H3 Hexagonal Analysis for Commercial Real Estate Revenue Optimization

H3 Hexagonal Analysis for Commercial Real Estate Revenue Optimization

Adverant Research March 2026

Executive Summary

Commercial real estate (CRE) revenue is fundamentally a spatial problem. Properties exist in specific locations. Tenants cluster by submarket. Lease expirations ripple outward across city blocks. Broker relationships map onto geographic territories. Yet the dominant tools used to manage CRE revenue --- generic CRMs, flat spreadsheets, zip-code territories --- strip away nearly all of this spatial structure before any analysis begins.

This paper argues for a geospatial-first approach to CRE revenue operations, built on Uber's open-source H3 hexagonal indexing framework. H3 provides a hierarchical, resolution-flexible grid that spans from continental market analysis (resolution 1, ~4.4 million km²) down to individual building footprints (resolution 15, ~0.9 m²), enabling consistent spatial aggregation across all scales that matter to a commercial real estate transaction.

We describe a proposed architecture --- the NexusROS platform --- that integrates H3 hexagonal indexing, Neo4j graph-based relationship mapping, DISC/Big Five psychological profiling, and GPU-accelerated lead scoring into a unified revenue operating system purpose-built for CRE. The system monitors geofenced trigger events (lease expirations, headcount changes, funding rounds), surfaces opportunity heatmaps at configurable H3 resolutions, and routes prospects to brokers based on relationship proximity within the broker-tenant-landlord network --- not arbitrary zip-code assignments.

We also confront the real constraints: CRE data quality varies sharply by market and asset class, broker adoption of analytics tooling has historically lagged other industries, and no geospatial system eliminates the relationship-driven nature of CRE transactions. These limitations are discussed honestly and directly.

1. Revenue Operations in Commercial Real Estate

1.1 Market Scale and Technology Investment

The global proptech market was valued at approximately USD 41.78 billion in 2024 and is projected to reach USD 140.67 billion by 2034, representing a compound annual growth rate of roughly 13.25%.[^1] Commercial real estate accounts for the majority of this spend: one analysis places CRE's share of proptech activity at approximately 56% of total market value in 2024.[^1]

Investment in CRE-specific technology is accelerating. Deloitte's 2025 Commercial Real Estate Outlook found that 81% of CRE executives planned to prioritize spending on data and technology in the coming year, driven largely by the maturation of generative AI tools.[^2] By 2025, over 70% of large U.S. commercial property portfolios had adopted some form of proptech asset-management platform.[^1]

Despite this investment momentum, adoption is uneven. The CRE brokerage community has historically moved more slowly than investors and operators. According to analysis cited by JLL, the percentage of CRE teams planning AI pilots grew from under 5% in 2023 to 92% in 2025 --- but most remain in the experimental phase, not production deployment.[^3] A key structural barrier: 54% of CRE technology leaders cite compatibility issues with legacy infrastructure as their top obstacle to adoption.[^3]

1.2 The CRM Problem in CRE Brokerage

The commercial real estate CRM software market was estimated at USD 2.5 billion in 2025, projected to reach USD 7 billion by 2033 at a 12% CAGR.[^4] Real estate leads all other industries in CRM purchase rates at 18% of the market.[^4] Yet despite this spend, fundamental productivity challenges persist.

CRM systems designed for general enterprise sales were not built with CRE-specific data structures in mind. As one industry analysis put it: "Standalone CRM applications lack the accompanying client and building data from property management databases that are essential for the CRE industry."[^5] Yardi identified this gap explicitly when it launched its CRM product in 2010, targeting "a lack of industry-specific CRM within the commercial real estate broker industry."[^5]

The consequence is data fragmentation at the point of revenue generation. Lease data lives in property management platforms (Yardi, MRI). Market comparables live in research databases (CoStar, Reonomy). Relationship history lives in a broker's inbox and personal contact list. Spatial context --- which submarkets are heating up, which buildings cluster together --- lives nowhere that automatically informs outreach priorities.

