Skip to content

Staging Models

Staging models are the clean interface between raw/corrected data and the analysis pipeline. They standardize names, preserve source grain, and add basic validation without introducing business scoring logic.

For a portfolio audience, the important point is simple: every analysis page starts from a stable set of source-facing models instead of repeatedly cleaning raw fields inside dashboard queries.

Layer purpose

The staging layer answers: can the rest of the project rely on consistent transaction, user, service, market, and external-context fields?

Nova Core Entities

Model Grain Pipeline role Main analyses supported
stg_interactions one row per transaction Standardizes transaction amount, status, timestamp, category, market, user, service, platform, lifecycle, and promo fields Executive Summary, Marketplace, Demand Drivers, Category Performance, Customer/RFM, Repeat & Risk
stg_users one row per user Standardizes membership, acquisition, lifecycle, device, market, and join-date fields Customer Analysis, RFM Analysis, Customer Repeat & Risk
stg_services one row per service Standardizes category, rating, market, service tier, and popularity fields Category Performance, Demand Drivers, service supply features
stg_markets one row per market Standardizes city market names, regions, coordinates, and market configuration fields Marketplace Analysis, Demand Drivers, market clustering

External Context Staging

Model Grain Pipeline role Main analyses supported
stg_geo_market_country_lookup one row per Nova market Bridges Nova city markets to country identifiers Marketplace Analysis, Demand Drivers
stg_geo_weather_daily one row per market per day Adds daily local weather features Demand Drivers, weather sensitivity
stg_geo_world_bank_indicators one row per country per indicator Provides macro indicators before pivoting Marketplace opportunity scoring
stg_geo_country_metadata one row per country Adds country labels, currency, region, and timezone context Marketplace Analysis and dashboard labeling

Business Relevance

The staging layer keeps business logic from being scattered across dashboards. For example, category names, transaction statuses, service tiers, and user lifecycle fields are standardized once, then reused throughout the project. That makes the analyses easier to trust and easier to explain.

Quality Controls

Staging tests cover the fundamentals: unique transaction, user, service, and market keys; accepted values for categories, statuses, user segments, and service tiers; required timestamps and dates; and relationships from transactions to users and services. These tests are the first quality gate before the data is aggregated into business metrics.