Independent Singapore VCC guidance
Direct answer
Treat every valuation model change as a controlled release, not a spreadsheet edit. Classify whether the change affects code, data, assumptions, methodology, overrides or reporting; preserve the current production version; test the proposed version on representative and stressed positions; explain every result difference; and obtain approval from people independent of development. Move it into production only with a named effective point, rollback route, user access controls and post-release checks tied to the correct VCC and sub-fund.
At a glance
- Separate methodological decisions from technical deployment.
- Compare old and new results across representative portfolios.
- Keep production code, inputs and approval evidence versioned together.
- Monitor the first live cycles and retain a workable rollback path.
Who this is for
- Internal or vendor valuation tools that influence a VCC NAV, price challenge or investor dealing decision
Important exclusions
- Selecting an accounting standard or replacing independent valuation, audit or legal advice
Classify the change before testing
Begin with a change statement that identifies the model, portfolio, VCC or sub-fund, instruments, users and decisions affected. A code correction, new market-data field, revised calibration, methodology choice, tolerance adjustment and user-interface change create different risks. Record whether the change fixes an error, responds to a new product, improves robustness or changes valuation judgement. If the team cannot state the purpose and expected result, it is not ready to test.
Sources: Monetary Authority of Singapore · Monetary Authority of Singapore| Change type | Primary risk | Evidence expected |
|---|---|---|
| Code or formula | Calculation behaves differently from design. | Unit tests, peer review and reproducible old-to-new comparison. |
| Data or mapping | Correct logic receives incomplete or mislabelled inputs. | Field lineage, stale-data tests and rejected-record handling. |
| Assumption or calibration | Judgement changes value without visible code movement. | Approval rationale, sensitivity analysis and effective date. |
| Workflow or access | Unauthorised users can alter or release outcomes. | Role test, change log and segregation of duties. |
Set acceptance criteria before seeing results
Define what must remain unchanged, which differences are intended, and which outcomes would stop the release. Acceptance criteria should cover calculation accuracy, input completeness, performance, access, exception handling, report presentation and downstream interfaces. Use absolute and relative tolerances only where their economic meaning is documented. A test that is adjusted after an unexpected result appears can become an explanation exercise rather than an independent gate.
Sources: Monetary Authority of Singapore- Name representative instruments, edge cases and stressed inputs before execution.
- Specify expected signs and directions for intended valuation changes.
- Require zero unexplained population loss between old and new versions.
- Define stop conditions for sub-fund attribution, stale data and override failures.
- Record who may approve exceptions and who may deploy the release.
Run parallel and sensitivity testing
Execute both versions from the same frozen input population. Reconcile every instrument, not only the portfolio total, and preserve model inputs, timestamps and outputs. Then vary the inputs that carry judgement or non-linearity to see whether the result behaves as designed. Include missing data, extreme but plausible values, stale feeds, zero positions, partial disposals and duplicated records. For an umbrella VCC, test ownership and expense allocation by sub-fund so offsetting errors do not disappear in the aggregate.
Sources: Monetary Authority of Singapore · Accounting and Corporate Regulatory Authority- FreezeLock the model versions, input files, market-data timestamp and instrument population used for the comparison.
- CompareCalculate both versions and explain differences at instrument, class, sub-fund and portfolio levels.
- StressChallenge boundary values, missing inputs, unusual lifecycle events and manual override behaviour.
- ReconcileTrace accepted outputs through administrator, accounting, risk and investor-reporting interfaces, recording every rejected or transformed field.
Related guidance: challenge stale data before VCC NAV
Approve methodology and deployment separately
A valuation committee may approve the economic methodology without being the right group to deploy code or administer user access. Keep those decisions linked but distinct. The approval pack should contain the change statement, intended impact, test population, results, unresolved differences, independent challenge, implementation window, communications and rollback route. Vendor assurance is useful evidence, but the VCC manager still needs to understand how the release affects its mandate and operating controls.
