A BIM-Based Integrated Model for Low-Cost Housing Mass Customization in Brazil: Real-Time Variability with Data Control ^†

Abstract

Addressing the growing demand for affordable housing requires innovative solutions that strike a balance between cost efficiency and user-specific needs. Mass customization (MC) presents a promising approach that enables the creation of tailored housing solutions on a scale. In this context, this study introduces a model for mass customization of affordable single-family housing units in the city of Teresina, PI, Brazil. Our approach integrates algorithmic–parametric modeling and BIM technologies, facilitating the flow of information and enabling informed decision-making throughout the design process. Since the early design stages, the work has assumed that these integrated technologies provide real-time control over design variables and associated construction data. To develop the model, the method proceeded through the following phases: (1) analysis of the context and definition of the design language; (2) definition of the design process; (3) definition of the cost calculation method and estimation of construction time; (4) definition of the computing model based on the specified technologies; and (5) quantitative and qualitative evaluation of the computational model. As a result, this research aims to contribute to the state-of-the-art by formalizing the knowledge generated through the systematic description of the processes involved in this workflow, with a special focus on the Brazilian context, where the issue of social housing is a critical challenge.

1. Introduction

Currently, the housing sector in the Architecture, Engineering, and Construction Industry (AEC) in Brazil addresses high housing demands through mass standardization, aiming to reduce production costs. While this is an efficient strategy for cost reduction, it overlooks the user’s needs. For example, Brazil’s social housing program, Minha Casa, Minha Vida (MCMV), ended in 2024 with 1.26 million contracted housing units [1]. However, the units delivered by the program, based on a mass production strategy, often fail to meet the needs of their residents and are frequently altered after occupancy [2].

On the other hand, mass customization (MC) is a strategy that enhances a product’s value by tailoring it to the user’s specific needs while maintaining production costs similar to conventional mass production. According to Piller [3], implementing mass customization requires flexible and robust operational and logistical processes, encompassing design, manufacturing, assembly, supply, and transportation. This requires leveraging contemporary design and manufacturing technologies that facilitate a seamless and automated flow of information.

Indeed, this is a pertinent strategy within the housing sector, as several studies indicate that alongside housing prices, matching the design with the user’s needs is considered one of the primary value criteria for prospective homebuyers [4]. However, the production process in the building sector is fragmented, operating under a linear, hierarchical model with centralized control. This fragmentation poses a challenge to implementing the strategy, which relies on a seamless and integrated supply chain.

As Kolarevic and Duarte [5] indicated, design technologies enabling the necessary variability of solutions to meet mass demands are readily available. Indeed, previous works have successfully implemented various levels of automation within the design process. Some important works in this direction include the discursive grammar of Duarte [6], the integrated system of Benros and Duarte [7], and the customization system for apartment plans by Veloso, Celani, and Scheeren [8]. However, the feasibility of MC depends on the precise information management associated with the range of solutions these design technologies can achieve. Hence, it is crucial to establish workflows that enable the ease of visualization and manipulation of complex building data. It is essential to ensure the feasibility of the solutions during the initial stages of the design process.

Building Information Modeling (BIM) is a comprehensive technology within the AEC sector that focuses on information management in the construction industry. It operates based on the paradigms of parametric modeling and interoperability, facilitating horizontal data exchange among all stakeholders involved in a project. This real-time exchange takes place within a centralized environment, promoting efficiency throughout the design, construction, and maintenance processes, ultimately aiming to enhance the lifecycle performance of buildings [9]. Therefore, BIM is particularly suited for integrating building MC models, as it facilitates dynamic information management and decision-making while enhancing communication and platform integration.

However, in terms of design automation, current BIM models lack the responsiveness needed for MC due to their limited algorithmic capabilities. Recent solutions have emerged that integrate algorithmic–parametric platforms with commercially available BIM software. These platforms expand BIM’s capabilities, enabling designers to utilize algorithms that efficiently execute complex design operations.

In this sense, this work presents a practical model for mass customizing affordable housing in the Brazilian context, integrating existing algorithmic–parametric modeling and BIM technologies. The model features a user-friendly visualization panel to help designers manage information from the early stages of the design process. Furthermore, it leverages the potential of BIM for integration with other models in an MC project context. This work aims to contribute to the state-of-the-art and support the adoption of MC by systematically outlining a workflow that relies exclusively on commercially available solutions.

1.1. Building Information Modeling and Algorithmic–Parametric Modeling

According to Andrade and Ruschel [9], BIM is founded on principles of parametric modeling and interoperability. One definition of parametric modeling is creating a geometric model by specifying fixed and variable parameters that influence changes within the model [9]. A parametric model enables associative design, meaning that relational and qualitative parameters propagate changes throughout the entire model based on changes in one parameter value [10].

A BIM model can generate associative solutions based on dimensional variations if the model maintains topological consistency. While this represents a significant advancement in construction information management, a mass customization strategy ideally involves more than just adjusting components or dimensions [10]. Thus, more complex design variations—such as space adjacencies, automatic positioning of building components, and topological changes—cannot be achieved due to the limitations of BIM platforms’ algorithmic capabilities, as noted in the work of Khalili-Araghi and Kolarevic, for example [11].

