Growth Insights Machine

Data Analyst take-home assignment (2023) for an email marketing SaaS company. The analysis explores growth patterns across two datasets: marketing site sessions and account signups, using sampled/anonymized data provided by the hiring team.

Preserved as a portfolio piece. Data is fake/sampled; company name anonymized. Not actively maintained.

What this demonstrates

• End-to-end analytical workflow: ingestion → EDA → feature engineering → insights → recommendations
• BigQuery + Python stack for analytics at scale (2.6M+ session records)
• Structured, self-documenting notebook with TOC, glossary, assumptions, and data quality log
• Business-first analytical framing: each insight is tied to a business question
• Proactive recommendations and open questions beyond the assignment brief

Analysis structure

Section	Contents
I. Environment	Package setup and dependencies
II. Pre-processing	Data ingestion, EDA, type analysis, feature engineering
III. Glossary	Dataset schema and field definitions
IV. Insights	Facts, trends, plan-level comparisons
V. Conclusions	Key findings, recommendations, open questions, assumptions

Key findings (summary)

• ~63% of accounts opted for the Free Trial Plan vs. Free Account Plan
• From July 2021, Free Trial sessions overtook Free Account sessions — consistent with a likely default plan switch at that date
• Conversion rate (sessions → signups) oscillates ~5%, trending lower post-2021
• The data points to potential plan cannibalization; several questions raised for stakeholders to validate

Stack

Layer	Tool
Data Warehouse	Google BigQuery
Analysis	Python (Pandas, statsmodels, plotnine, seaborn), Jupyter
IDE	GitHub Codespaces
Data transport	GCP service account (credentials gitignored — see .devcontainer/gcp-credentials.example.json)

Running locally

1. Create a GCP service account with BigQuery read access on your project
2. Export the key as JSON and save to .devcontainer/gcp-credentials.json (gitignored)
3. Update the project_id and GOOGLE_APPLICATION_CREDENTIALS path in the notebook
4. Install dependencies: pip install -r requirements.txt
5. Launch: jupyter notebook notebooks/insights_growth_accounts_sessions.ipynb

Known workarounds (documented in original submission)

• Dependencies installed directly in notebook cells (vs. a proper virtual environment manager)
• Large CSV (~750 MB) hosted on a GCS bucket due to GitHub/BigQuery upload limits