# Mapping the real estate data stack from parcels to portfolios
Pick any office tower in a downtown core. Before a single lease is signed, that building already exists as dozens of disconnected data records: a parcel number at the county assessor, a set of comparable rents in a commercial database, a rent roll in the owner's accounting system, and a stream of temperature readings from sensors in the ceiling.
None of these systems talk to each other by default. The job of the real estate data professional is to stitch them into one coherent picture of the asset. Let's trace a single hypothetical building, call it 100 Market Street, through the full stack.
Everything starts with the parcel: the legally defined piece of land, identified by a unique APN (Assessor's Parcel Number, the ID your local government uses for tax purposes).
Public records give you the foundation:
This data is free but messy. Every county formats it differently. Some publish clean open data portals; others still require an in-person visit. The federal government's
For 100 Market Street, the parcel layer tells us: it sits on a 0.8 acre lot, is zoned for high-density commercial use, and last changed ownership three years ago. That is our anchor. Every other data source will link back to this parcel.
In database terms, the APN is your primary key: the stable identifier that lets you join messy sources together. Addresses are unreliable (100 Market St, 100 Market Street, Suite 400) but a parcel number is precise. Good real estate data teams normalize everything back to the APN.
Public records tell you what the government thinks. Comps (comparable properties) tell you what the market thinks.
This is where paid providers dominate. CoStar is the largest commercial real estate data provider in the US, offering lease comps, sale comps, and tenant information. Competitors include Reonomy, CompStak, and Moody's CRE (formerly Real Capital Analytics). These are subscription products, often expensive, and their coverage varies by market.
Comps give us:
For 100 Market Street, comps might show that similar Class A office space in the submarket asks a certain rent per square foot, and that the building last traded at a specific cap rate. Cap rate (capitalization rate) is annual net operating income divided by property value, the standard yield metric in commercial real estate.
A caution: comps are estimates assembled from broker reports, public filings, and proprietary research. Two providers can disagree on the same building's rent. Treat comps as a range, not a fact.
Now we go inside the building. The rent roll is the owner's master list of every lease: tenant name, suite, square footage, rent, lease start and end dates, and escalation clauses.
This is the most valuable and most private layer. It lives in property management and accounting systems like Yardi, MRI, or RealPage. Unlike public records, you only get it if you own or operate the asset (or are underwriting a purchase).
The rent roll answers the questions that actually drive value:
For 100 Market Street, the rent roll might reveal that a major tenant occupying 30 percent of the building has a lease expiring next year, at a rent well below current market. That single fact reframes the whole investment: is that upside (you can re-lease higher) or risk (the tenant might leave)?
The newest layer is the building itself, instrumented. IoT (Internet of Things) sensors now track occupancy, temperature, energy use, elevator traffic, and air quality in real time.
This matters for three reasons:
For 100 Market Street, sensor data might show that occupancy peaks midweek at a fraction of leased capacity, a pattern common in post-2020 office markets. That insight shapes both energy strategy and re-leasing assumptions.
🎬 [VIDEO: "How Smart Buildings Use IoT and Data" — youtube.com — a clear overview of building sensor systems and what the data enables for operators]
Here is the core challenge: four sources, four formats, four update frequencies. Public records change rarely. Comps update monthly. Rent rolls update as leases sign. Sensors stream continuously.
A coherent property data model joins them on the parcel key and layers time on top. Conceptually:
Property (APN: 12345-678)
├── Parcel [source: county, updated yearly]
│ └── zoning, lot size, assessed value
├── Market [source: CoStar, updated monthly]
│ └── submarket rent, cap rate, vacancy
├── Leases [source: Yardi, updated on signing]
│ └── tenant, sqft, rent, expiry, WALT
└── Operations [source: IoT, updated continuously]
└── occupancy, energy_kwh, temperatureThe value is not in any single layer. It emerges from the joins. Market rent alone is a number. Market rent compared to in-place rent from the rent roll, adjusted for a lease expiry you found in the rollover schedule, weighted against actual occupancy from sensors: that is an investment thesis.
Knowledge check
1. Why is the APN described as a 'primary key' in the real estate data stack?
2. A professional finds that a building's assessed value is significantly lower than the rents and sale comps suggest it should be worth. What is the most reasonable conclusion?
3. The excerpt emphasizes that public parcel data is 'free but messy.' What is the primary practical challenge this creates for a data professional?
4. Select ALL correct answers. Which of the following are typically part of the public records (parcel) layer of the real estate data stack?
Select all the correct answers.
5. Select ALL correct answers. Why does the excerpt treat the parcel as the 'anchor' for an asset's data, rather than the rent roll or sensor feed?
Select all the correct answers.
Everything above describes one asset. Owners hold portfolios: dozens or hundreds of buildings. The stack scales, but new problems appear.
Standardization becomes critical. If each building's rent roll uses different suite-naming conventions, you cannot roll them up. Portfolio teams enforce a common data dictionary (agreed definitions for every field) so that "occupied square feet" means the same thing across all assets.
Aggregation unlocks portfolio metrics: total NOI (Net Operating Income), blended WALT, geographic concentration, tenant concentration (how much rent comes from a single tenant across all buildings). A retailer that is a small tenant in ten buildings can become a hidden concentration risk.
Benchmarking lets you compare assets against each other and against the market. Which buildings underperform their submarket comps? Which have the highest energy cost per square foot? Data at portfolio scale turns individual quirks into patterns you can act on.
The direction of travel in 2026 is clear: firms are investing in centralized data platforms that pull all four layers into a single warehouse, so an analyst can ask one question and query across every building at once.