For instance, changing the window type and associated cost throughout the model is possible, thereby adjusting the overall building cost and other associated parameters. However, the designer must manually make more intricate modifications—such as adding extra rooms, creating an outdoor area between rooms, or swapping positions.

As Henriques [10] stated, an algorithmic modeling design process systematically uses logical operations to generate form and space based on architectonic rules, such as typologies and building codes, to codify the design intention in 3D modeling software using available programming languages. Thus, an algorithmic model relies on creating and manipulating an algorithm characterized by traceable architectonic relationships.

In a purely parametric model, manipulating object variables results in instances that lack the explicit requirement for algorithm manipulation and lack traceability of the relationships underlying the object’s construction.

The second BIM principle, interoperability, enables seamless data exchange between applications, eliminating duplication and streamlining information flow across disciplines. As Andrade and Ruschel [9] noted, a key advantage of BIM interoperability is the integration of information from various AEC stakeholders into a single parametric model. It enhances communication and enables diverse analyses throughout the design process, including construction costs, energy efficiency, structural performance, and multidisciplinary compatibility.

These two paradigms enable the definition of BIM as both a technology and a working methodology. From a technological standpoint, BIM harnesses and processes information from a database to facilitate project documentation and simulate building construction and operation activities. Methodologically, BIM embodies a collaborative, coordinated, and integrated work approach [9].

Therefore, this study assumes that the combination of BIM with algorithmic–parametric modeling technology can facilitate the MC of housing units by assuming two roles: (1) providing input data that feed the processing of a parametric algorithm and (2) serving as an algorithmically designed responsive central model, enabling simultaneous information sharing across disciplines.

1.2. Selected Approaches to BIM-Based Mass Customization Models

The first model, proposed by Khalili-Araghi and Kolarevic [11,12,13], employs the Autodesk Revit BIM application as both a design tool and a platform for user communication in a housing MC strategy. This approach revolves around manipulating dimensional parameters and constraints. Notably, this work explores the potential of parametric modeling, visualization, and documentation to establish a customization model that operates independently within a BIM environment without relying on external models. It highlights the model’s capability to introduce variability in solutions while ensuring the feasibility of construction, project regularization, and control of construction information. However, a notable drawback is its reliance on a specific application, which makes the model susceptible to its limitations, as well as design variability restricted to dimensional variations.

Hwang et al. (2023) [14] propose a client-engaged design framework for modular housing that leverages a BIM-informed digital platform to support mass customization. The platform allows clients to input their preferences through a web interface, using a Self-Organizing Map algorithm to generate and recommend compatible modular design samples. Grounded in DFMA principles (Design for Manufacture and Assembly), the system aims to formalize early-stage decision-making through data-driven customization while enhancing communication and reducing design iterations. This study does not use BIM as a comprehensive modeling environment but rather as a conceptual framework to structure and visualize modular combinations. While the work advances in developing a user interface that enhances user control and engagement in the design process, the platform cannot display information that is crucial to aid decisions in the early design phases, such as cost estimation or preliminary performance indicators.

Bianconi, Filippucci, and Buffi [15] introduced an MC model for single-family units in Italy, employing generative design techniques with performance-based criteria. The outcome is a web catalog showcasing a range of design solutions tailored to diverse environmental conditions. This catalog features an interface that enables users to explore and compare design options, considering factors such as construction cost, energy consumption, and thermal comfort. The authors primarily emphasize the role of BIM in facilitating interoperability and information exchange among industry stakeholders. They describe how to export the outcomes of the optimization process as IFC (Industry Foundation Class) elements to ensure compatibility with BIM, enabling seamless integration of the models generated during the research phase into design workflows and facilitating sharing with other professionals in the construction field. However, this study focuses only on the building’s envelope and its wooden structure without defining the interior layout. The interior space is simplified, lacking defined divisions or functionality, which limits the comprehensive development of a housing program tailored to the specific needs of each family.

Bakhshi et al. [16] developed a design model that utilizes BIM and algorithmic–parametric modeling via the Revit and Dynamo platforms. They aimed to incorporate the client’s involvement in customizing prefabricated buildings within a design process emphasizing manufacturing and assembly. The result was a design algorithm using the BIM platform for visualizing solutions and delivering information. The model automatically constrains customer choices during the building configuration process, guaranteeing fulfillment of their preferences while mitigating production, assembly, and rework challenges during the design and construction stages.

However, this research faces specific limitations that affect its applicability. One of them is its exclusive focus on high-performance prefabricated buildings, thereby restricting the model’s applicability to contexts where this building technology is suitable. Additionally, the lack of solutions for customizing the building layout limits a comprehensive approach to customization. Furthermore, the authors needed to develop a customized user interface to assess client collaboration during the building configuration process.

