One Customer Number, Ninety Days, Three Definitions
Northgate Financial is eleven months past its acquisition of a regional insurer. The board has committed to a single customer count in its next regulatory filing, due in 90 days. Three definitions are already in production: Retail Banking counts anyone with a transaction in the last 90 days (2.1M); Insurance counts anyone with a policy in force in the last 12 months (740k); Digital counts anyone who logged into the app in the last 30 days (1.3M). The CFO sponsors the data programme, wants one number, and does not care which. Fourteen data domains have been declared; four have accountable owners. The data steward council meets monthly for two hours and has no decision rights. The CDO has $400k and six staff. Separately, General Counsel has flagged that the same customer appears under three different party identifiers across the two core systems, and no one can currently prove which identifier belongs to a real person.
1. The CDO must decide what is deliverable in 90 days. What belongs in scope, what must be explicitly declared out of scope, and which governance artefacts must exist before the regulatory number can be defended to an auditor?
Read the worked answer
Start from the obligation, not from the fourteen domains. The driver is a filing date, so the deliverable is one defensible metric with traceable lineage, not an enterprise dictionary.
In scope (90 days)
- One conformed Party concept: identifier rules, survivorship, and a resolution routine that proves the three identifiers map to one person.
- One Regulatory Customer metric: purpose-built definition, inclusion/exclusion rules, versioned.
- A reconciliation bridge showing how 2.1M, 740k and 1.3M relate to that figure.
Explicitly out of scope
- The other ten undeclared domains, catalogue tooling, and enterprise-wide vocabulary. Say so in writing; unstated scope becomes the reason the programme is judged to have failed.
Mandatory artefacts
- A decision record signed by the CFO naming the accountable owner of the metric.
- The definition itself with business rules and a version number.
- Documented source-to-figure lineage for each input field.
- A change-control route: no rule change after the freeze date without CFO sign-off.
The critical judgement is that a governance council with no decision rights cannot own this. Escalate decision rights for this one metric to a small CFO-chaired forum for the 90 days. A number produced without an accountable decision-maker is a number nobody will defend when the regulator queries it.
2. Should the organisation converge on one active-customer definition, or keep three? Recommend a position and justify the trade-offs, including what happens if the regulator prescribes the term.
Read the worked answer
Keep three operational measures; publish one regulatory measure. Dissolve the ambiguity by making the relationship between them explicit rather than by deleting information.
Reasoning
- The three numbers answer different questions: recency of transaction, policy liability, and digital engagement. Collapsing them into one destroys the operational meaning that Retention, Underwriting and Product actually use, and teams will quietly rebuild their own versions - producing shadow metrics, which is worse than the current state.
- The real defect is not the count of definitions. It is the absence of a traceable bridge between them, so no one can explain why 2.1M and 1.3M differ.
Recommended structure
- Level 1: conformed Party (one person, resolved identifiers).
- Level 2: purpose-specific metrics, each with a named owner, a stated purpose, and a mapping rule into the Party set.
- Level 3: the regulatory metric, derived only from Level 2 with documented inclusion rules.
Regulatory override Check first whether the filing prescribes the term. Where a regulator defines the measure, that definition wins and internal preference is irrelevant; the internal variants then become diagnostics, not alternatives. If the term is prescribed, the 90-day work shrinks considerably and the bridge becomes the main deliverable.
Trade-off accepted Three published numbers create interpretive burden. Mitigate with a single bridge table and a rule that no external communication uses a Level 2 metric without the Level 3 figure alongside it.
3. The CFO asks how governance success will be measured after the filing. Propose a measurement set that would distinguish real progress from activity.
Read the worked answer
Reject activity counts - meetings held, glossary terms published, stewards trained - as primary evidence. They rise whether or not anything improves. Use outcome and control indicators with baselines taken now.
Control effectiveness (lagging)
- Reconciliation break rate between the regulatory figure and source systems, expressed in rows and currency, reviewed monthly.
- Number of external or audit findings attributable to definition or lineage gaps.
- Share of critical data elements with both an accountable owner and an approved, versioned definition (baseline is four domains of fourteen).
Speed and behaviour (leading)
- Median time to resolve a definition dispute, from ticket open to decision recorded. This is the single best proxy for whether the council has real decision rights.
- Number of production metrics in use that are not traceable to an approved definition - the shadow-metric count. It should fall, and it will only fall if the official route is faster than the unofficial one.
- Percentage of metric changes that passed change control before release.
Value framing for the CFO Translate at least one indicator into money or risk exposure: cost of manual reconciliation effort, or the capital implication of reporting 2.1M versus 1.3M.
Sequencing caveat Do not launch all of these at once with six staff. Instrument dispute-resolution time and shadow-metric count first; they are cheap, they diagnose the governance mechanism itself, and they will tell you within two quarters whether the operating model is working or merely present.