[ad_1]
Based on Tatango’s session at Coalesce 2025, the premier analytics engineering conference where Jean Paul Azzopardi, Data Analytics Engineer- Tatango, presented to conference attendees.
Tatango powers high-scale, permission-based messaging for nonprofits—helping teams connect with supporters, inspire action, and raise critical funds. To keep pace with new AI capabilities, we rebuilt our data foundation so AI and analytics work together in real time for fundraisers.
The Challenge: Outgrowing Legacy Systems
As message volumes and AI-driven personalization grew, our legacy Redshift environment became a bottleneck. Campaigns needed time-sensitive data processing and rapid iteration on models—yet queries slowed and developer velocity suffered.
The Solution: A Data Migration Built for AI
We moved to Snowflake for warehousing and standardized on dbt for orchestration—giving us elastic scale, versioned transformations, consistent tests, and documentation across 1,000+ models, tests, and macros.
- Dual environments: Redshift and Snowflake ran in parallel to ensure continuity during the transition.
- Validation & testing: We used Google’s open-source Data Validation Tool (DVT) to compare results across systems for trust and repeatability.
- Backfill & optimization: Historical data was rebuilt and tuned for Snowflake’s architecture; we also implemented macro-based dual execution to keep a single codebase compiling correctly for both platforms.
- Cutover & continuous improvement: We switched during a low-traffic window, then iterated for further gains.
Where AI Accelerated the Work
We used internal AI agents (e.g., KoreyAI for project flow; Devin for code assistance) to automate tedious refactors and safely remove legacy patterns—shortening the migration timeline by about a month.
The Impact for Nonprofits
- 68% faster queries—performance-critical pipelines now execute in seconds instead of minutes.
- 3× developer velocity—new segments and models deploy much faster.
- +23% engagement—AI-personalized messages show significantly better response rates.
These improvements make it easier to build programs that are timely, relevant, and truly donor-centric.
What This Enables Today
- Dynamic segmentation: Real-time grouping by engagement and giving patterns.
- Hyper-personalization: AI-generated message variations tailored to each supporter.
- Instant ROI analysis: Immediate campaign feedback to guide next steps.
- Predictive modeling: Identify who’s most likely to respond to specific appeals.
Explore how this foundation shows up in product:
Key Learnings
- Treat dbt like an application: CI/CD, tests, and validation are essential across environments.
- Lean on AI for migration speed: The right agent mix can accelerate refactors and reviews.
- Invest in documentation: Well-documented models moved far faster to migrate and validate.
Looking Ahead
We’re expanding beyond SMS/MMS to support richer messaging like RCS and WhatsApp, with deeper AI observability and governance for transparency and safety at scale.
Closing Thought
Great donor experiences start with great data. By rebuilding our foundation, we made AI more reliable, more contextual, and more useful—so nonprofit teams can do more with less, and supporters feel more seen in every message.
See it in action: Explore Tatango’s AI Smarter Texting, Power Segments, and Inbox.
[ad_2]
Source_link