Finally, Zhang et al. [17] propose a model to enhance the client’s co-design experience and streamline production information flow for implementing mass customization in the housing industry through a prototype of an information model for manufacturing cabinet products. In this model, they integrated BIM to gather and store all required information and knowledge concerning product development. Subsequently, they generated automated manufacturing commands based on cabinet specifications and kitchen layout, facilitated by a Revit add-on programmed with pattern-cut optimization algorithms.

In the Brazilian context, Gazel et al. [18] developed a mass customization model in response to the environmental disaster in the city of Mariana, where the homeless population remains displaced. The modular housing model offers principles of variability, flexibility, and prefabrication for single-story and two-story housing, accommodating different land sizes. The authors propose using BIM as a future development to support detailed modeling, cost control, phase planning, and environmental comfort simulations. However, the article does not advance in that direction and focuses solely on the development of the design language and the parametric model.

Brasil and Franco [19] present a novel parametric wood frame construction system designed to support the MC of affordable housing in Brazil. The system is embedded within an integrated CAD/CAE/CAM (Computer-Aided Design, Engineering, and Manufacturing) workflow that automates design, structural analysis, and digital fabrication. However, the paper does not demonstrate how the system is applied within a specific housing typology in a particular context. Dalla Vecchia and Kolarevic [2] and Dalla Vecchia and Medvedovski [20] discuss the possibilities and limitations of applying mass customization in Brazil as a pre-occupancy and post-occupancy design strategy but do not propose a workflow or design language. Santana and Alves [21] propose a workflow for mass customization of window frames using interoperability between Rhinoceros 3D/Grasshopper/VisualARQ. Nevertheless, the process can be adapted for Revit or Archicad, which are widely adopted in the industry. This study does not present a new housing project but instead explores the exchange of window frames (and potentially any other elements) in social housing using the same workflow.

While mass customization offers a promising strategy for addressing Brazil’s social housing deficit, previous studies have not yet explored real-time variability with cost control in Brazil. In this context, the proposed model emphasizes the integration of BIM as an information technology to feed the algorithmic–parametric model with data. This integration aims to support the customized design process of low-cost single-family houses at dimensional, geometric, and topological levels. Additionally, the work considers the designer as the model’s primary user, tasked with mediating and communicating the residents’ requirements. Lastly, although not the primary focus of the work, the model showcases the potential of BIM as a methodology. It achieves this by integrating an interactive central model into the MC model, allowing designers to seamlessly incorporate it with other platforms.

1.3. Production Technologies for Mass Customization

The production process of a building involves fabrication, assembly, and occasionally subassembly of components, with different configurations depending on the technology, method, and location. The technology employed is the main factor influencing the production process of a building [22,23]. Thus, it has a significant influence on the definition of a design language for mass customization.

According to Piller (2019) [3], two factors increase the cost of producing customized goods: (1) increased complexity and (2) increased uncertainty of business operation. Regarding the first factor, increased variability necessitates a greater number of parts, processes, suppliers, and distribution channels.

In the second case, the increase in uncertainty arises from surprises stemming from the varying demands of the end-user, the point in the production chain at which they occur, and their impact on the cost of manufacturing and distribution [3]. As a solution, Piller (2019) [3] suggests that implementing computational technologies for production automation can increase the level of certainty and stability in production, as it enables a high level of variability with minimal human intervention.

For a better understanding of the role of production technology in the MC strategy, it is important to establish that there are three main building systems technology types: (1) artisanal production, (2) mechanical industrial prefabrication, and (3) digital industrial production. The first option offers high flexibility in production but comes with low stability and performance. It is not the most suitable option for MC [23], except in cases where cultural aspects favor its use, such as when there are traditions of community engagement and the utilization of local materials in the execution of the MC product [5].

The second technology is associated with a modular production strategy, which ensures flexibility and stability while also benefiting from the advantages of scale economy [5,24]. In industrial mechanical prefabrication, the correct synchrony of its processes depends on the coordination or compatibilization between design and the end-user’s insertion into the supply chain. Therefore, the design solution requires direct input data from the production process, which will generate design constraints. By manipulating the influence of the end-user on the supply chain, the designer can delay the end-user’s involvement to a later phase in industrialized production. In this case, the manufacturer can stabilize supplies and achieve savings by storing pre-assembled parts and later, based on specific demands, forming a customized product to be distributed to the end-user [25].

The third technology is controlled by digital information from a computational model, allowing high flexibility and efficiency. The manufacturer can produce objects from different techniques, classified as additive, subtractive, and conformation [26]. Therefore, design and manufacturing are interdependent, and the manufacturer cannot treat them as separate entities. Hence, the connection between the systems is so close that, eventually, one subsystem overlaps with the other [27]. Although digital production in the AEC industry has experienced significant advances recently, these technologies have not yet reached their full potential or are not yet economically viable [5].

Finally, the production technology employed should consider the post-occupancy process, when families expand or renovate their houses to suit their living conditions better [2]. After moving in, families often expand or modify their houses themselves, typically in small, incremental stages, using manual labor and locally purchased materials. This process reflects a high level of dependence on traditional construction techniques that are accessible and familiar to the residents. In this way, the designer of an MC design language should consider building systems that are easy to understand, flexible, and compatible with traditional practices [2].

