By Siddharth GangalA reporting-platform migration fails when dashboards look similar but numbers, permissions, refresh timing, filters, subscriptions or lineage changed. The portable unit is the reporting estate: source connections, transformations, datasets, semantic models, metric contracts, calculations, reports, filters, schedules, alerts, row-level access, subscriptions, lineage, credentials, audit history and consumer decisions.
This guide uses official documentation reviewed August 8, 2026. Gangly did not migrate a BI tenant or compare rendering. Export, API, history, configuration, permissions and lineage vary by platform, plan and contract. Buyer-verify every artifact and output.
Separate reporting migration from reporting automation
This page owns incumbent exit and cutover. The refreshed sales reporting automation guide owns architecture, metric governance and ongoing implementation. CRM reporting for managers owns manager usage, while the dashboard guide owns dashboard design. Migration preserves existing decision capability or deliberately replaces it under signed acceptance.
Freeze source/target, licenses/capacity, sources, warehouses, CRM/finance systems, gateways, regions, archive, holds, snapshot, cutoff, parallel period, decision authority, rollback, cancellation, retention and deletion. Name reporting, semantic, source, data engineering, CRM, finance, security, privacy, consumer, validation and rollback owners.
Hard gates include approved metric values and time semantics, correct row-level access, complete critical reports/subscriptions, traceable lineage, fresh data, visible failures, one decision authority, tested rollback and closure evidence.
Inventory sources, models, reports, delivery, and governance
Inventory data sources, accounts, APIs, files, warehouses, schemas, tables, views, extracts, gateways, refresh credentials, service accounts, connectors, query parameters, transformations, pipelines, schedules, caches and incremental state. Add datasets, semantic models, dimensions, measures, calculations, relationships, hierarchies, date/currency/timezone rules, joins, filters, row-count expectations and owners.
Inventory metric names, definitions, formulas, grain, authoritative source, inclusion/exclusion rules, status, version, effective date, tolerance and approver. Add dashboards, reports, pages, tiles, visuals, queries, drill paths, parameters, filters, bookmarks, exports, embedded views, accessibility and mobile behavior.
Include workspaces/projects/folders, users, groups, roles, row/object-level permissions, shares, subscriptions, schedules, alerts, recipients, delivery channels, usage, favorites, ownership, lineage, audit events, comments, certifications, endorsements, refresh/error history, credentials, API keys and deletion state. Each item needs stable ID, parents/edges, portability label, target, test and evidence.
Map lineage and metric decision authority
Build edges from source→connection→pipeline/transformation→dataset→semantic model→metric→report/visual→filter→schedule/subscription→consumer/decision. Add workspace→role→row policy; credential→connection; report→owner; alert→metric/threshold/recipient; and metric version→approver/effective interval.
Google documents LookML as a semantic modeling layer for dimensions, aggregates, calculations and relationships. It also documents Git options for LookML projects. Version-controlled model code is valuable, but does not automatically contain every dashboard, schedule, permission, credential, audit event or data snapshot.
Publish a metric authority contract. CRM may own current opportunity fields, a snapshot table “pipeline as known then,” and finance realized revenue. The semantic layer joins them under approved rules. Use the integration guide and CRM migration checklist when source authority also changes.
Classify portability and reconstruction
Use documented exportable, buyer verified, support/contract dependent, reconstruct, and unavailable. Record URL/ticket, plan/role, export/API, filters, format/schema, relationships, history, permission coverage, limits, sample, test date and exclusions.
Microsoft’s portability documentation distinguishes customer data from some configuration and protected operational data. Apply the boundary per artifact. A PDF preserves appearance but not formulas, filters, lineage or row security. Model code preserves logic but not necessarily data, credentials, subscriptions or audit history.
Request vendor manifests and perform representative exports before cancellation. Label unsupported transformations and visuals for reconstruction; label unreturnable usage/audit history explicitly. Record whether IDs and relationships survive or require crosswalks.
Freeze configuration and preserve snapshot plus final delta
- Freeze sources, schemas, transformations, semantic definitions, reports, filters, permissions, schedules, alerts, credentials and ownership.
- Record active refreshes, queued deliveries, errors, expiring credentials, open changes and incidents.
- Export configuration/model code before reports and data; preserve authoritative source snapshots at one UTC cutoff.
- Capture late source rows, corrections, refreshes, permission/owner changes and subscription events in an append-only final delta.
- Keep source read access until reports reconcile, parallel decisions end and rollback expires.
For every file/batch store source, query/filters, schema/version, UTC cutoff, rows, bytes, errors, exclusions and digest. Preserve raw exports unchanged and version transformations, crosswalks, rejects and manual repair.
Reconcile counts, values, permissions, rendering, and delivery
- Count variance = |target count − authoritative count| ÷ max(authoritative count, 1).
