delta-and-beta/analyst-business-intelligence

★ Insight ───────────────────────────────────── The writing-skills skill emphasizes CSO (Claude Search Optimization) — structuring skills so future Claude instances can find and correctly apply them. Key techniques: verb-first section names, symptom-based language, and a quick reference table that lets Claude scan before committing to a full read. ─────────────────────────────────────────────────

Here is the generated skill file:

---

Business Intelligence Reporting

Overview

This skill covers configuring, integrating, and optimizing BI tools and dashboards for marketing and product analytics. It applies a braze-analyst lens: every configuration decision is evaluated through dashboard design, KPI definition, data visualization clarity, and the reliability of automated reporting pipelines.

Use this skill when:

• Connecting a BI platform (Looker, ThoughtSpot, Mixpanel, Heap, etc.) to Braze data exports or Currents
• Defining KPIs and metric hierarchies for executive, operational, or marketing stakeholders
• Designing or reviewing dashboards for campaign performance, user behavior, or funnel analysis
• Building or troubleshooting automated reporting pipelines (scheduled reports, data refreshes, alerting)
• Choosing between BI tools based on use case (self-serve analytics vs. governed reporting vs. real-time observability)

Do not use this skill for raw Braze SDK instrumentation, data warehouse schema design, or general SQL authoring — those fall under data engineering or platform integration skills.

---

Topics Synthesized

This skill draws from the following reference areas:

| Topic | Focus | |-------|-------| | Looker BI Dashboard Integration | Primary governed BI layer; LookML modeling, Braze Blocks, scheduled deliveries | | ThoughtSpot BI / Analytics Integration | AI-powered search analytics; SpotIQ anomaly detection, live query mode | | Tellius BI Integration | Decision intelligence; natural language query, automated insight narratives | | Mozart Data BI Integration | Modern data stack consolidation; reverse ETL into BI layers | | Mixpanel BI / Product Analytics Integration | Event-based funnel analysis, retention cohorts, A/B test reporting | | Heap Digital Analytics Integration | Retroactive event capture, session replay, conversion path analysis | | Kubit BI Integration | Warehouse-native analytics; no-ETL reporting directly on data lake | | Clarisights Marketing BI | Performance marketing reporting; cross-channel attribution, spend efficiency | | Datadog Observability Integration | Infrastructure and pipeline monitoring; alerting on data delivery SLAs | | Business Intelligence Category Overview | Platform selection criteria, integration patterns, capability matrix |

---

Lens: How This Skill Approaches Problems

Dashboard Design

Dashboards serve specific audiences — executive, operational, or analytical. Prioritize:

• Signal-to-noise ratio: surface the 3–5 metrics that drive decisions, not every available field
• Audience-appropriate granularity: executives need trends and anomalies; operators need actionable drill-downs
• Consistent visual hierarchy: KPIs at top, supporting context below, raw tables last

KPI Definition

Before connecting any BI tool, establish metric definitions:

• North Star metric per reporting surface (e.g., 30-day retention for lifecycle, send-to-conversion rate for campaigns)
• Leading vs. lagging indicators: open rate is leading; revenue attribution is lagging
• Agreed denominator: unambiguous unique user/event counting prevents stakeholder disputes

Data Visualization

Match chart type to analytical question:

• Trend over time → line chart
• Part-to-whole → stacked bar or pie (only when ≤5 segments)
• Correlation → scatter plot
• Distribution → histogram or box plot
• Ranking → horizontal bar

Avoid decoration: 3D charts, excessive color, and dual Y-axes obscure rather than clarify.

Automated Reporting Pipelines

Reliable pipelines require:

• Idempotent refresh logic: re-running a pipeline must not duplicate data
• Alerting on data freshness: stale dashboards are worse than no dashboards
• Schema change tolerance: Braze Currents schema evolves — downstream models need versioned references
• Delivery confirmation: scheduled report delivery (email, Slack) should log success/failure

---

Quick Reference: Tool Selection

| Need | Recommended Tool | |------|-----------------| | Governed self-serve with semantic layer | Looker | | AI-powered ad hoc search analytics | ThoughtSpot | | NLP-driven insight generation | Tellius | | Cross-channel paid marketing attribution | Clarisights | | Product funnel and retention cohorts | Mixpanel | | Retroactive behavioral analytics | Heap | | Warehouse-native, no-ETL reporting | Kubit | | Data pipeline observability and alerting | Datadog | | Modern data stack consolidation layer | Mozart Data |

---

Common Mistakes

Connecting BI tools before defining KPIs. BI tools amplify whatever you measure. Undefined KPIs lead to dashboard sprawl and stakeholder confusion. Define the metric dictionary first.

Using real-time query mode for all dashboards. Live query (ThoughtSpot, Kubit) is powerful but expensive at scale. Reserve it for operational dashboards; pre-aggregate for executive summaries.

Ignoring Currents schema versioning. Braze Currents event schemas evolve. LookML views and Mixpanel event mappings built against an older schema silently break. Pin schema versions and monitor Braze changelog.

Building dashboards for data availability, not decisions. Build only dashboards with a named owner who acts on them. Unmaintained dashboards erode trust in the entire BI layer.

Skipping pipeline alerting. A dashboard that shows stale data without surfacing staleness is a liability. Every automated pipeline needs a freshness check and failure notification.