2. Materials and Methods

To develop the model, this study proceeded through the following phases: (1) definition and analysis of the context in which the design took place, along with the specification of a design language appropriate to that context; (2) establishing an explicit logical design process; (3) determining the method for cost calculation and estimating construction time; (4) defining the computer model based on the specified technologies; and (5) conducting both quantitative and qualitative evaluations of the model. It is essential to note that the subsequent phases of the process were not strictly linear.

2.1. Context Definition and Design Definition—Residential Complex Parque Brasil in Teresina, PI, Brazil

Since one of our primary goals was to avoid a model offering generic solutions, the research adopted a real and specific context to extract the information for defining a suitable design language. For this purpose, the housing complex Residencial Parque Brasil in Teresina, located in the state of Piauí, was selected. The planning and implementation of this complex were part of the initiatives of the urban and environmental program Lagoas do Norte, led by the City Hall, aimed at the involuntary relocation of the families affected by the construction works or residing in areas at risk.

Due to its large-scale development, encompassing 1022 residential units, this case study supports the applicability of a mass customization strategy. Additionally, the area had a strong local construction culture shaped by longstanding clay extraction and processing activities. According to Oliveira and Lopes [28], the clay extraction, which has been prevalent in the region for over 50 years, served as the primary catalyst for settlement in the area.

Finally, the City of Teresina compiled a comprehensive report detailing the users and housing profiles in the locality [29]. This demographic data revealed a highly diverse composition of families in the housing complex, further justifying this case selection. In addition to single-family residential use, there are many cases of cohabitation or extended family, where different generations share the same dwelling or develop post-occupancy strategies to redefine specific degrees of privacy and individualization within the household. Furthermore, although the use is predominantly residential, as the affected local families are generally low-income, many residents adapt their living spaces to serve as workspaces for commercial activities, such as sewing, farming, and food services. Most existing housing units were self-built using ceramic bricks, roof tiles, and reinforced concrete.

In this sense, despite local government efforts to address varying needs, the housing units produced through standardized mass production did not fully reflect the diversity observed in terms of design and construction technologies.

In terms of the local climate, Teresina has a semi-humid tropical climate with high temperatures and two characteristic seasons: the rainy season (which occurs in summer and autumn, with an annual maximum temperature of 33.8 °C) and the dry season (which happens in winter and spring, with a minimum yearly temperature of 22.4 °C) [29]. Therefore, residents incorporate outdoor areas as an extension of their houses, whether for drying clothes or family gatherings. Shading devices, particularly through the roof design, are necessary in this context.

The preliminary research during this phase gathered information on the properties of the affected families, as well as local cultural, geographical factors, and building codes. As the focus was on the formal and spatial aspects of the housing units, the research adopts the existing land subdivision defined in the City Hall project, where most lots are 10 m × 25 m, located on predominantly flat terrain with minimal topographic variation.

2.2. Design Process Definition

The design logic follows a hierarchical and modular approach, where the designer defines spatial modules and assigns functions to spaces rather than relying on predefined rules tied to specific spatial functions. While this approach offers greater creative freedom, it also imposes qualitative challenges. However, by prioritizing flexibility, the model can accommodate a broader range of family configurations, including multi-family residences or home-based businesses.

2.3. Cost and Time Estimation

This research calculates the construction costs using two distinct methods. The first method, known as the CUB (“Custo Unitário Básico”), represents the Basic Unit Cost of Construction, which is the average cost per square meter to build a standard building, calculated monthly by state in Brazil for different building typologies. The second method relies on references from the National System of Research on Costs and Indices [30] in Brazil.

While CUB/m² is a commonly used benchmark, cost estimations derived from SINAPI references are more precise. Consequently, this work adopts the cost generated by SINAPI as the official value and the estimated cost with CUB/m² as the reference value to establish a maximum parameter.

The CUB/m² utilizes a standard reference value per square meter, calculated and published monthly by the state Construction Industry Unions following the criteria outlined by NBR 12.721. However, it does not account for the costs associated with foundations and building installations. To address this limitation, the work opted to augment the CUB/m² value with percentages that closely approximate the costs of these construction stages. Additionally, it modified the calculation of roof coverage to use the SINAPI cost instead of the Basic Unit Cost per square meter. This adjustment is necessary because the coverage area can significantly exceed the housing area, potentially distorting the final cost. Therefore,

Cost = ((cub × A) × (F1 + F2 + F3 − F4)) + C3

(1)

where A represents the built area, and Fn is the cost factor that adjusts the base cost based on specific construction components. Specifically, F1 accounts for the foundation cost factor (4.1); F2 accounts for the electrical installations cost factor (4.8); F3 accounts for the plumbing installations cost factor (12.7); and F4 is the cost factor for the roof coverage, deducted from the overall cost (15) [31]. Finally, C3 represents the cost of the roof calculated using the SINAPI method.

SINAPI references consist of reports detailing inputs (materials, labor, and equipment) and compositions for each federation unit.

Table 1. Method for calculating estimated costs utilizing values derived from SINAPI references. Source: authors.

Cost Code	Building Component	Cost Calculation
C1	Wall	Area × cost per m2
C2	Waterproofing of the walls	(area 2) × cost per m2
C3	Roofing	(area × slope factor) × cost per m2
C4	Cast-in-place concrete slab	slab volume × concreting volume cost
		slab volume × armor rate) × cost per kg of the frame
		slab cost = concrete + frame
C5	Prefabricated slab	area × cost per m2
C6	Superstructure	weight of beams × cost per kg
		weight of the columns × cost per kg
		superstructure cost = beams + columns
C7	Door	number of doors × unit cost
C8	Window	window span area × m2 cost
C9	Total cost	C1 + C2 + C3 + C4 + C6 + C7 + C8
C10	Infrastructure	total cost × f1
C11	Electrical installations	total cost × f2
C12	Hydraulic installations	total cost × f3
C13	Unit final cost	C9 + C10 + C11 + C12

illustrates the calculations derived from the SINAPI values and compositions. In cases where specific input values were unavailable in SINAPI, this work incorporates approximate values gathered from market research conducted in the city of Teresina, PI.

The analytical composition costs provided by SINAPI references helped to determine the estimated construction time. These compositions enable consideration of the hourly rates of the professionals involved and the associated costs per square meter of a specific construction component. For instance, constructing 1 square meter of vertical masonry using ceramic blocks measuring 9 cm × 19 cm × 39 cm requires 0.59 h of a mason’s time and 0.295 h of an assistant’s time.

Based on these values, the total time required (Tt) to construct the entire wall is defined by multiplying the area (A) by the time needed per square meter:

Tt = (A × t/m²)

(2)

In cases involving multiple professionals, such as a mason and an assistant, the model considers the highest time value as a reference. The values and units vary from volume to mass, depending on the construction component rather than the area. Subsequently, the model calculates the cost of the professional per square meter (c), which involves multiplying the professional’s hourly rate (p) by the total time required per square meter. Finally, the costs consider the sum of all professionals involved in executing the component:

c = (p1 × t1/m²) + (pn × tn/m²)

(3)

Hence, the model estimated the total construction cost (Tc) by the product of the professional’s rate and the total area of construction:

Tc = A × c

(4)

2.4. Computational Model Configuration

The model relies on commercially available tools for architectural and engineering designers. Using the Revit Inside Rhino plugin, the model integrates Grasshopper within Rhinoceros 3D 7 and Autodesk Revit 2023. This tool enables bidirectional data transfer between the two platforms, ensuring seamless workflow conversion between Revit’s data and Rhinoceros3D’s geometry and metadata [32]. This approach is due to its information transmission protocol and Grasshopper’s superior processing performance compared to Dynamo [33].

This workflow allows Grasshopper geometries to serve as references for positioning BIM objects in Revit. The resulting model comprises a central BIM model in Revit and a CAD model in Rhino 3D, utilizing the HUMAN UI plugin for Grasshopper to create a user-friendly interface for data visualization and parameter manipulation. Additionally, the BIM model is connected to Twinmotion’s virtual reality software via the Datasmith export plugin by Epic Games. Figure 1 illustrates the operational flow process behind the model and its main parameters.

In this diagram, the designer’s role is highlighted as a central mediator in the process, gathering residents’ demands and managing the algorithmic–parametric model, which exchanges data bidirectionally with the BIM environment. Blue arrows represent building component data generated in Autodesk Revit that feeds into the model, while yellow arrows indicate design parameters and rules embedded in the algorithm.

The columns on the right list the parameters defined in Rhinoceros/Grasshopper, including display settings, design inputs, and key constraint rules. The columns on the left represent the designer’s visualization resources: the parameters and data visualization panel (violet) and the 3D-rendered visualization software (pink). Below, the diagram shows how BIM component geometry, cost parameters, and 3D reference geometries are integrated to generate the 3D model.

3. Results

3.1. Building System Definition

The design was based on a modular grid to ensure regularity in component arrangement and installation. Additionally, the design language combines traditional building techniques with prefabricated elements to promote efficiency while respecting local construction practices and traditions (Figure 2).

Modularization and prefabrication, key production strategies for MC, enable flexible designs while maintaining production stability by incorporating the user’s needs in later stages of the supply chain.

The walls also consist of ceramic brick masonry due to its local availability and the regional construction tradition. This well-established building method promotes independence, allowing residents to undertake maintenance and modifications themselves or hire a local, affordable workforce. The external envelope is of load-bearing bricks.

The roof is suspended and composed of corrugated fiber cement panels installed on cold-formed metal profiles due to its lightweight nature and ease of installation and maintenance. The panel doors and windows are made of lightweight aluminum, occupying the full ceiling height of the building. The residents can easily remove and relocate them to different parts of the structure, allowing new configurations. Additionally, these panels eliminate the need for lintels. Complementary metal framing supports the roof in unbuilt areas, ensuring the roof’s stability and maintaining its rectangular shape consistency, regardless of the house’s volume. It also defines sheltered spaces that can be enclosed in the future if necessary.

The foundations are of the raft type, which is cost-effective and easy to execute. The roof slab is a precast slab system with lattice joists to support ceramic tile infill. This system is low-cost and quick to build.

3.2. Design Language

The house design consists of a modular grid of eight parametric modules. This grid defines the building footprint, with the recommended setbacks, as per local urban legislation. These setbacks include a minimum frontal (5 m), lateral (1.5 m), and rear (2.5 m) distance. The number of modules was defined based on subdivision tests that considered lot sizes (10 m × 25 m) and proportions and the ease with which users could manipulate the modules to create spatial configurations. The placement of the modules determines the massing and floor plan layout. The intersection of a central guide axis (y) and three perpendicular axes (x) establishes the origin points of the modules, arranged in pairs: AB, CD, EF, and EG. The longitudinal arrangement of the modules aims to adhere to the proportions of the existing lots, maximizing land usage and providing a clear indication of potential house expansion (Figure 3).

Defining the massing through modulation involves the following three key actions: (1) adjusting dimensional parameters based on multiples of the ceramic block size; (2) adding or removing modules to create open or sheltered areas that connect with internal spaces and allow for future modifications; and (3) managing interactions between adjacent modules to extend one space into another or fully integrate them. This interaction can also create additional access and circulation spaces between modules if required. It involves shifting the division axis between modules and increasing the minimum circulation space in one module while decreasing it in the other (Figure 4).

The design language considers a wet core by placing a bathroom within one of the modules selected by the designer. Thus, the positioning of the toilet establishes a hydraulic axis aligned with either the x or y axis adjacent to the vertex of the module containing the toilet. This placement determines and configures a wet core, ensuring that at least one side of the bathroom faces the exterior of the building. The hydraulic core may consist solely of the module housing the toilet or may include an adjacent module.

Dimensional guidelines are applied to ensure that the bathroom and other wet areas are neither too small nor too large compared to the module that contains them. These guidelines also maintain minimum circulation space requirements. Additionally, this core dictates the location of the house’s water tank, placed above the bathroom.

The roof will consistently maintain a rectangular shape, covering the entire built area of the structure, ensuring stability in construction regardless of the house’s layout. It extends over all active modules. This modular planning strategy optimizes land use by enabling a single, rectangular roof to span the dwelling, creating functional, partially covered open spaces in between. These intermediary areas reduce costs and offer flexibility, as they can be enclosed and repurposed by the user over time. While the basic roof shape remains stable, there are three variations based on the number of slopes and trim direction: a single-gable roof, a double-gable roof with edge trim, and a double-gable hipped roof with central trim. Doors and window frames are positioned at coordinates centered on the modular grid edges and at the ends, with a minimum distance determined by the designer. Frame dimensions are also standardized. Figure 5 illustrates the sequence of design decisions by the architectural language.

3.3. Computational Model

The model enables multiple house solutions designed by qualified professionals who interpret user needs and assess the quality of the results generated. The model allows the creation of rapid, informed housing solutions by automating the design process and providing real-time visualization of building data from the earliest stages. This approach enables designers to quickly define a house design that meets cost and schedule constraints while leveraging the potential of MC strategy.

The model operates by using Design Parameters, Building Component Data Parameters, and Model Display Parameters. The user manipulates the Design Parameters, which the algorithmic–parametric model then processes to generate the modular grid that allocates the building components. The BIM components specification feeds the model with building component data. By manipulating the Design Parameters, the user establishes the modular layout, the interaction between the interface edges of the modules, and the necessary subdivisions.

Hence, by simultaneously processing building data and design parameters, the model generates a 3D-informed model in Rhinoceros 3D and Revit viewports. This process is made possible through the bidirectional information protocol of the Rhino Inside Revit plugin. The data extracted from the BIM model, including construction time, hourly workforce costs, and material quantities, are used to calculate and display the unit cost and construction schedule. Additionally, although not explored in this work, the BIM model can be used for automatic drawing extraction and interoperability with other platforms, such as structural and energy evaluation software. Exporting the model in IFC (Industry Foundation Class) format would also allow interoperability and integration with other BIM software, facilitating integration with multidisciplinary workflows.

The user interface panel groups all the model’s parameters in one window, facilitating the design selection even for users unfamiliar with Grasshopper’s interface. Moreover, the panel displays cost and time information with user-friendly graphs. Thus, all essential building information is gathered in a single window to expedite decision-making while the designer tests design options, as shown in Figure 6c below. Regarding the 3D model visualization, the Model Display Parameters Group controls the model’s level of detail. It displays auxiliary information, such as site boundaries, setback distances, modular grid axes, and other relevant information. Additionally, the Datasmith plugin for Revit enables real-time renders using the Epic Games Twinmotion software.

