Flowmetric — SaaS Churn & Retention Analytics by Gabriel SouzaFlowmetric — SaaS Churn & Retention Analytics by Gabriel Souza

Flowmetric — SaaS Churn & Retention Analytics

Gabriel Souza

Completed work

Data Analyst

Microsoft Power BI

Python

SQL

Computer Software

Flowmetric — SaaS Subscription Health & Churn Analytics

Portfolio Project | Data Analytics | Python · SQL · Power BI

About This Project

This project simulates the work of a data analyst embedded at Flowmetric, a fictional B2B SaaS company selling project management, time tracking, and invoicing software to creative and consulting agencies across the United States, United Kingdom, Germany, and Portugal.

The goal was to build a complete end-to-end analytics case study — from raw data generation to executive dashboard — covering the four business questions any SaaS company needs to answer:

Revenue Health — Is MRR growing? What is driving the change?

Churn Analysis — Who churns, when, and in which segments?

Customer LTV — Which segments generate the most lifetime value?

Engagement & Early Warning — Can we predict churn before it happens?

The analysis is structured in three layers:

Python notebooks — data generation, cleaning, statistical analysis, and 12 KPI calculations

SQL (SQLite) — same KPIs replicated using advanced SQL techniques

Power BI dashboard — 4-page interactive executive dashboard

AI assistance (Claude by Anthropic) was used throughout this project as a development partner — for code generation, debugging, and methodology review. All analytical decisions, business interpretations, KPI definitions, and conclusions are the author's own.

Dataset

The dataset is synthetic, generated from scratch using Python with realistic B2B SaaS business logic calibrated to industry benchmarks published by ChartMogul and ProfitWell.

Table Rows Description customers 1,000 Client companies — country, industry, employee size band plans 3 Starter (€49/mo) · Pro (€199/mo) · Enterprise (€799/mo) subscriptions 1,094 Subscription history including upgrades, downgrades and churn events 200,322 Granular product usage log (login, time tracking, invoicing, etc.) payments 11,410 Monthly billing records including failed payments and retries

Data generation script: generate_data.py — fully reproducible with SEED = 42.

Churn rates calibrated to real SaaS benchmarks:

Starter: ~4.5%/month

Pro: ~1.6%/month

Enterprise: ~0.5%/month

Intentional data quality issues were injected to simulate real-world conditions:

Inconsistent country name formatting (casing, whitespace, abbreviations)

Missing industry values (~2% of customers)

Duplicate event rows (~1% — simulating tracking pipeline double-firing)

Negative subscription durations (33 records — day-of-month jitter)

All issues are identified, documented, and corrected in the EDA notebook before any analysis is performed.

Project Structure

Tools & Techniques

Python (EDA Notebook)

Libraries: pandas, numpy, matplotlib, seaborn, scipy

Data cleaning: string normalisation, duplicate removal, null handling, date correction

Analysis: 12 KPIs across 4 business blocks

Statistical validation: t-test for engagement signal (p=0.0002)

SQL (SQLite Notebook)

Common Table Expressions (CTEs)

Window Functions: SUM() OVER, RANK() OVER, LAG() OVER

Date arithmetic with JULIANDAY() and STRFTIME()

Multi-table JOINs (up to 4 tables)

UNION ALL for inline reference tables

NULLIF() for null handling in cohort analysis

Power BI (Executive Dashboard)

4-page interactive dashboard

DAX measures for KPI calculation

Conditional formatting on cohort retention matrix

Cross-page navigation with bookmark buttons

Key Findings

1. Revenue Health

MRR grew from €3,689 → €216,550 over 36 months (+5,770%)

Total cumulative revenue: €3.44M

NRR: 82% — below the 100% industry benchmark, meaning churn exceeds expansion revenue

Growth is acquisition-driven; if acquisition slows, the retention problem is immediately exposed

2. Churn Analysis

Plan Churn Rate Avg Days to Churn Starter 52.1% 201 days Pro 28.2% 216 days Enterprise 10.1% 274 days

208 active Starter customers currently show low engagement (at-risk cohort)

Geographic and industry churn differences are small (31–36%) — churn is a product problem, not a market problem

3. Customer LTV

Plan Avg LTV % of Total Revenue Enterprise €10,670 68% Pro €2,416 25% Starter €618 7%

Enterprise LTV is 17x higher than Starter

All plans have healthy LTV/CAC ratios (>3x)

Dev Shop / Software Consulting segment has the highest LTV (€4,190)

4. Engagement & Early Warning

Churned customers average 11.4% fewer product events in their first 30 days (statistically significant, p=0.0002)

Customers who never use time tracking or invoicing in the first 90 days are 1.40x more likely to churn (48.0% vs 34.3%)

Month 9 is the critical retention threshold — cohorts that survive past month 9 stabilise

Strongest cohort: 2024Q1 (59.6% retained at 24 months)

Recommendations

Build a 30-day onboarding flow specifically for Starter customers to accelerate time-to-value

Define deep feature activation (time_logged + invoice_sent within 90 days) as a north-star onboarding metric

Trigger customer success outreach when engagement score is below 14 events at day 14

Shift sales investment toward Enterprise — one Enterprise customer = 17 Starter customers in LTV

Investigate 2024Q2 cohort — weakest retention at 24 months (44.3% vs 59.6% for 2024Q1)

How to Run

Run in order:

Note: faker is required only to regenerate the dataset. If you only want to run the analysis notebooks, faker is not needed — the clean CSVs in data/clean/ are ready to use.

Author

Gabriel Souza — Data analyst Lisbon, Portugal · 2026 LinkedIn: https://www.linkedin.com/in/gabriel-souza-5bb6123a8/ GitHub: https://github.com/gabriel-souza-data

Like this project

Completed work

Posted Sep 8, 2026

End-to-end B2B SaaS case study: 12 KPIs across revenue health, churn, LTV, and engagement, with a 4-page Power BI dashboard and SQL analysis.

Likes

Views