residual risk and who accepted it.
The depositary should have clear accountability for oversight and access to people with sufficient authority to challenge the manager or delegate. Identify conflicts involving affiliates, service fees or shared personnel. Establish independent review, separate access or other measures suited to the risk, then test whether the process works in practice.
Escalation routes should avoid filtering material concerns through the party whose conduct is under review. Staff need a way to raise unresolved issues promptly. Minutes and decision records should capture evidence considered, dissent, conditions and follow-up. A control framework is stronger when a second reviewer can understand the reasoning without relying on informal explanations.
Data quality, reconciliations and monitoring
Control owners need complete and timely data. Define source systems, cut-offs, fields, frequency and validation checks. When information arrives late or changes unexpectedly, trigger an exception rather than letting a spreadsheet silently overwrite the prior record. Reconcile cash, positions, investor flows and valuation inputs against independent evidence according to the relevant risk.
Monitoring metrics can show missed reports, aged breaks, repeated manual adjustments, late settlements or valuation overrides. Metrics help prioritize review; they do not replace investigation. A green dashboard can conceal a poorly designed threshold or incomplete population. Test samples back to source records and question unexplained trends.
Delegated service controls
Due diligence should examine a delegate’s capability, financial and operational resilience, conflicts, systems, staff, subcontracting and access to records. Contract terms need timely reporting, audit or information rights and clear service expectations. The depositary should know how it will respond when the provider misses a service level or the underlying control fails.
Ongoing oversight can include incident reporting, reconciliations, control attestations, site or system reviews and analysis of exceptions. Use frequency and depth proportionate to the function and risk. An annual certificate is one input, not proof that every transaction was processed correctly. Retest material remediation to confirm that it fixed the cause.
Incident response and remediation
For a recurring exception, identify whether the root cause is people, process, data, technology or oversight. A delayed report caused by a broken data interface needs a different fix from repeated manual override by a user with excessive access. Assign a control owner, due date and success measure. “Retrain staff” is not enough if the system still permits the error.
Retest the control using a period that includes normal variation and exceptions. Check that the changed procedure works for different schemes and does not shift risk to another step. Close remediation only after evidence demonstrates the fix and the responsible reviewer approves closure. Preserve the original incident trail so the firm can explain how it learned from the failure.
When a control fails, first understand whether assets, valuation, dealing or investor outcomes are affected. Preserve records, contain ongoing risk and identify the population of transactions or schemes involved. Set a named owner and timetable for investigation. Escalate promptly when the impact could be material or the issue suggests misconduct.
Remediation should address the cause. If reconciliations repeatedly fail because source data arrive late, adding another sign-off may conceal the problem. Improve data delivery, system validation or responsibility allocation, then test the new control. Document whether any investor correction, disclosure or regulatory notification is needed under the relevant rules.
Testing a control with an example
Review who can change the control parameters and who approves changes. A threshold that is manually increased to clear aged breaks can mask a real problem. Keep version history, approval evidence, test results and periodic recertification. This is especially important when the same system supports several funds with different terms.
Control design should consider continuity: what happens when the primary reviewer is absent, a system is unavailable or a delegate misses a reporting cycle? Document an alternate owner and a safe fallback. A workaround should preserve evidence and be time-limited; otherwise, it can become an unreviewed permanent process.
A reviewer should select samples that cover risk, not just convenience. Include new accounts, changed instructions, high-value transfers, aged reconciliation breaks, manual adjustments and transactions processed during system incidents. Document the selection logic and any limitation. A handful of routine transactions may confirm that the normal path works while missing the failure paths.
For each exception sample, verify closure evidence rather than relying on an owner’s status field. Check the source record, correction, approval and downstream effect. If the same discrepancy appears again, reopen the root-cause analysis. Repeated closure without prevention can indicate the control is treating symptoms rather than reducing risk.
A control should have a meaningful escalation threshold and closure criteria. If every small break goes to the same senior committee, significant events can be buried in noise. If thresholds are too high, repeated small discrepancies may never be recognized as a systemic failure. Review trends and qualitative context together.
A control test should challenge both design and operation. Design asks whether the control could prevent or detect the risk if performed as intended. Operation asks whether the right person performed it, with complete information, at the right time, and whether exceptions were resolved. A signed checklist may pass a superficial test while source evidence reveals missing accounts.
For a sample cash movement, trace from investor instruction to authorization, bank settlement and scheme ledger. Check for independent verification of changed details, duplicate detection, reviewer access and evidence that a rejected instruction did not later enter through another channel. Test exceptions, not only successful transactions.
When a control test fails, determine whether the sample is isolated or indicates a broader issue. Expand the population where warranted, identify impacted schemes and assess whether the failure changes NAV, investor dealing or asset ownership. Document why the test scope is sufficient. Remediation should be retested by someone able to assess the fix independently.
Consider a nightly file that feeds cash balances into a depositary dashboard. A control design should define expected totals, reject incomplete files, preserve prior data, alert on exceptions and prevent unverified amounts from authorizing payments. Test the process with a missing file, a duplicate record, an unexpected currency and a stale balance. Each should create a different traceable result.
Common questions
Which source should I use if a study note conflicts with a regulator page?
Record the discrepancy and use the applicable official source.
Does passing Paper 18 itself authorize Type 13 activity?
No. It is an examination component. Licensing also depends on the relevant SFC and firm requirements, including the candidate’s role and competence.
Are the topic percentages published?
The public Paper 18 syllabus identifies four topics but does not give a per-topic question allocation. Treat any unofficial weighting as an estimate.