3.4. Model Evaluation

The model’s results were evaluated using two distinct methods. Firstly, a quantitative assessment consists of calculating the combinatorial possibilities of all design parameters. Secondly, a qualitative evaluation by creating five different solutions to fulfill the same housing program within a specified cost range of BRL 75,000.00 to BRL 82,000.00 (Brazilian Reais) per dwelling, which is equivalent to a range of USD 13,700.00 to USD 14,950.00.

For the first method (

Table 2. Quantitative evaluation synthesis—combinatorial calculation of design parameters.

Aspect	Description
Method Type	Quantitative (Combinatorial Calculation)
Objective	To assess the model’s design flexibility and variability by calculating total possible combinations
Considered Parameters	- 14 parameters with 4 options each - 22 parameters with 2 options each - 2 parameters with 6 options each - 3 additional parameters (cardinalities: 8, 7, and 3)
Excluded Parameters	Dimensional variations (to avoid distorting results)
Equation	(4)^14 × (2)^22 × (6)^2 × 8 × 7 × 3
Total Combinations	6,809,442,636,584,189,952 (approximately 6.8 quintillion)
Key Insights	Demonstrates the model’s high flexibility and adaptability but lacks quality assurance and contextual filtering

), all design parameters are calculated combinatorially by multiplying their respective cardinalities (the number of elements in a set). For instance, the additional wet core can be in any module from A to H. Thus, the parameter ‘additional wet module’ has a cardinality of 8. We excluded the dimensional parameters, as they would distort the results. In this case, a one-centimeter difference in a wall within a house of the same design does not constitute a significant outcome.

The vast number of design variations (Table 2) demonstrates the model’s flexibility and adaptability (Figure 7), enabling it to accommodate diverse configurations and address various site conditions, user preferences, and functional requirements. The model’s comprehensive parameters enable the handling of complex design challenges and highlight the potential for MC, allowing users to tailor solutions efficiently. However, an ample solution space does not guarantee quality, as many configurations may need to be more functional or meet regulatory and design goals. Additionally, this method does not account for site-specific constraints or support iterative refinement, both of which are crucial for improving design quality. Therefore, while the model’s flexibility is evident, constraints and refinement are equally important for ensuring feasible, high-quality designs.

Additionally, since the design language lacks an objective quality control schema for the design solution, it was crucial to assess whether the model could fulfill its purpose, which is to enable designers to create fast, variable, and feasible solutions easily. For this, the researchers developed five different design solutions tailored to a fictitious young couple with two children, also incorporating a workspace for a home office with a separate entrance (

Table 3. Qualitative evaluation synthesis—design solution experimentation.

Aspect	Description
Method Type	Qualitative (Design Experimentation)
Objective	To evaluate if the model supports the creation of diverse, feasible, and fast design solutions
Scenario	A fictitious family (young couple with two children) requiring a workspace with a separate entrance
Design Constraints	- Area: within target range - Cost: from BRL 75,000.00 to BRL 82,000.00 - Time: 3 h total for five designs
Assessment Criteria	Feasibility, diversity, adaptability to program needs, real-time feedback on cost and area
Findings	- Model enabled fast, varied solutions - Real-time data helped balance design intent with practical limits

). Five design solutions were developed and documented within approximately three hours, adhering to the established cost and area ranges (Figure 8).

The model enables the quick identification of an adequate design solution based on cost and schedule, showing the real-time impact of changes. As a result, real-time data create a fast feedback loop between design ideas and practical outcomes. Consequently, designers and stakeholders can easily see the effects of their choices, making it easier to adjust and refine the design. This workflow streamlines the process, striking a balance between esthetics and practical needs.

Furthermore, considering detailed cost and time information from the start, the design can be adjusted dynamically to improve efficiency, especially in fast-track mass customization projects where timing and budget are critical. For example, even though the first and third designs have the same built area (Figure 8), there is an increase in the costs of the third one due to the rise in the use of the steel framing structure. It is essential to highlight that this qualitative approach is inherently subjective, relying on the interpretation of designers, stakeholders, or evaluators.

It is essential to note that these five examples are just a few of the numerous possibilities generated by the model. Labeling the examples in Figure 8 from left to right, solution 1 consists of a more compact walk-in configuration, offering the flexibility to convert the house from three bedrooms to two bedrooms and a workspace with a separate entrance. It also defines a more private, covered back area that can serve as a family gathering space, a laundry line for drying, or an extension of a workspace.

In solution 2, the recessed covered front creates a transition between public and private spaces. The workspace at the front allows easy client access and can also benefit from extending into the covered entry. The U-shape creates a flexible space that can serve multiple purposes, such as a kitchen extension, dining area, or even a laundry room. Other possibilities include adding a future bedroom directly connected to the central circulation.

Solution 3 separates the living and private spaces by creating a buffer zone with two independent outdoor spaces. These spaces can also serve as enclosed rooms in the future.

Solution 4 places the main entrance in the central void rather than on the main facade. It also creates an intermediate space that can serve as a multipurpose space.

Finally, solution 5, which is more streamlined, allows for future expansion by adding rooms along the central corridor.

