Ga naar hoofdinhoud

Portfolio rationalization (Gartner TIME)

Adds Gartner TIME classification (Tolerate / Invest / Migrate / Eliminate) to each application-in-use (gebruik), and a per-organisation portfolio rationalization report that combines TIME quadrant counts with existing end-of-support exposure (application-lifecycle-tracking), cloud-transition share (the existing cloudDienstverleningsmodel field), and annualised cost overlay (contract-administration). See VNG Softwarecatalogus issue #54.

Specification: openspec/specs/portfolio-rationalization-time/spec.md.

Screenshots of the report page are still pending a live-instance capture — this document ships with the implementation; the Playwright-captured images will follow in a subsequent docs pass (see the design.md Open Questions note on the report's live-verification follow-up).

Classifying a gebruik​

Three new optional fields on the existing gebruik schema — recorded per organisation's usage of an application, never on the module/application itself (mirroring how geplandeVervanging is scoped per gebruik):

FieldTypeNotes
timeClassificationenum (string)Tolerate | Invest | Migrate | Eliminate
timeRationalestringFree-text justification for the classification
timeReviewDatedate (string)When the classification should next be reviewed

A gebruik with no timeClassification set is Unclassified — it is never hidden from the report, and is excluded from every quadrant count until classified.

Editing happens through the app's generic schema-driven object editor (ObjectModal.vue, object-type-key="gebruik"), which now renders any enum-on-string schema property (not just the three TIME fields — this also improves status and every other enum field project-wide) as a clearable dropdown instead of free text. Because the editor reads the full current object before submitting (formData = cloneDeep(activeObject)) and only mutates the edited key, a TIME-only edit carries every other gebruik field forward unchanged — OpenRegister's saveObject is PUT-semantic, so omitted fields would otherwise be nulled out.

Cloud-transition share reuses the existing Hosting field​

No new deployment-model field is introduced. gebruik already carries cloudDienstverleningsmodel ("Hosting": On-premises (self-managed) / IaaS / PaaS / SaaS, facetable: true) — the report's cloud-transition metric reads that field directly, so the existing Hosting column and this report never fork into two competing sources of the same fact.

The portfolio rationalization report​

GET /apps/stackiq/api/portfolio-report?organisation={uuid}

Returns a bounded, organisation-scoped aggregate:

{
  "organisation": "org-a",
  "generatedAt": "2026-07-23T12:00:00+00:00",
  "pageSizeCeiling": 500,
  "totalGebruiken": 42,
  "includedGebruiken": 42,
  "truncated": false,
  "quadrants": {
    "Tolerate": { "count": 5, "eolExposedCount": 1, "cloudTransition": { "SaaS": 3, "On-premises (self-managed)": 2 }, "annualisedCost": 12000, "oneOffCost": 0 },
    "Invest": { "...": "..." },
    "Migrate": { "...": "..." },
    "Eliminate": { "...": "..." },
    "Unclassified": { "...": "..." }
  },
  "rows": [
    { "uuid": "g1", "moduleName": "Example App", "timeClassification": "Migrate", "quadrant": "Migrate", "timeRationale": "Vendor lock-in, successor selected", "timeReviewDate": "2027-01-01", "lifecyclePhase": "In productie", "eol": { "passed": false, "withdrawn": false, "endDate": "2027-06-01" }, "eolApproaching": false, "hostingModel": ["SaaS"], "annualisedCost": 4800, "oneOffCost": 0 }
  ]
}

Every figure is computed at query time, never persisted — the same "derive, don't cache" principle as the lifecycle-tracking phase/EOL rules. PortfolioReportService composes three existing derivations rather than forking a second implementation:

  • TIME quadrant counts — from gebruik.timeClassification.
  • EOL exposure — reusing the application-lifecycle-tracking end-of-support rule (a moduleVersie's datumEindeOndersteuning passed, or within the 180-day approaching window).
  • Cloud-transition share — from cloudDienstverleningsmodel.
  • Annualised cost overlay — reusing the contract-administration cost derivation (kosten × kostenPeriode) over each gebruik's linked contracts.

Pure phase/EOL/cost/relation-id rules live in PortfolioReportDerivation (no I/O), separated from the OpenRegister-querying orchestration in PortfolioReportService to keep both testable and under the project's complexity thresholds.

Bounded, always​

Every OpenRegister query the report issues carries an explicit _limit (or uses searchObjectsPaginated) — never an unbounded searchObjects() call, per the bound-unbounded-searchobjects-scans rule. The gebruik query is capped at a configurable page-size ceiling (portfolio_report_page_size_ceiling app config, default 500); when an organisation's gebruik count exceeds it, the response's truncated: true + totalGebruiken / includedGebruiken disclose the truncation — the report never presents a bounded subset as a silently-complete total. The linked-contract lookup is bounded to 5 × the gebruik ceiling (capped at 5000), since one gebruik may have several linked contracts (renewals, multiple services).

RBAC: scoped to the caller's authorised organisation(s)​

The controller resolves and checks the caller's organisation access before the service ever issues a query for the requested organisation (fail closed, per vendor-visibility-rbac):

  • admin / ambtenaar may request any organisation's report (the existing unrestricted-read bypass those roles already have on gebruik reads).
  • Every other authenticated user may request only their own active organisation's report — a report synthesises another organisation's gebruik/contract data, which is not granted beyond the caller's own organisation.
  • A request for an unauthorised organisation returns 403 with no organisation data in the response body, and the CSV export variant is denied the same way.

Re-verify this enforcement point once vendor-visibility-rbac lands as its own change — the gating mechanism may move from the IConfig::getUserValue('core', 'organisation') lookup used here to whatever canonical mechanism that change introduces.

CSV export​

GET /apps/stackiq/api/portfolio-report?organisation={uuid}&format=csv

The same bounded, RBAC-scoped row set as the JSON report, serialised as CSV (one row per gebruik: organisation, module, TIME classification, rationale, review date, lifecycle phase, EOL status, hosting model, annualised + one-off cost) — never a second, unbounded/unscoped export path.

Report page​

/portfolio-report (manifest page PortfolioReport, component PortfolioReportView) renders:

  • An organisation picker.
  • A TIME quadrant bar chart (CnChartWidget, apexcharts via @conduction/nextcloud-vue), one bar per quadrant including Unclassified, coloured with NL Design System semantic tokens (--color-success / --color-warning / --color-error / --color-text-maxcontrast / --color-border-dark — never a hardcoded hex).
  • A quadrant summary table (count, EOL-exposed count, cloud-transition mix, annualised + one-off cost).
  • Gebruik-level detail tables, one section per quadrant (Unclassified always rendered, even when empty).
  • A truncation banner when the report is bounded.
  • An Export CSV button.

Unlike LifecycleRoadmapView / LicensePostureView (which derive their figures client-side over a full collection fetched into the browser), this page is a thin renderer over the single composed backend endpoint above — the aggregation and RBAC scoping live server-side, per design.md Decision 3/4.