Powered by

# Transition Plan

UME sits on a wealth of data. Credit operations, customer interactions, payment reconciliations, operational metrics - all flowing through systems that power a growing business. But data without governance is a liability, not an asset.

Today, the same indicator can have over a hundred different versions scattered across dashboards. A single misconfigured query against log tables can cost tens of thousands of reais. Teams build reports by copying and tweaking existing ones, creating a maze where no one knows which number to trust. When leadership spots a discrepancy of a few cents - and they will - trust erodes.

This isn't a technology problem. It's a behavior and visibility problem that technology can help solve.

# The Real Cost of Ungoverned Data

The symptoms are out there:

  • Duplicated effort - Data scientists expand into data engineering just to get reliable inputs. The small DE team runs after bad queries instead of building strategic capabilities.
  • Inconsistent truth - The same KPI means different things to different teams. Reports built at different times can't be reproduced.
  • Security gaps - Metabase gives everyone access to everything. Granular permissions don't exist. External sharing is blocked by LGPD concerns.
  • Operational friction - FinOps sends reconciliation data to fund managers knowing some values are wrong. Atendimento can't track how indicators evolved over time.

These aren't hypotheticals. They're conversations happening right now.

# A Path Forward

The Architecture and Tools section describes what a governed data platform looks like - the layers, the tools, the design principles. This Transition Plan describes how we get there - the phases, the first projects, and the validation approach.

The approach is pragmatic:

  1. Start with research, end with production - We've spent time understanding the pain, mapping tools, and documenting architecture. This material is shared with stakeholders for alignment before implementation begins.

  2. Iterate in focused verticals - Rather than a big-bang transformation, we pick specific areas where data problems cause real pain, solve them end-to-end, and expand from there. What we build is production-grade from day one.

  3. Make governance visible and valuable - A data catalog becomes the central hub where everyone sees what exists, who owns it, and whether it's trustworthy. Good practices spread when they're easy to follow.

# What's Ahead

Start here, then follow the links in order - or jump to what matters most:

  • Rollout Structure - How we phase the work: the research and alignment phase we're completing now, the MVP implementation phase starting with two verticals, and the strategy for broader adoption.

  • First Projects - The two areas where we begin: Customer Support (Atendimento) and Financial Operations (FinOps). Both have clear pain points, engaged stakeholders, and the potential to demonstrate value quickly.

  • Tooling - What gets deployed for the MVPs and what comes later. We're not adopting tools speculatively - each choice is validated in a real vertical before broader rollout.

The goal isn't perfection. It's building a foundation where data is trustworthy, discoverable, and governed - and where the organization learns to expect nothing less.