How the Phygii engine works: a technical view of structured property intelligence

Most organizations don’t have a data problem when it comes to their property portfolio, they have a structure problem.

Across Store Planning, Construction, Facilities, Brand, and Real Estate, data exists everywhere: site surveys, PDFs, CAD drawings, photos, spreadsheets, work orders, and assessments. But that data is typically captured for a specific project, used once, and then left disconnected from future decisions. That is what creates risk: Starting new initiatives without full context.

The Phygii platform solves that problem by standardizing how property data is captured, structured, and reused, and then applying machine learning and expert validation to generate actionable insights.

1. Data ingestion: capturing physical and digital inputs

Phygii begins with ingestion across both physical and digital sources.

Today, the primary inputs include:

• Mobile field capture (structured assessments, photos, 360 imagery)
• Reality capture (LiDAR-based scans and point clouds)
• Derived building data (floor plans, Revit models, CAD outputs)

Phygii doesn’t treat these inputs as isolated artifacts. Instead, the platform normalizes them into a consistent format so that downstream systems can interpret them reliably.

While integrations with external systems (e.g., CMMS, ERP, building management systems) are expanding, the current strength of the platform is in how it captures and standardizes field and spatial data with high fidelity.

2. The Phygital Twin: structuring data into a usable model

Once ingested, data is transformed into what we call a Phygital twin, a structured representation of each property that acts as a persistent source of truth.

This is not just a visualization layer. It is a normalized data model built around a clear hierarchy:

• Location (site-level context)
• Building
• Areas (e.g., sales floor, back-of-house, maintenance rooms)
• Assets (e.g., flooring, HVAC units, fixtures)
• Systems (e.g., mechanical, electrical, fire protection)

Phygii maps every data point into this structure, whether it originates from a survey response, image, or model.

For example, instead of recording “flooring = luxury vinyl” as a one-time answer, Phygii represents:

• A specific flooring asset
• Located in a defined area
• With attributes such as material, condition, and potentially manufacturer or lifecycle data

This structure is influenced by industry standards (including IFC) but is optimized for operational use, not just design or construction.

“The key outcome is consistency,” says Immersion Data Solutions Head of Platform Jared Juusola. “A door is always a door. A parking lot is always a parking lot. That allows data to be compared across time, across locations, and across programs without rework.”

3. Insight engine: calculating from structured data

With structured data in place, Phygii can generate what we call calculable insights.

These are deterministic, queryable outputs derived directly from the asset model. Examples include:

• Condition summaries by asset type (e.g., flooring across office areas)
• Age-based filtering (e.g., HVAC units older than 10 years)
• Scope identification (e.g., assets below a defined condition threshold)

“Today, clients generate most insights at the initiative level rather than across the entire portfolio in a fully unified way,” Juusola says. They are looking at a brand compliance program, a parking lot assessment, or a refresh initiative.

“Now, because all insights originate from the same structured model, they are inherently more consistent and traceable,” he says. “Every output can be tied back to a specific asset, location, and source data. This naturally extends clients’ ability to look across the whole portfolio to identify risk.”

4. Machine learning: targeted, high-value applications

Phygii applies machine learning selectively, where it can materially improve speed, scale, or consistency.

A primary use case today is asset identification and brand compliance.

For example:

• Identifying specific fixtures across thousands of stores
• Detecting brand inconsistencies (e.g., outdated signage, non-compliant layouts)
• Narrowing large image datasets to relevant subsets before human review

The workflow is intentionally hybrid:

1. The structured data model reduces noise (e.g., filtering by asset type or area)
2. Machine learning identifies likely matches or anomalies
3. Domain experts validate and refine the outputs

“This human-in-the-loop approach is critical,” according to Juusola. “It ensures that outputs are not just automated, but trusted. In many cases, the value comes from reducing the problem space by 80 to 90%, allowing experts to focus on validation rather than discovery.”

This is how processes that previously took months (such as large-scale brand compliance reviews) can be reduced to days.

5. Knowledge layer: connecting internal and external context

An important area of expansion is the knowledge layer, which connects internal asset data with external reference information.

This enables more complex questions, such as:

• Understanding lifecycle expectations based on asset type and age
• Incorporating external standards (e.g., ADA requirements, brand guidelines)
• Relating asset condition to performance or risk indicators

Today, this capability is emerging and applied selectively. Over time, it will expand the types of questions that can be answered without requiring manual data joins or custom analysis.

6. Orchestration and query handling

From a system perspective, Phygii uses API-driven orchestration to handle queries and data access.

Key components include:

• Query translation (mapping user questions to structured data queries)
• Context management (ensuring responses align with user role and intent)
• Controlled use of generative AI for interpreting and returning results

A key design principle is minimizing hallucination risk. Outputs are grounded in the underlying data model, and users can trace insights back to source data when needed.

7. Security and data governance

Phygii is designed with enterprise data controls in mind:

• Single sign-on (SSO)
• Role-based access control (RBAC)
• Logical data separation between clients
• Encryption and ongoing alignment with SOC 2 standards

Importantly, customer data is used to generate insights for that customer, not to train generalized models across clients.

8. Why structure comes before AI

The most important takeaway is this: Phygii does not rely on AI to create order from chaos. It creates order first.

By structuring property data into a consistent, reusable model, Phygii enables machine learning and analytics to operate on clean, contextualized inputs. That is what allows teams to:

• Reduce surprises during acquisition and construction
• Improve planning accuracy and reduce change orders
• Maintain brand and operational consistency at scale
• Move faster with greater confidence across the lifecycle

As integration points expand and portfolio-level analysis matures, that foundation will support increasingly advanced use cases.

But today, the value is already clear: validated truth, structured for reuse, and transformed into trusted intelligence.