Marketing Performance Analysis

End-to-end analytics stack to analyze marketing performance for the Maven Fuzzy Factory e-commerce dataset. The project ingests CSVs with DLT, transforms with dbt, orchestrates with Dagster, and surfaces KPIs consumable in Metabase.

Business Questions Covered

• Trends in sessions and orders
• Session-to-order conversion rate
• Performance by acquisition channel (utm_source)
• Revenue, gross margin, and refunds
• Revenue per order and per session

Architecture (High Level)

CSV (data/)
  → DLT ingestion → Snowflake (raw)
  → dbt (staging → intermediate → marts)
  → KPIs + tests + Elementary report
  → Metabase dashboard

Modern Data Stack Diagram

Stack

• Ingestion: dlt + polars
• Transformation: dbt-core + dbt-snowflake
• Orchestration: dagster + dagster-dbt
• Data Quality: dbt tests + elementary-data
• BI: Metabase
• Warehouse: Snowflake

Repository Structure

• data/ – raw CSVs + data dictionary (maven_fuzzy_factory_data_dictionary.csv)
• ingestion/ – DLT pipeline (CSV → Snowflake raw)
• dbt_marketing_perf/ – dbt project (staging, intermediate, marts, tests, exposures)
• infra/ – Dagster orchestration + ops scripts
• dashboard/ – Metabase dashboard export (PDF)
• docs/ – user guide, runbook, optimization notes

Data Model (dbt)

• Staging: source normalization (stg_*)
• Intermediate: business joins (int_*)
• Marts:
  • Core: fct_sessions, fct_orders, fct_order_items, fct_refunds, dim_products, dim_users
  • KPIs: kpi_daily_overview, kpi_daily_marketing_channels

KPI marts are incremental with a rolling 3‑day recompute window to keep data fresh.

Prerequisites

• Python 3.10+
• Snowflake access (account, warehouse, database)
• dbt and dagster installed via requirements.txt

Configuration

Snowflake environment variables used by dbt (see dbt_marketing_perf/profiles.yml):

• SNOWFLAKE_ACCOUNT
• SNOWFLAKE_USER
• SNOWFLAKE_PASSWORD
• SNOWFLAKE_ROLE
• SNOWFLAKE_DATABASE
• SNOWFLAKE_WAREHOUSE

Notes:

• DLT reads its configuration from ingestion/.dlt/ (or infra/.dlt/ when orchestrated).
• ingestion/filesystem_pipeline.py uses an absolute path to data/. Update DATA_PATH if the repo moves.

Quick Start (Local)

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# DLT ingestion
python ingestion/filesystem_pipeline.py

# dbt build + tests
cd dbt_marketing_perf
DBT_PROFILES_DIR=. dbt deps
DBT_PROFILES_DIR=. dbt build

# dbt docs
DBT_PROFILES_DIR=. dbt docs generate
DBT_PROFILES_DIR=. dbt docs serve

# Elementary report (HTML)
./infra/elementary/run_report.sh

Dagster Orchestration

• Ingestion job: dlt_pipeline_job
• Transformation job: dbt_build_job (auto-triggered after successful DLT run)
• Schedule: daily at 06:00 UTC

Start the Dagster UI:

cd infra
DAGSTER_HOME=$(pwd)/.dagster dagster dev -w dagster_project/workspace.yaml

Metabase Dashboard

• PDF export: dashboard/Metabase - MARKETING PERFORMANCE DASHBOARD.pdf
• dbt exposure: marketing_kpis_dashboard (default local URL in dbt_marketing_perf/models/exposures.yml)

Quality & Monitoring

• Constraints and relationship tests in schema.yml
• Freshness/anomaly monitoring via Elementary
• Dagster sensors to trigger dbt and alert on failures

Useful Docs

• docs/user_guide.md – daily workflow
• docs/ops_optimization.md – Snowflake optimization recommendations
• docs/kt_checklist.md – handover checklist

Recommended Run Order

1. DLT ingestion
2. dbt build + tests
3. Generate the Elementary report
4. Review the Metabase dashboard

---