1.3 Lease Expiration as the Core Revenue Trigger

The primary revenue event in CRE brokerage is the lease expiration. Leases covering more than 265 million square feet of commercial space were set to expire in 2025 alone, comprising 100 million square feet of industrial space, 85.5 million square feet of office space, and 58.5 million square feet of retail space.[^6] Each expiration represents a re-leasing decision --- a decision the tenant must make, that landlords must prepare for, and that brokers on both sides must anticipate far enough in advance to be useful rather than reactive.

The challenge is timing. Commercial lease negotiation cycles range from three to nine months on average, with larger or more complex deals extending to a year or more.[^7] That means effective outreach must begin 12--18 months before a lease expires, not at the 90-day notice window. Systems that surface these opportunities only when a lease is already in renewal are structurally too late.

2. The Geospatial Intelligence Gap in CRE

2.1 Why Geography Defines CRE Revenue

In commercial real estate, location is not merely one variable among many --- it is the primary organizing principle of value. Submarkets command dramatically different rents, vacancy rates, and absorption speeds even within the same metro. A tenant's decision to lease a particular building is inseparable from its proximity to workforce pools, transportation nodes, supplier networks, and competitive peers. A broker's territory is, by definition, a spatial claim.

And yet the dominant data structures used to manage CRE revenue --- zip codes, county boundaries, MSA definitions --- are administrative artifacts, not market realities. A zip code boundary may bisect a contiguous office corridor. An MSA definition may lump together submarkets with entirely different demand profiles. These boundaries were drawn for postal delivery and census counting, not revenue optimization.

CBRE's data platform draws from 39 billion data points across more than 300 sources.[^3] JLL has built a large language model --- JLL GPT --- specifically for CRE market analysis.[^3] Platforms like ArcGIS, CARTO, and Reonomy have brought geospatial analytics to property research and site selection.[^8] But the translation from market intelligence to individual broker outreach --- who should call which tenant, when, based on what spatial intelligence --- remains largely manual.

2.2 The Limitation of Flat Territory Assignment

Most CRE brokerage operations divide territory by property type, geography, or some combination of the two. A broker "owns" office buildings in the Financial District; another covers industrial in the I-78 corridor. These assignments are typically drawn manually, negotiated through internal politics, and updated infrequently.

The structural problem with flat territory assignment is threefold:

Static boundaries in a dynamic market. Submarket boundaries shift as development cycles play out. A flat territory structure updated annually cannot track the emergence of new activity clusters or the decay of over-supplied corridors.

No relationship context. A broker may have deep relationships with tenants whose current space falls outside their formal territory. Rigid geographic assignment routes these opportunities to a different broker --- or to no one --- because the spatial boundary overrides the relationship reality.

No resolution hierarchy. The same CRE market needs to be analyzed at multiple spatial scales simultaneously: metro-level trends, submarket dynamics, block-level competitive analysis, individual building status. Zip codes provide only one resolution, and it is rarely the right one for any given question.

2.3 Current Tooling and Its Gaps

Esri's ArcGIS platform offers sophisticated GIS capabilities for CRE research and site selection.[^8] Reonomy (acquired by Altus Group in 2021) aggregates property data across 50 million U.S. properties with ownership intelligence useful for off-market deal sourcing.[^9] CoStar holds an estimated 70%+ market share among institutional users for market analytics and lease comps.[^9]

These tools excel at market research and due diligence. They are not designed to drive daily broker workflow --- prospect prioritization, outreach timing, relationship routing. The gap between sophisticated market intelligence and actionable, spatially-aware sales execution is where CRE revenue leakage most commonly occurs.

3. An H3 Hexagonal Framework for CRE Revenue

3.1 The H3 System

H3 is an open-source geospatial indexing system developed by Uber and released under the Apache 2.0 license. It partitions the Earth's surface into a hierarchical grid of hexagonal cells, supporting 16 resolution levels (0--15) from approximately 4.4 million km² at the coarsest level to approximately 0.9 m² at the finest.¹ Each finer resolution has cells with one-seventh the area of the coarser resolution, enabling consistent hierarchical aggregation.

Uber developed H3 to optimize dynamic pricing and driver dispatch across city-scale geographies.¹ The system has since been adopted across machine learning, disaster response, environmental monitoring, and geospatial analytics use cases.¹

Three properties of H3 make it particularly suited to CRE revenue applications:

Uniform adjacency. Every hexagonal cell has exactly six neighbors (except for twelve pentagonal cells used to resolve the topology of the sphere). This means spatial distance relationships are consistent across the grid --- an important property when building proximity-based scoring models or neighborhood-level clustering.

Multi-resolution consistency. A cell at resolution 6 (roughly 36 km²) is composed of exactly seven cells at resolution 7 (roughly 5 km²), which are composed of seven cells at resolution 8 (roughly 0.7 km²), and so on. This hierarchical structure allows analysis to zoom seamlessly from submarket to block to building without changing data models.

Index compactness. H3 cell identifiers are 64-bit integers, making them computationally efficient for database indexing, spatial joins, and real-time filtering across millions of property records.

3.2 NexusROS: Proposed CRE Architecture

The NexusROS platform proposes to integrate H3 spatial indexing with four additional technical layers to address the CRE revenue operations gap:

Layer 1: H3 Property and Tenant Indexing

Every property in the system is assigned H3 cell IDs at multiple resolutions (6 through 11 by default, configurable per deployment). This allows instant spatial queries at any granularity: "show all properties with lease expirations in the next 18 months within H3 resolution-7 cells adjacent to this target building" is a single indexed lookup rather than a radius calculation.

Tenant locations, when available, are indexed at resolution 9--10 (neighborhood to block scale), creating a spatial layer of tenant demand density that overlaps with the property supply layer.

Layer 2: Trigger Event Monitoring in Geofenced Zones

The system defines geofenced monitoring zones around properties, submarkets, or broker territories using H3 cell clusters. Within these zones, the system monitors for configurable trigger events:

• Lease expirations sourced from property management system integrations (Yardi, MRI) and lease abstraction services
• Headcount changes (hiring surges, layoffs, office location additions) sourced from employment data feeds
• Funding rounds and corporate expansions sourced from financial data providers
• Building permit activity indicating renovation or repositioning
• Ownership transfers indicating potential tenant disruption

When a trigger fires within a monitored H3 zone, it surfaces in the broker's priority queue ranked by predicted deal probability. Triggers are geofenced --- they only fire for brokers whose relationship graph includes nodes proximate to the triggering property.

Layer 3: Broker-Tenant-Landlord Relationship Graph (Neo4j)

The relationship layer represents the social and contractual network of CRE as a graph database. Nodes include brokers, tenants, landlords, properties, and companies. Edges represent relationship types: REPRESENTED_BY, PREVIOUSLY_LEASED, CURRENTLY_OCCUPIES, NEGOTIATED_WITH, OWNS, MANAGES, REFERRED_TO.

Neo4j's graph query language (Cypher) enables relationship-distance queries that are computationally expensive in relational databases: "find all tenants with leases expiring in 18 months who are within two relationship hops of Broker X through either direct prior representation or shared landlord relationships."

This graph layer addresses the flat-territory problem directly. Opportunity routing is informed by relationship proximity --- who in the broker's network knows this tenant, who previously represented this company, which landlord has the densest existing relationship with this tenant type --- rather than by which zip code the building happens to sit in.

Relationship graph data is spatially tagged with H3 cell IDs, enabling spatial-relationship queries: find tenants clustering in a specific H3 submarket zone who are already within the broker's relationship graph.

Layer 4: Psychological Profiling for Relationship-First Buyers

Commercial real estate is an extreme example of a relationship-driven transaction. Broker selection --- by both tenants and landlords --- is heavily influenced by trust, prior experience, and personal rapport. Academic and practitioner research consistently shows that CRE buyers (unlike, say, SaaS buyers) weight relationship and cultural fit highly relative to analytical proof of value.²

DISC personality profiling (Dominance, Influence, Steadiness, Conscientiousness) is one of the established frameworks for adapting sales communication to buyer psychology in complex B2B cycles.² A Dominant buyer (results-focused, time-compressed) needs a different engagement approach than a Steadiness buyer (trust-seeking, risk-averse). The difference is material in a transaction with a 6--18 month cycle and a 5--10 year contractual commitment downstream.

The NexusROS architecture proposes to infer DISC profile signals from available data --- communication patterns, engagement behavior, stated preferences in prior representations --- and surface these as context for broker outreach planning. Big Five personality dimensions and Cialdini's persuasion principles (social proof, authority, scarcity, reciprocity) provide additional behavioral context layers.

This is a probabilistic inference layer, not a psychological assessment. The profiles are signals to inform approach, not deterministic characterizations of individuals.

Layer 5: GPU-Accelerated Lead Scoring with Geospatial Weighting

The final layer scores prospects across all available signals --- behavioral, temporal, spatial, relational, firmographic --- into a single prioritized outreach queue. Machine learning-based lead scoring models have demonstrated the ability to increase marketing conversions by 9--20% and reduce churn identification time significantly when properly calibrated.³

The geospatial weighting component adds spatial proximity as a first-class scoring dimension. A prospect located within an H3 cell cluster experiencing high lease expiration concentration scores higher than an equivalent firmographic prospect in a low-activity zone. A trigger event within a broker's relationship-dense H3 zone scores higher than the same event in a zone where the broker has no relationship graph nodes.

GPU acceleration enables real-time re-scoring across large property and prospect datasets --- a requirement when trigger events can fire at any time and prioritization must update before the next broker planning session.

4. Projected Applications

4.1 Lease Expiration Pipeline

The most direct application is a rolling 24-month lease expiration pipeline, spatially organized by H3 submarket clusters and ranked by relationship proximity and predicted deal probability.

Each pipeline entry represents a specific lease, with the following context: property address and H3 cell membership, expiration date, tenant firmographic data, current rent relative to market, H3 submarket activity heat (based on recent comparable transactions), relationship distance from the viewing broker, and trigger events that have fired in the last 90 days.

This pipeline view does not eliminate the broker relationship requirement --- it directs limited broker attention to the highest-probability opportunities, with spatial and temporal context that makes initial outreach more informed.

4.2 Submarket Opportunity Heatmaps

At a portfolio or team level, H3 resolution-7 and resolution-8 heatmaps provide visual aggregation of opportunity concentration. Cells with high concentrations of near-term lease expirations, recent trigger events, and low current broker relationship density represent underserved opportunity zones --- areas where the team's relational coverage is thin relative to the deal activity present.

This inversion --- from "where are our relationships?" to "where are the opportunities we're not covering?" --- enables proactive relationship development rather than reactive response to known contacts.

4.3 Relationship Network Preservation

A structural challenge in CRE brokerage is relationship continuity through broker turnover. When a senior broker leaves a firm, their tenant and landlord relationships typically leave with them --- represented only in email archives and personal contact lists, not in any institutional data structure.

A graph-based relationship model with spatial tagging creates institutional memory of relationship networks that persists independently of individual brokers. When a broker departs, the graph retains: which tenants they represented, in which properties, at what H3 locations, with which landlords as counterparties, and through which referral chains those relationships were sourced.

This does not replace the relationship itself --- that transfers with the person. But it informs succession planning, client outreach priorities for remaining brokers, and the spatial coverage gaps created by the departure.

4.4 Proximity-Based Close Rate Analysis

Over a sufficient transaction history, the system can analyze whether spatial proximity between a broker's highest-relationship-density zones and prospect locations correlates with close rate. This analysis is not predictive in a causal sense --- many factors outside spatial proximity determine transaction outcomes --- but it can surface whether territory assignments are structured to match broker relationship strengths to spatial opportunity patterns, or whether there are systematic mismatches.

