# Building the data governancedata governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.Voir la définition complète → operating model
A NAV (Net Asset Value, the per-share price of a fund calculated daily) publishes 12 basis points too high because someone changed a fee accrual in a spreadsheet and no one owned the change. The fund restates. The client complains. The regulator asks who was responsible, and the honest answer is: nobody, specifically. That gap is what a data governancedata governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.Voir la définition complète → operating model closes.
Governance fails when it lives in a policy PDF nobody reads. It works when specific humans own specific datasets, meet on a schedule, and hold defined decision rights. This lesson shows you how to stand that up.
A data governancedata governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.Voir la définition complète → operating model is the answer to four questions:
Notice none of that is technology. Tooling comes later. First you assign people.
Data Owner. A senior accountable person, usually a business lead. The Owner does not clean data. The Owner is answerable for it. For NAV feeds, the Owner might be the Head of Fund Accounting. For fee schedules, the Head of Product or Client Servicing.
Data Steward. The hands-on custodian. Stewards define what "correct" means for a field, run checks, and triage exceptions. A benchmark steward knows that the S&P 500 rebalances quarterly and that a constituent add/drop must flow into the portfolio system before the next rebalance date.
Data Custodian. Usually IT or a platform team. Custodians run the pipelines and storage. They keep the data flowing and secure but do not decide business meaning.
Split them cleanly. When the Owner, Steward and Custodian collapse into one overworked person, accountability dies the moment that person leaves.
"We own dataown dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.Voir la définition complète → quality" means nothing. Assign at the dataset level. Below is a starter accountability mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.Voir la définition complète → for a mid-size asset manager.
| Dataset | Example content | Owner (role) | Steward (role) |
|---|---|---|---|
| Benchmark constituents | Index members, weights, rebalance dates | Head of Investment Data | Index Data StewardData StewardA business-side owner responsible for the quality, consistency and appropriate use of data in their domain.Voir la définition complète → |
| Fee schedules | Management fees, performance fees, breakpoints | Head of Product | Product Data StewardData StewardA business-side owner responsible for the quality, consistency and appropriate use of data in their domain.Voir la définition complète → |
| NAV feeds | Daily fund prices from the fund accountant/administrator | Head of Fund Accounting | Fund Ops Steward |
| Security master | Instrument identifiers (ISIN, CUSIP), asset class | COO Data function | Reference Data StewardData StewardA business-side owner responsible for the quality, consistency and appropriate use of data in their domain.Voir la définition complète → |
| Client/investor records | Beneficial owner, jurisdiction, KYC status | Head of Client Servicing | Client Data StewardData StewardA business-side owner responsible for the quality, consistency and appropriate use of data in their domain.Voir la définition complète → |
Two definitions for the non-technical:
The mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.Voir la définition complète → is the artifact you keep current. When someone resigns, you reassign the role, not rebuild the knowledge.
Roles without forums drift. You need standing decision bodies with clear scope.
Data Governance Council. Meets monthly or quarterly. Chaired by a senior sponsor (often the COO or a Chief Data Officer). It approves policy, arbitrates cross-team disputes, and signs off major changes (new data vendor, new golden source).
Data Quality Working Group. Meets weekly or biweekly. Stewards and custodians review open exceptions, aging issues, and upcoming events (index rebalances, fee changes, fund launches).
Change Advisory. Reviews proposed changes to critical datasets before they go live. A fee schedule change that hits client billing should never reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.Voir la définition complète → production without a documented approver.
Keep minutes. Regulators and auditors ask "who decided this and when." Minutes are your answer.
Write down who is Responsible, Accountable, Consulted and Informed for each decision type. Example for a fee schedule change:
The single rule that saves you: exactly one Accountable per decision. Two Accountable parties means zero.
This is asset management, so the operating model must mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.Voir la définition complète → to real rules. Name them, so your roles trace to obligations.
MapMapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.Voir la définition complète → each dataset to its regulatory hooks. Client records touch GDPR. NAV and fee data touch SEC/FCA records rules. Transaction reference data touches MiFIR reporting. Now your Owner knows why their dataset matters beyond "it should be right."
The operating model becomes real when each dataset carries automated checks that flag to its steward. A simple daily NAV sanity check, expressed as pseudo-SQLSQLSales Qualified Lead: a prospect the sales team has validated as ready for direct outreach and a proposal, having passed clear qualification criteria.Voir la définition complète →, might read:
-- Flag NAV moves that exceed a threshold vs prior day
SELECT fund_id, nav_date, nav_per_share, prior_nav,
ROUND((nav_per_share - prior_nav) / prior_nav * 100, 2) AS pct_change
FROM nav_daily
WHERE ABS((nav_per_share - prior_nav) / prior_nav) > 0.05 -- 5% flag
ORDER BY ABS(pct_change) DESC;A 5 percent single-day move on a diversified bond fund is almost certainly a data error, not a market move. The check routes to the Fund Ops Steward, who confirms or corrects before publication. The threshold itself is a governance decision, owned and documented.
Worked example of a materiality trigger: if a fund's NAV is 20.00 and the error is 0.02 per share, that is 0.10 percent, or 10 basis points. Many fund administrators use a NAV error tolerance around 0.5 percent (50 bps) as a common industry threshold for investor compensation, though the exact figure varies by fund, jurisdiction and prospectus (treat as an illustrative estimate, not a legal standard). Your governance model decides the threshold and who signs off breaches.
Vérification des acquis
1. The lesson opens with a NAV misstatement where a regulator asks who was responsible and the answer is 'nobody, specifically.' What core problem is this example designed to illustrate?
2. A Data Owner discovers a quality issue in the NAV feed. According to the operating model, what is the Owner's proper role?
3. Why does the lesson insist that the Owner, Steward, and Custodian roles be split cleanly rather than combined in one person?
4. Select ALL correct answers about what a data governance operating model defines.
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers that correctly distinguish the three core roles.
Sélectionnez toutes les réponses correctes.
The whole point is durability. Three practices make governance outlast the people who built it.
1. Document the definition, not just the value. For each critical field, record what it means, its source of truth ("golden source"), its owner, and its check. When the benchmark steward leaves, the definition of "constituent weight as of rebalance close" stays.
2. Name backups. Every Owner and Steward has a named deputy. No single point of failure.
3. Run a data catalog. A catalog is a searchable inventory of datasets with owner, steward, source, sensitivity and lineage. Open tools like OpenMetadata or DataHub do this for free. When an auditor asks "where does the fee schedule live and who owns it," the answer is one search, not one week.
Governance you cannot evidence is governance you do not have. Schedule light-touch checks: