Analyst & Steward guide

This guide covers two personas whose work is closely linked: Marek Nowicki, the Business Analyst who produces and consumes data extracts without writing code, and Tomasz Wiśniewski, the Data Steward who governs and certifies that same data. Every journey is a numbered, step-by-step walkthrough mapped to concrete screens and routes.

Part 1 — Business Analyst journeys

Marek's defining constraint is simple: no coding required. The platform targets 30 minutes from his first login to a working pipeline, and his interface is a deliberately simplified subset of the engineer's.

His backend role is ANALYST (level 50); his Keycloak realm role is analyst; his UX persona is analyst. His allowed route prefixes are /, /my-pipelines, /data-browser, /marketplace, /monitor, /governance/quality, and /governance/lineage. He can view and run pipelines and use the AI Copilot, but he cannot create, edit, deploy, or delete pipelines through engineer-grade tooling — the AI builds them for him.

Journey 1 — 30-minute onboarding

The goal: from first login to a scheduled, monitored pipeline in under 30 minutes.

1. SSO login. Marek signs in. The platform greets him — "Welcome Marek!" — with his role shown as Analyst and his workspace set to the Analytics Team.
2. Guided tour. A 5-minute interactive tour walks him through the Design Studio.
3. AI-guided first pipeline. In the AI Chat he types a plain request: "I need daily sales from SAP HANA." The AI generates a 3-node visual pipeline — Source → Filter → Target — and Marek previews the first 100 rows at each node.
4. Schedule and monitor. He sets the schedule to "Daily at 8 AM". His dashboard now shows the pipeline's running status.

Screens: Keycloak login, onboarding tour, AI Copilot chat, Design Studio, schedule dialog, HomeDashboard (analyst variant).

Journey 2 — AI-guided self-service extract

This is Marek's everyday journey: turning an ad-hoc data request into a working extract through conversation.

1. Open the dashboard. Marek's Home Dashboard (analyst variant) shows "My Pipelines" — all 5 ran successfully — and a DataFreshnessCard confirming the data is fresh.
2. Ask in AI Chat. He clicks the sparkle trigger to open the AI Copilot, then the Chat tab, and types: "Export customer data from Teradata for Q1 2026."
3. Review the AI plan. The AI replies with a MiniPipelineDiagram — a lightweight SVG source → transform → target sketch — and a plan that creates a pipeline with an Excel output to GCS.
4. Confirm or adjust. Marek clicks "Yes, create". The pipeline is scaffolded in the Design Studio, where he fine-tunes it by dragging a filter node in the visual designer — no code.
5. Schedule. For a recurring report he sets a schedule; for a one-off he runs it immediately.

For a monthly report he chains this further: a scheduled extract from multiple sources, aggregated by region, product, and date, output to Snowflake to feed the BI dashboard.

Screens: HomeDashboard (analyst), MyPipelinesCard, DataFreshnessCard, AIChatQuickEntry, AI Copilot chat, MiniPipelineDiagram, Design Studio visual designer, ScheduledPipelinesTable.

Journey 3 — Self-service catalog exploration

Before building anything, Marek often needs to find the right data. The Data Browser is his catalog.

1. Open the Data Browser. He navigates to /data-browser and uses the large search bar.
2. Search. He searches "churn". Twelve related tables and views come back, ranked by relevance and popularity.
3. Browse a table. He opens a result to see its columns, statistics, and sample data, including column-level profiling with histograms.
4. Ask the AI. Unsure what a column means, he asks the AI Copilot "What does this column mean?" — the AI explains, e.g. "CHURN_SCORE is a 0–100 index calculated monthly...".
5. Build from the table. He clicks "Create pipeline from this table", which drops an instant pipeline stub into the Design Studio.

Screens: Data Browser landing, search results, Table Detail (Columns, Sample Data, Profiling tabs), column-level lineage, AI Copilot.

Part 2 — Data Steward journeys

Tomasz's role is investigative and gate-keeping. He ensures data quality and governance across Polkomtel's estate.

His backend role is STEWARD; his UX persona is steward. His allowed route prefixes are /, /monitor, /governance, /data-browser, /marketplace, /admin/audit-log, and /admin/access-reviews. The front-end persona model grants him rich governance, quality, and catalog permissions — review, approve, policy, lineage edit, quality create/edit, and catalog classify.

Journey 4 — Lineage investigation

A request arrives: "Where does CUSTOMER_CHURN_SCORE come from?"

1. Open the Lineage Explorer. Tomasz navigates to Governance Hub → Lineage Explorer (/governance/lineage).
2. Search. He searches "CHURN_SCORE" in the typeahead search bar.
3. Read the graph. An interactive D3.js graph renders the full path: SAP_HANA.CRM.CUSTOMER_ACTIVITY → pipeline wf_customer_360 (transform ML_SCORE_CALC, Python UDF churn_model.py) → Snowflake.DWH.DIM_CUSTOMER.CHURN_SCORE → Databricks.Analytics.churn_dashboard.
4. Adjust the level. He uses the Lineage Level toggle to switch between pipeline, table, and column granularity.
5. Inspect a node. Clicking a node opens the 320px detail panel — type, owner, quality score, tags, description, and upstream/downstream counts.
6. Run impact analysis. He clicks Show Impact Analysis to open a modal grouping downstream dependencies by severity (Critical / High / Medium / Low), exportable for a change-review.

Screens: LineageExplorerPage, LineageSearchBar, LineageLevelToggle, LineageGraph, LineageDetailPanel, ImpactAnalysisModal.

Journey 5 — Governance review and certification

Walkthrough of reviewing the pipeline wf_Subscriber_Churn_v2 before it is deployed.

1. Engineer submits. An engineer submits the pipeline for governance review before staging or production deployment.
2. Open the review queue. Tomasz opens the Governance Review Queue (/governance/reviews) and selects the pending pipeline, which opens Review Detail (/governance/reviews/:reviewId).
3. Read the automated checks. Governance is non-blocking — automated checks run in the background and produce a governance score out of 100. For this pipeline the checks return:
   • PASS: all 12 PII columns identified and tagged; data masking applied for the Analyst role; a 36-month retention policy; audit logging for PII access; complete source-to-target lineage; 12 quality rules defined; no cross-region transfer.
   • WARN: 2 columns lack business descriptions.
   • Score: 94 / 100 (the threshold is 90).
4. Review in detail. Tomasz inspects the diff view, the impact analysis, and the discussion thread.
5. Decide. He chooses Approve, Request Changes, or Reject with Reason. An approved pipeline may then be scheduled for production.

Screens: GovernanceReviewPage, ReviewQueueList / ReviewQueueItem, GovernanceReviewDetailPage (diff view, impact analysis, discussion thread, action bar).

Journey 6 — Quality monitoring

1. Open the Quality dashboard. Tomasz navigates to /governance/quality. The Overall Quality Score gauge shows, for example, 97.3% — up 0.2% week-over-week.
2. Drill into a domain. He sees four domain cards — CRM, Billing, Network, CDR. The CDR card has a red border and a pulsing dot: its score is 94.1%, down 1.2%.
3. Find the failing rule. Clicking the CDR card filters the Quality Rules grid. The rule IMEI_FORMAT_CHECK is failing at a 2.1% rate.
4. Read the AI insight. The AI Quality Insight card explains the cause: "IMEI format changed in new handset batch — update the regex pattern."

Screens: QualityMonitoringPage, QualityScoreGauge, DomainQualityCard, QualityTrendChart, QualityRulesGrid, AIQualityInsight.

Journey 7 — Glossary and audit

Business glossary

Tomasz maintains telecom-specific terms in the Business Glossary — for example, an ARPU rule validating values in the PLN 0.00–999.99 range. He searches and edits terms through GlossarySearchBar, GlossaryGrid, and GlossaryTermDetail.

Audit trail

1. He opens the Audit Trail (/governance/audit) to review governance-related activity.
2. He filters by category — for example, SECURITY.
3. He exports the filtered events as CSV.

Every state-changing operation is logged with the user ID, IP address, action, resource, timestamp, and a JSON detail blob. Audit logs are retained for 365 days and are exportable to a SIEM.

Compliance reports

Tomasz generates compliance reports — a GDPR Data Map, a PII Inventory, and an Access Audit — and exports them as PDF or CSV.

Screens: AuditEventRow, AuditFilterBar, compliance report generators.

Journey → screen cross-reference

Where to go next

• User journeys — the full journey index with ASCII maps.
• Data Engineer guide — the journeys for the persona who builds the pipelines.
• Administrator guide — platform operations and legacy migration.
• Personas & roles — how the Analyst and Steward personas map to backend roles.