5. Implementation Considerations

5.1 Data Source Requirements

A functional H3-based CRE revenue system requires four categories of input data:

Lease and property data. Property addresses and H3 cell assignment can be computed programmatically from any geocoded address dataset. Lease expiration data requires integration with property management platforms (Yardi, MRI, CoStar) or lease abstraction services. Data quality varies: institutional portfolios with Yardi or MRI implementations typically have structured, accessible lease data. Smaller landlords and broker-held data are often in PDFs or spreadsheets, requiring extraction before integration.

Market transaction data. Comparable transaction data (rents, concessions, absorption) is primarily held by CoStar, which holds an estimated 70%+ of the institutional market for lease comps.⁴ CoStar data access requires subscription and API agreement. Alternative sources include Reonomy (ownership and off-market intelligence), local MLS datasets, and county recorder records for transaction prices.

Firmographic and trigger event data. Employment data, funding rounds, and corporate expansion signals are available through commercial data providers (LinkedIn, ZoomInfo, Cognism, Apollo). Data quality and coverage vary by company size --- large enterprises are well-covered, SMB data is patchier.

Relationship and interaction data. Initial population of the relationship graph requires CRM migration or manual data entry. Ongoing enrichment requires integration with email and calendar systems to capture interaction signals. This is the highest-friction data layer for initial deployment.

5.2 Integration with Existing CRE Platforms

The CRE technology stack at most institutional brokerage firms includes a combination of: Yardi or MRI for property management, CoStar for market data, a general-purpose or CRE-specific CRM (Salesforce, HubSpot, Apto, ClientLook), and productivity tools (Microsoft 365 or Google Workspace).

NexusROS is architected as a plugin layer --- it does not replace these systems but ingests their data through connectors and surfaces spatial and relational intelligence back to brokers through a dashboard or CRM integration. The 26-category connector framework includes CRM platforms, property management systems, employment data providers, and financial data feeds.

Data flow is bidirectional where possible: outreach activity recorded in NexusROS updates the relationship graph, which in turn updates lead scoring for future opportunities.

5.3 Broker Onboarding Approach

Broker adoption is the primary operational challenge for any CRE technology deployment. CRE brokers have historically been among the most resistant technology adopters in professional services --- the industry's traditional emphasis on personal relationships and relationship-protection culture creates active skepticism toward systems that appear to systematize those relationships.⁵

Successful deployment patterns in adjacent industries suggest several principles:

Show the pipeline first. The most immediately credible value proposition is the lease expiration pipeline --- a prioritized list of known upcoming opportunities that the broker did not previously have systematically organized. This is additive (it finds opportunities, not directive about how to work existing ones) and directly tied to commission.

Do not lead with the profiling layer. Psychological profiling and personality inference are valuable for experienced practitioners but create friction as a first-touch feature. Surface these as an optional depth layer after baseline adoption is established.

Integration reduces resistance. Brokers who receive spatially-ranked pipeline updates inside their existing CRM interface are more likely to adopt than those required to log into a separate system. API integration with Salesforce or HubSpot reduces the workflow friction substantially.

6. Limitations

6.1 CRE Data Availability Varies Sharply by Market

The analysis in this paper assumes reasonable availability of structured lease data, market comparables, and property records. This assumption holds reasonably well for major gateway markets (New York, Los Angeles, Chicago, Boston, San Francisco) with dense institutional ownership and established CoStar coverage. It holds less well in secondary and tertiary markets, where lease data is frequently held in paper files or informal formats, ownership is fragmented among smaller operators, and comparable transaction data is sparse.

An H3 heatmap built on incomplete data produces misleading spatial intelligence --- cells appear low-activity because the data is absent, not because deal activity is genuinely low. Any deployment must include explicit data quality indicators at the cell level, distinguishing between low-activity cells and no-data cells.

