Client impact · Provincial state capital
Three tiers stop producing three different numbers
A provincial group, its municipal platforms and its county entities each defined "invested capital" their own way, and every filing season ended in manual reconciliation. We changed three things: definitions signed off in one data dictionary, one exit for filings, permissions cut down to the field. Acceptance uses the same four measures published on the client impact index — no private formula.
The institution
One provincial group, three layers of filing duty
The capital is provincial state-owned operating and fiscal money. The structure is three tiers: the provincial holding, municipal platforms, county-level investment entities. Investment work sits in the group investment department, the finance arm and each platform’s own development office. The group reports to the shareholder authority; the municipal platforms report separately to their local counterpart; the county entities report project by project.
The estate was not empty. Accounting, office automation and approval workflows had run for years. What never got systematised were the four objects that matter to supervision: investments, funds, shareholdings and valuations. A single equity transfer therefore appeared as two rows with two amounts — and both rows were defensible.
The constraints were hard: a localised IT stack, private or intranet deployment, horizontal isolation between municipal platforms, and no missed filing window during the cutover.
Before the change: where the effort went
This describes mechanism and workload structure, not percentages. A time saving for your group can only be computed from your own ledger during the assessment; any figure quoted before that is decoration.
| Step | Who carried it | How the effort was spent | Why it never closed |
|---|---|---|---|
| Defining metrics | A statistics or finance post in each tier | The same metric defined three times over three spreadsheets; differences settled by email chains | No owner of the definition and no signed version, so one change meant recomputing history |
| Filing | Group investment department plus platform liaisons | The same investment reported on two templates at two reference dates, reconciled by hand at quarter end | Exits were spread across tiers, so the consolidated pack could only compare, not constrain |
| Consolidation and checks | One or two long-serving group staff | Personal memory decided which version was right; validation rules lived outside the system | The knowledge sat in people, so rotation or leave produced a backlog |
| Permissions and trace | IT function plus business units | Menu-level grants, one full export available to many roles, boundaries enforced by policy | Menu permissions cannot express "which tier sees which field" |
The judgement
The bottleneck is definitional authority, not tooling
Every tier already knows how to build a table. What was missing was one definition set, signed by business, finance and risk, enforced inside the system, plus a single exit that can only draw from it.
So the first cut lands on the data dictionary, not on report development. That step writes no code and commits you to no vendor choice.
- Authority over definitions comes before tooling
- Name the field owner before automating
- Only one filing exit may be authoritative
- Keep both versions when tiers disagree; never average
Three steps we took
Fix the definition, close the exit, cut to the field
Each step names its deliverable and its reviewer. Everything outside these three steps kept running unchanged, so we never moved definitions and the organisation at the same time.
01 · Blueprint
Consolidate definitions
Invested amount, disbursed amount, ownership ratio and investment income are written once, with an owner, a scope and a source. Open disagreements go into a separate ruling sheet instead of being quietly resolved in software.
- Deliverable: data dictionary v1
- Deliverable: disagreement ruling sheet
- Reviewed by: investment, finance, audit
02 · Build
Converge the filing exit
One underlying pack maps to several filing templates. Template versions and submission timestamps are recorded in the system, so municipal and county filings are generated from the same fields rather than transcribed.
- Deliverable: template map and filing register
- Deliverable: exports leave a trace
- Reviewed by: group executive and tier liaisons
03 · Govern
Field-level rights and a supervision ledger
Permissions are set at function, data and field level with horizontal isolation between tiers. Shareholding changes enter the ledger by effective date, and sensitive operations trace back to the approval node.
- Deliverable: permission matrix
- Deliverable: look-through shareholding ledger
- Reviewed by: IT and compliance
Modules and deployment shapes in this scenario
Module names follow the company’s published product list. Actual scope follows your tender documents and the assessment findings; nothing here is a standard-function guarantee.
Acceptance measures
Same formula, different sample plan
This page does not invent metrics. Four measures are defined on the client impact index together with the trap in each; this engagement uses three of them — filing cycle, definition consistency and manual fallback hours. The wording of the formulas is unchanged. What we add is how the numbers were taken here, at what scope and on what date.
The rule still holds: numerator and denominator come from objects that already exist in the system. Anything else decays into a spreadsheet someone has to feed.
How we recomputed them here
| Measure | Calculation (as published) | Recomputation in this engagement | Sample and timing |
|---|---|---|---|
| Filing cycle | Calendar days from template issued to consolidated group submission | Taken from the issue and submission timestamps of the filing task, not from a rehearsal; one full filing season before and after go-live | Group plus every municipal platform, quarterly peak included, ad-hoc returns excluded |
| Definition consistency | Share of identical metrics where the three tiers agree with zero variance | Each metric agreed in the dictionary is compared across tiers; a variance only closes when it is explained by a definition source or a data source | County tier must be in the sample; sampling head offices alone inflates the figure |
| Manual fallback hours | Overtime hours during filing season ÷ people involved, from system-captured effort | Recomputed from logged effort records, never from recalled effort; filing fallback time is kept separate from implementation time | Reported per filing season across two consecutive seasons |
Without an owner for the definition, a system only copies disagreements faster and makes them look tidier.Catalpa delivery methodology
500+
institutions served on the platform (company disclosure, not this engagement)
14,000+
funds covered (platform book, not this client’s book)
5.6RMB tn+
assets covered (RMB, platform book)
15
offices nationwide, delivery near the client
Institutions publicly referenced
Names come from the company’s own published sites. Order carries no meaning, and listing them does not attribute this case to them — the institution described above is an anonymised archetype.
Named references with client authorisation are available on request during the assessment.
Where this approach does not apply
If the group has no controlled second or third tier, there is nothing to consolidate and no isolation to enforce; a disciplined template with a named owner costs far less. If no tier will staff the maintenance of the dictionary, we recommend the organisational agreement first and the system second — otherwise the definitions diverge again within a year and the platform simply fossilises the disagreement.
Five questions the procurement stage always asks
These are mechanism answers. They do not substitute for the written position of your supervising authority, and they are not a commitment on project duration.
Where do you draw the line on cleaning historical data?
When the three tiers disagree on a definition, who rules?
How do you keep filings running during cutover?
County entities have no dedicated investment staff. Who enters the data?
How is the localised IT stack validated before go-live?
Next step
Bring the report that gives you the most trouble
In a 45-minute assessment we take that one filing form apart live: field divergences, owners, and the exits that can be closed. No product demo — definitions only.