Giga Innovate Case Study | Metric Harmony

Company Overview

Industry & Size:

A fast-moving SaaS scale-up (~200 employees)
FinTech focus, particularly real-estate portfolios (asset valuation, ROI, etc.)
Heavy on machine learning (ML) capabilities

Team Setup:

Data Analysts producing daily dashboards
Core Modeling Team governed by data engineers
Machine Learning Group building advanced features

The Challenge

Competing Definitions

The ML team built a "feature store" for advanced models while finance uses a different approach for "asset value," and analytics keeps producing new "AssetValue" permutations.
They ended up with 15+ variations of "Asset Value," none fully matching the CFO's perspective.

Speed vs. Control

Product managers demanded agility – they can't wait for a central team to define new features.
Data engineers worry about "model decay" if new metrics bypass their version-controlled pipelines.

Proliferating Dashboards

Real-estate funnel metrics were inconsistent; one dashboard calls a renovation "complete" once final inspection is done, while another does so at purchase.
Meetings often derailed into debates over "Which funnel figure is right?"

Why Giga Innovate Called Metric Harmony

They needed to bridge ML, finance, and product data. Giga Innovate heard that Metric Harmony specializes in human-centric alignment – especially where ML or advanced analytics teams clash with finance.

The CEO was tired of real-estate funnel confusion in board meetings, and the CFO and Head of ML wanted a pragmatic, short solution to unify "Asset Value" once and for all.

Engagement Scope

Top 5 Metrics to Align:

Asset Value
Renovation Completion
Annualized Return
Churn Rate
ML Feature Adoption Rate

Core vs. Experimental:

Identify which definitions must be fully governed in the "core model" and which remain experimental but clearly labeled.

Duration & Approach:

Interviews with the CFO, Head of ML, and product analysts, followed by a 2–3 week alignment sprint that culminates in a short document or wiki clarifying "official" vs. "experimental" metrics.

Process & Key Activities

1

Model Mapping & Discovery

Catalog the various "Asset Value" definitions along with other conflicting metrics by gathering 8–10 versions from dashboards, the ML feature store, and spreadsheets.
Ask each team: "Who uses this variant? For what decisions?"

2

Conflict Resolution Workshop

Conduct a 90-minute session with data engineers, finance, ML, and product management.
Use real examples (e.g., a renovation flagged as 80% complete in the ML model vs. 100% in the CFO's sheet) to highlight discrepancies.
Agree on a single "core" definition for "Asset Value" in official dashboards while allowing ML to extend it as "experimental" if needed.

3

Semantic Layer & Documentation

Build a short Notion "Data Model Wiki" documenting each newly aligned metric with its formula, data source references, and business logic.
Flag "experimental" metrics as "Not Governed" to indicate they are for ML use only and not for official board reporting.
Propose monthly check-ins where data engineers review any new or overlapping definitions.

Results & Insights

Clear Core vs. Experimental: The CFO's official numbers no longer clash with ML prototypes, and everyone can see which definitions are "certified" versus "in pilot."
Less Tension in Data Teams: Analysts and the ML group can experiment with new ideas, but label them "experimental" until validated.
Better Communication: There are no more Slack flurries debating which version to present—the Notion doc clearly tags each metric.

Reflection & Takeaways

Agile vs. Governed

They learned that you can maintain ML agility while having a core set of official metrics.

Short, People-Focused Approach

In under 3 weeks, they overcame a major data-model tug-of-war that was stalling decisions.

Sustainable Maintenance

Regular check-ins ensure that "experimental" metrics are monitored and only adopted if they prove valuable.

Case Study: Giga Innovate