6.2 Relationship Graph Cold Start

The relationship graph is most valuable when it contains sufficient historical relationship data to generate meaningful relationship-distance scores. A new deployment starts with whatever can be migrated from existing CRMs and manually enriched --- typically a sparse, inconsistently structured dataset. The graph improves as the system captures ongoing interactions, but the period before sufficient density is accumulated represents a window where relationship routing adds limited value over conventional territory assignment.

Firms considering deployment should plan for a 6--12 month data enrichment period before relationship-based routing reaches operational reliability.

6.3 Broker Adoption Challenges Are Structural, Not Merely Cultural

The technology adoption challenge in CRE brokerage is not only cultural resistance to change. There are structural incentives that create friction: commission-based compensation creates strong disincentives to share relationship data that might route deals to a colleague; territory disputes make brokers reluctant to document relationship overlaps; and the personal-brand nature of senior broker practices means institutional data accumulation feels like losing a competitive moat.

These are organizational design problems that technology cannot solve. A system that surfaces institutional relationship intelligence is only as useful as the firm's willingness to act on it --- which requires compensation structures, territory policies, and data-sharing norms that most brokerage firms have not yet addressed.

6.4 Geospatial Intelligence Does Not Replace Judgment

H3 hexagonal scoring, trigger event monitoring, and relationship graph routing are tools for directing broker attention more effectively. They are not substitutes for the market knowledge, negotiating skill, and relationship depth that determine transaction outcomes. CRE deals are won and lost on factors that no scoring model reliably captures: the particular chemistry between a broker and a tenant's CFO, the landlord's undisclosed flexibility on concessions, the tenant's undisclosed preference to stay put despite the expiring lease.

The appropriate posture for this technology is augmentation of broker judgment, not replacement of it. Systems that present scores as directives rather than as prioritized inputs for human evaluation will encounter well-founded resistance from experienced practitioners.

6.5 Privacy and Data Ethics Considerations

Psychological profiling inferred from behavioral signals raises legitimate privacy considerations. DISC and Big Five inference from email patterns, meeting behavior, or engagement data is not equivalent to a consented personality assessment, and should be disclosed as such to both brokers and, where applicable, to the individuals being profiled. Regulatory frameworks including GDPR and CCPA impose constraints on behavioral inference from personal data that any commercial deployment must address explicitly.

7. Conclusion

Commercial real estate revenue operations sit at the intersection of geography, relationship networks, behavioral psychology, and temporal triggers. Current tooling addresses each of these dimensions in isolation --- market research platforms for geography, generic CRMs for relationships, deal management software for pipeline. None integrates them into a unified operational layer that can direct broker attention based on where spatial opportunity, relationship proximity, and trigger event timing converge.

The H3 hexagonal framework provides the spatial foundation for such integration: a resolution-flexible, computationally efficient indexing system that enables consistent aggregation from metro to building scale. Combined with Neo4j-based relationship graph modeling, behavioral profiling, and GPU-accelerated lead scoring with geospatial weighting, it represents a technically coherent architecture for CRE revenue operations that is genuinely spatial rather than administratively bounded.

The limitations are real. CRE data quality is uneven. Broker adoption faces structural as well as cultural headwinds. The relationship cold-start problem is non-trivial. And no scoring system replaces the judgment, credibility, and trust that close CRE transactions.

What this architecture offers is not certainty --- it is better-directed effort. In an industry where a senior broker manages hundreds of active relationships across dozens of potential opportunities at any given time, directing that attention toward the opportunities where spatial timing, relationship proximity, and trigger event convergence align is, at minimum, a higher-probability starting point than the alternatives currently available.

References

This paper was produced by Adverant Research. The NexusROS system described herein is a proposed architecture. All market statistics are sourced from third-party research as cited. No outcomes or customer results are claimed. All projections reflect forward-looking market forecasts published by the cited research organizations, not Adverant measurements.

© 2026 Adverant. Released under Creative Commons Attribution 4.0 International License.