- Value variance = |target value − authoritative value| ÷ max(|authoritative value|, approved floor).
- Metric fidelity = accepted metric/cut combinations ÷ eligible combinations.
- Permission fidelity = user/role/test rows with expected allow/deny result ÷ eligible tests.
- Render fidelity = visuals whose labels, axes, units, filters, totals, nulls and interactions pass ÷ eligible visuals.
- Delivery fidelity = correct on-time subscription/alert deliveries ÷ eligible deliveries.
Test every critical metric across dates, stages, owners, teams, currencies, time zones, empty/small/large populations, late facts and corrections. Full-reconcile executives, finance-sensitive outputs and row-level access. Baseline source quality with the CRM data-quality guide.
Inject credential expiry, revoked field permission, API throttle, schema rename, missing partition, duplicate/partial batch, transform failure, semantic error, stale cache, daylight-saving change, currency gap, deleted user, group change and row-security leakage. Require alert, quarantine, bounded retry, backfill, correction, user notice and reconciliation.
Stage and run parallel reports with one decision authority
Build target connections, models, reports, groups, row policies, schedules and monitoring with external subscriptions/alerts disabled. Seed golden records and expected results. Microsoft documents Power BI refresh dependencies and schedules; use the actual target’s equivalent controls and failure evidence.
During parallel reporting, declare one authoritative output for each real decision. The target is validation-only until signed promotion. Label dashboards “source authoritative” or “target test,” suppress duplicate alerts/subscriptions, and log who consumed which version. Compare refresh completion, data timestamp, metrics, filters, permissions, rendering, exports and delivery.
Canary one team and bounded report pack only after hard gates pass. Collect consumer validation for workflow-specific features such as drill, export, accessibility and scheduled delivery—not opinions about colors alone.
Cut over and prove rollback without split-brain decisions
Cutover disables source subscriptions/alerts, captures final delta, completes target refresh, transfers decision authority, enables accepted delivery and monitors. Do not allow executives to act on two differently governed pipeline or forecast numbers.
Rollback disables target delivery first, preserves target-only incidents/corrections, restores source credentials/schedules/subscriptions, applies accepted source deltas, announces authority reversal, and prevents duplicate alerts. Stop for material metric variance, permission leak, stale critical data, missing report, wrong recipient, untraceable lineage, silent refresh failure or untested rollback.
Revoke connections, archive evidence, and request deletion
After acceptance and hold review, revoke source refresh credentials, gateways, service accounts, OAuth, API keys, webhooks, embeds, shares, subscriptions, SSO, users/admins and support access. Rotate reusable secrets and test access/refresh fails. Retain signed metric contracts, snapshots, reconciliation, lineage and audit evidence under limited custody.
Request deletion covering imported/cached data, extracts, semantic models, reports, exports, schedules, alerts, subscriptions, credentials, logs, support artifacts, subprocessors, backups, holds and exceptions. Cancellation is not deletion. Record retained scope, purpose, protection, expiry and completion evidence.
Apply the worked example and calculate TCO
This fictional example is not a platform result. Of 500 metric/cut cases, 490 pass; 1,000 permission cases yield 995 correct allow/deny results; 400 visuals yield 384 accepted renders; and 200 deliveries yield 196 correct on-time deliveries. One executive row-level leak remains a hard stop despite the aggregate permission result.
| Control | Calculation | Result |
|---|---|---|
| Metric fidelity | 490 ÷ 500 | 98% |
| Permission fidelity | 995 ÷ 1,000 | 99.5% |
| Render fidelity | 384 ÷ 400 | 96% |
| Delivery fidelity | 196 ÷ 200 | 98% |
TCO = source overlap/exit + target licenses/capacity + connectors/gateways + warehouse/pipeline + export/support + semantic/report reconstruction + security/privacy + testing/reconciliation + training/change + parallel run + monitoring/admin + rollback + archive + deletion. Use written quotes and measured labor.
Print the reporting-platform migration checklist
| Control | Evidence | Status |
|---|---|---|
| Estate, owners, cutoff, decision authority frozen | Charter | □ |
| Sources, models, metrics, reports, permissions, delivery inventoried | Graph | □ |
| Portability/reconstruction labels verified | Docs/tickets/samples | □ |
| Snapshot/final delta preserved | Manifests | □ |
| Counts, values, permissions, rendering, delivery reconcile | Signed report | □ |
| Parallel reporting has one decision authority | Consumer log | □ |
| Canary and rollback pass | Rehearsal | □ |
| Credentials revoked and deletion evidenced | Closure pack | □ |
Gangly is not a general BI, semantic-model or finance-ledger migration destination. Evaluate any Gangly reporting input with the same metric authority, permission, refresh, reconciliation, rollback and exit controls.