A data warehouse is a centralized repository of structured, curated data optimized for analytical queries and reporting. It stores data in a predefined schema, applied before the data is written, so every record conforms to a known format. In commercial real estate it holds cleaned portfolio data such as rents, expenses, and occupancy that analysts query for underwriting and reporting.
How Does a Data Warehouse Work?
A data warehouse works by loading data through a pipeline that enforces a fixed schema on write, meaning each value must match a defined field and type before it lands. Rent, expense, and occupancy figures arrive already cleaned, mapped, and typed. The warehouse then stores them, often in columnar format, so analytical queries that aggregate across thousands of rows run fast.
Because structure is enforced up front, a warehouse trades flexibility for query performance and trust. It cannot easily hold a raw lease PDF or an unmapped export; those must be transformed first. The data warehousing market grew from $37.42 billion in 2025 to a projected $43.48 billion in 2026 at a 16.2% CAGR, per The Business Research Company, with the cloud segment growing faster, from $11.78 billion in 2025 toward a forecast $49.12 billion by 2031, per Mordor Intelligence.
Property | Data warehouse |
Data shape | Structured, curated |
Schema timing | Schema-on-write, enforced before load |
Primary use | Analytics, BI, reporting |
Query language | Mostly SQL |
Storage cost | Higher per unit than object storage |
Why a Data Warehouse Matters
A data warehouse matters because underwriting and asset management decisions depend on data that is consistent, current, and fast to query across a whole portfolio. When every property's rents and expenses conform to one schema, an analyst can compare assets, roll up portfolio NOI, and refresh a report without reconciling formats first. The warehouse becomes the trusted layer models read from.
The cost of not having one is duplicated, drifting spreadsheets. Warehouse query pricing is metered by data scanned, for example roughly $5 per terabyte processed on Azure Synapse and $6.25 per terabyte on BigQuery, per 2026 vendor pricing, so schema and columnar layout that reduce scanned data lower cost directly. A well-modeled warehouse is often the single source of truth a CRE team reports from.
Example
A data warehouse is clearest when three properties feed one queryable table. Each property's monthly figures load into a shared schema, and a portfolio query totals them instantly.
Property | annual_noi (loaded) | occupancy_pct (loaded) |
Property A | 1,200,000 | 94 |
Property B | 850,000 | 88 |
Property C | 640,000 | 91 |
Because all three conform to one schema, a single query totals portfolio NOI as 1,200,000 plus 850,000 plus 640,000, which equals $2,690,000. A weighted-by-NOI occupancy blends to roughly (1,200,000 x 94 plus 850,000 x 88 plus 640,000 x 91) divided by 2,690,000, which equals about 91.3%. Both numbers come from one query because the warehouse enforced a shared structure on write.
Variations and Edge Cases
Data warehouses differ by deployment and design. The cases below change cost, speed, or what the warehouse can hold.
Case | Behavior |
Cloud warehouse | Snowflake, BigQuery, or Synapse; compute scales on demand |
On-premise warehouse | Fixed hardware, higher upfront cost |
Data mart | A subject-specific slice of a warehouse |
Lakehouse | Adds warehouse features over a data lake's storage |
Schema drift | A new source field has no matching warehouse column |
Data Warehouse vs Data Lake
A data warehouse is often confused with a data lake, but they store data differently. A data warehouse holds structured, curated data under a schema enforced on write, so it is fast to query and trusted for reporting. A data lake holds raw data of any format at low cost and applies structure only on read.
The warehouse answers known questions fast; the lake keeps everything, including data no one has modeled yet. A CRE team might land raw lease PDFs and market feeds in a lake, then transform the subset it underwrites on into a warehouse. Warehouse storage costs more per unit than the object storage a lake uses, which is why raw, high-volume, or unstructured data usually stays in the lake.
Frequently Asked Questions
What is a data warehouse?A data warehouse is a centralized repository of structured, curated data optimized for analytical queries and reporting. It enforces a predefined schema before data is written, so every record conforms to a known format. In commercial real estate it holds cleaned portfolio data such as rents, expenses, and occupancy for underwriting and reporting.
What is the difference between a data warehouse and a data lake?A data warehouse holds structured, curated data under a schema enforced on write, making it fast to query and trusted for reporting. A data lake holds raw data of any format at low cost and applies structure only on read. The warehouse answers known questions fast; the lake keeps everything, including unmodeled data.
What does schema-on-write mean?Schema-on-write means data must conform to a predefined schema before it is loaded into the warehouse. Each value is checked against a defined field and type on the way in, so records are consistent and query-ready once stored. It trades ingestion flexibility for query speed and data trust.
Related Terms
Data Lake
ETL
Data Pipeline
Single Source of Truth
Data Normalization