The Challenge
For four years, the company's internal teams tried and failed to calculate accurate Annual Recurring Revenue across their portfolio of Atlassian Marketplace products. The Atlassian Marketplace frequently changed data structures and obfuscated transaction details, making consistent ARR calculation a moving target. BI and Product teams maintained pricing and SKU lookup tables that had ballooned to over 25,000 entries, introducing errors with every manual adjustment. The legacy workflow relied on antiquated Airflow pipelines with complex Python scripts embedding decision-tree logic — constant maintenance just to keep running, with accuracy gaps persisting despite the effort.
The Constraint
The company needed to unify numerous product lines, each with unique subscription models, renewal patterns, and revenue recognition rules, into a single trusted source. The migration had to run alongside an active Salesforce CPQ rollout across multiple instances. Complex edge cases like early and late renewals, multi-year contracts, and non-standard revenue recognition meant that a generic migration playbook would not work. The team needed to deliver enterprise-grade accuracy without a six-month timeline.
The Approach
Mammoth Growth migrated the company to Snowflake using a three-layer data architecture (raw, standardized, reporting-ready) with helper models in dbt, then built a semantic layer via MetricFlow to centralize business logic once and distribute it everywhere. Streamlit apps let stakeholders validate calculations and surface edge cases in real time, compressing feedback cycles from weeks to days.
The Outcome
Less than 2% delta variance across the entire book of business — a milestone unreachable after four years of internal attempts. The solution automated processing for all 25,000+ SKUs, fully eliminating manual lookup table maintenance. Complex edge cases including overlapping dates, mid-cycle changes, refunds, and renegotiated pricing are handled automatically through recursive SQL and dbt lineages.
What This Unlocked
Durable ARR foundation. Withstands Atlassian Marketplace data changes and obfuscation without manual intervention.
Extensible medallion architecture. Future product additions and marketplace shifts absorbed without rebuilding the pipeline.
High-recency reporting unlocked. ARR calculations run frequently without process queueing or high compute costs.
Full rollout underway. Stakeholder approval completed. The solution is replacing the legacy Airflow process across all BI dashboards.
A replicable path forward. For companies managing multiple acquired products through marketplace channels, this engagement proves the route from fragile manual revenue processes to automated, trustworthy reporting leadership can rely on.
Services
Tech Stack
Results
ARR forecast variance
<2% delta across full book of business
SKU coverage
25,000+ automated (zero manual lookups)
Delivery timeline
8 weeks total (core in 5 weeks)
Legacy system replaced
Airflow + Python → dbt medallion architecture