# Governing investment data under regulation
A prospective client asks a simple question: "How did you calculate the 12.4% return you put in the pitch deck?" If your firm cannot reconstruct that figure from source data within hours, showing every input, adjustment, and approval, you have a governance problem, not a math problem.
Regulators ask the same question, only with subpoena power. This lesson is about building the controls, lineage, and audit trails that let you answer confidently before anyone asks.
A marketed track record is not just marketing. It is a factual claim about past results that three regimes scrutinize:
The common thread: every number you show must be reproducible, substantiated, and retained. That is a data problem before it is a compliance problem.
Take that 12.4% return. It is the end of a pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète → with many failure points.
Inputs:
Transformations:
Each step is a place where an error, an undocumented override, or a stale price can silently corrupt the marketed number. Governance means making each step visible, controlled, and replayable.
Data lineageData lineageData lineage maps how data moves and transforms across systems, from origin to consumption, showing where it came from, what changed it, and where it goes.Voir la définition complète → is the documented trail showing where a data point came from and every transformation it passed through. For performance data, good lineage lets you answer: "Which price file, which fee schedule, which calculation version produced this composite return on this date?"
A concrete test: pick any figure in a client presentation from 18 months ago. Can you regenerate it exactly, using the data and code as they existed then? If a vendor later restated a price, or someone patched the return engine, naive reproduction will produce a different number and undermine your credibility.
The fix is point-in-time (bitemporal) data: storing both when an event occurred and when you recorded it. This lets you query the data "as it was known" on any past date.
-- Reconstruct a composite return as it was known on 2024-06-30,
-- ignoring later restatements
SELECT composite_id, return_pct, calc_version
FROM composite_returns
WHERE composite_id = 'GLOBAL_EQ'
AND period_end = '2024-05-31'
AND knowledge_date <= '2024-06-30' -- as-of the marketing date
ORDER BY knowledge_date DESC
LIMIT 1;The knowledge_date column is what separates a defensible archive from a rolling number that quietly changes under you.
Lineage tells you what happened. Controls stop errors before they 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 → a client deck. Build them at each pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète → stage.
Input validation:
Calculation controls:
Composite controls (GIPS-specific):
Marketing controls:
The CFA Institute publishes the full standards and guidance; the GIPS Standards resources are free and worth bookmarking as the authoritative source.
An audit trail is an immutable, timestamped log of every action taken on the data and the process. When a regulator asks how a figure was produced, the audit trail is your evidence.
For each marketed number, you should be able to show:
Manual overrides deserve special attention. They are legitimate (a vendor price is obviously wrong, so an analyst corrects it) but they are also where fraud and error hide. Every override should be logged, justified, and approved by someone other than the person making it (segregation of duties). An override with no recorded reason is exactly what an examiner circles.
Both MiFID II and the SEC Marketing Rule impose record-keeping obligations. The precise retention periods and formats depend on your jurisdiction, entity type, and the specific record, so confirm current requirements with compliance counsel rather than assuming a single number.
The governance principle is stable regardless of the exact period: retain the full reconstruction package, not just the final PDF. That means the data snapshot, the calculation version, and the approval log, stored so they cannot be silently altered. A marketing document you can show but cannot substantiate from source is a liability, not an asset.
Vérification des acquis
1. The lesson argues that being unable to reconstruct a marketed 12.4% return from source data is fundamentally what kind of problem?
2. What is the 'common thread' the lesson identifies across GIPS, MiFID II, and the SEC Marketing Rule?
3. Why does the lesson describe a marketed track record as 'a regulated asset' rather than mere marketing?
4. Select ALL correct answers. Which of the following are described as inputs feeding into a performance figure's pipeline?
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers. Which statements accurately characterize the regulatory regimes discussed in the lesson?
Sélectionnez toutes les réponses correctes.
Controls, lineage, and audit trails only work if someone owns them. A workable model assigns clear roles:
Data owners are accountable for the accuracy of a data domain (for example, the head of portfolio accounting owns transaction data). Data stewards run the day-to-day quality checks. Compliance owns the marketing sign-off and the mapping of controls to specific rules. A GIPS verifier (an independent third party) can review your claim of compliance, which many firms use to strengthen credibility with institutional clients.
Tie each control to the rule it satisfies. A simple control-to-regulation mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.Voir la définition complète → turns a vague policy into an auditable system:
| Control | Satisfies |
|---|---|
| Net-of-fees presentation check | SEC Marketing Rule |
| "Fair, clear, not misleading" review | MiFID II |
| Composite inclusion rules | GIPS |
| Point-in-time reconstruction | All three |
| Override approval log | All three |
When an examiner arrives, you hand them this mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.Voir la définition complète → and the evidence behind each row. That is the difference between a scramble and a routine exam.
If your firm is early on this, sequence the work:
1. Inventory every performance figure that gets marketed and trace its pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète →.
2. Fix reproducibility first: implement point-in-time storage and version-locking so numbers stop drifting.
3. Add input and calculation controls at the highest-risk stages (pricing, composite construction).
4. Formalize audit trails and override approvals.
5. Map controls to regulations and layer in compliance sign-off.
Reproducibility is the foundation. Controls and audit trails are worth little if the underlying number cannot be regenerated in the first place.