Sources: Monetary Authority of Singapore · Monetary Authority of SingaporeRelated guidance: VCC valuation governance review
Release with rollback and access control
Store the production package, configuration, input schema, approval reference and deployment record under one immutable release identifier. Limit who can alter assumptions, promote versions and approve prices. A rollback must restore not only code but also compatible data mappings, configuration and reports. Schedule the effective point around the valuation cycle, and tell administrators and reviewers which version owns each date. Do not permit silent vendor changes to become production merely because the hosted system updated successfully.
Sources: Monetary Authority of Singapore- Before releaseConfirm backups, user roles, interface readiness, deployment owner and conditions that trigger cancellation.
- At deploymentRecord the release identifier, effective timestamp, checksums or equivalent integrity evidence, and initial smoke tests.
- First live cycleReperform selected valuations, inspect exceptions, reconcile interfaces and compare realised behaviour with testing.
- StabilisationClose defects, review overrides, confirm user access and retain the former version for the approved retention period.
Related guidance: VCC NAV release approval matrix
Use post-release evidence to improve governance
Review the first live cycles for new overrides, slower calculations, unexplained dispersions, data fallbacks and user workarounds. Compare those observations with the release assumptions and decide whether the model is stable, needs remediation or should be rolled back. Record lessons by root cause rather than blaming the visible exception. Repeated emergency fixes often indicate weak inventory, unclear ownership or inadequate test design. Feed those findings into the next release gate and provider review. Compare realised valuation differences with the range presented to approvers, and reopen the decision if a new instrument, data source or market condition falls outside the tested population. A stable model also needs stable governance around inputs and human judgement. Retain enough evidence for audit, investor questions and later model comparisons, including the reason an apparently successful release was accepted.
Sources: Monetary Authority of Singapore · Monetary Authority of Singapore- Override volume and reasons by instrument and user.
- Data-source fallback frequency and unresolved stale inputs.
- Differences between test and live populations.
- Downstream breaks in accounting, risk and reporting.
- Open defects, compensating controls and the next formal decision.
Related guidance: VCC compliance checklist
Frequently asked questions
Does a vendor release remove the need for VCC testing?
No. Vendor evidence can support the review, but the manager must test the actual portfolio, configurations, data mappings, users and interfaces that affect its VCC. Hosted deployment success does not prove suitability for the mandate.
What is a material valuation model change?
Materiality depends on the decision affected, instruments, possible value movement, investor dealing and control impact. A small code change can be material if it changes ownership attribution or an important exception path.
Can the model developer approve production release?
Development expertise is necessary, but independent challenge and release authority should be preserved. The person who built or changed the model should not be the only person deciding that its results are acceptable.
How long should the old model remain available?
Use the fund’s records, audit and change-management policy. Keep enough code, configuration, inputs and output evidence to reproduce affected valuations and to execute the approved rollback while it remains a credible option.
Official sources and further reading
- VCC key features, eligibility and requirements (Accounting and Corporate Regulatory Authority)
- Overview of managing a VCC (Accounting and Corporate Regulatory Authority)
- Operational resilience (Monetary Authority of Singapore)
- Fund management licensing (Monetary Authority of Singapore)
Discuss your fund, family office or trust structure
Contact Raffles Corporate Services about Singapore VCC setup, fund structures, family offices and corporate administration.
- Email: hello@rafflescorporateservices.com
- WhatsApp: +65 8501 7133
- WhatsApp: +65 9018 0072
Where legal assistance is required, we can coordinate an introduction to an associate independent law firm, with your consent. The law firm advises you under a separate engagement. Raffles Corporate Services is not a law firm. Tax, investment management and trustee services, where needed, require appropriately qualified or authorised providers.
General information only. This article is not legal, tax, regulatory or investment advice and does not imply affiliation with or endorsement by ACRA, MAS or IRAS.