The proximity of the bathroom and kitchen, as a design constraint, helps define a hydraulic wall to concentrate the plumbing system and the upper water tank. Models 2, 3, and 4 also allow flexible entrance designs, particularly for corner lots.

It is essential to note that this description outlines potential evaluation scenarios, but they do not constitute the primary focus of this study.

Possible quality criteria to expand on this study could include the following: (1) evaluation based on the residents’ demands (household size, age groups, usage demands, projected usage over time, etc.); (2) environmental performance criteria (thermal comfort, daylighting, ventilation, energy efficiency, water efficiency, etc.), based on lot orientations and placement (center of a block, corner, etc.). This model does not include integration with other simulation and optimization software and plugins, despite (3) the potential for accessibility, (4) affordability (based on time and cost), and (5) lifecycle (the ease of adaptability and resilience of the design in accommodating future changes or expansions). Although some criteria are objective and can be defined with automated solutions from simulation software, others are subjective and require the architect’s input based on the dialog with residents.

The architectural language bears a resemblance to the local construction pattern. Future research could develop other design languages to enhance the esthetic variability of the model.

However, it is essential to note that, according to the affordability criterion (cost) focus of this work, all examples presented here have estimated values below BRL 108,000.00 Brazilian Reais, the price of a standard unit, as per the Teresina City Hall [29], indicating that design variation is possible without compromising affordability.

4. Conclusions

This work contributes to the development of low-cost housing in the Brazilian context by demonstrating a comprehensive workflow for a mass customization strategy based on the integration of BIM and algorithmic–parametric models. It clarifies and formalizes the design and development process of a specific model in Teresina, PI, Brazil. This workflow can be generalized to other cases, contemplating the formalized processes: (1) developing the design using specific design technology and contextual information, including local demographic and geographic data, in addition to integrating design strategies associated with mass customization to specify building technologies and establish formal and functional relationships among potential solutions; (2) the model generation based on algorithmic–parametric modeling and BIM technologies; (3) calculating the cost and time associated with the design solutions; and (4) displaying the project information, such as cost and construction time, and the outcomes through a user interface to aid the design process and speed up the decision-making. It is essential to note that while SINAPI provides cost references specific to Brazil, the workflow is generalizable and can be adapted to other regions by substituting local cost data references.

This study demonstrates how combining BIM with algorithmic–parametric modeling enables designers to manipulate data directly embedded in building components. With the help of automated routines, it becomes possible to quickly generate design alternatives that respond to specific project needs while still maintaining control over key information such as cost and construction time. The core contribution here is how the workflow transforms BIM into a more interactive design tool, where real-time data feedback supports decision-making throughout the process rather than serving merely as a reporting tool of a single solution at the end.

It also highlights how this integration turns the BIM platform into a two-way system—capable of both supplying data to drive design routines and absorbing output data to build a central model. While this study primarily focuses on architectural workflows, the same logic also opens opportunities for linking with other disciplines, such as structural and building systems. This approach sets the foundation for a more connected and collaborative environment, where different teams can work with shared information and respond more effectively to project demands.

Regarding the work limitations, we emphasize the reliance on external commercial tools. Integration between Revit and Rhinoceros 3D relied on the Rhino Inside Revit plugin, which had limitations that might impede its real-world application. In addition, we only used Grasshopper’s acyclic components, which lacked support for recursive routines. Automating design processes with a more robust language and establishing an independent connection between technologies would yield better results. Additionally, juggling multiple different software in separate windows may occasionally lead to confusion in the user experience.

Furthermore, advancements in cost detailing are crucial, as this work does not consider kitchen finishes and fixed items. Additionally, the workflow does not include methods for qualitative evaluation of the generated solutions. Therefore, designers must mediate the code and assess the solutions analogically.

As future developments, we suggest creating an objective evaluation system to assess the results generated after defining specific functions, considering factors such as residents’ demands, building performance (thermal comfort, daylighting, ventilation, energy efficiency, water efficiency, building’s carbon footprint, etc.), accessibility, and lifecycle (the ease of adaptability and future changes or expansions).

We also propose developing algorithmic–parametric systems for complementary projects, such as structural design and electrical and plumbing installations, integrated into the central model to respond dynamically to architectural changes. Additionally, implementing structural sizing tools will enhance budget accuracy, further improving project precision. Furthermore, integrating performance assessment can help define objectives for optimizing and implementing generative design techniques, which could significantly enhance the design process.

Ultimately, there is a promising research opportunity to integrate this model with machine learning algorithms, enabling dynamic, data-driven prediction and analysis. Identifying variations within a single design language for low-cost housing presents a significant challenge in training machine learning models. In this context, the proposed computational model offers valuable potential for generating a dataset designed to train machine learning algorithms focused on automating design solutions based on specific user demands.

Other opportunities include incorporating the Grasshopper interface into a website, which could evolve into an online platform offering greater accessibility and user interaction. Finally, considering that mass customization involves not only the conception and design process but also the fabrication stages, future research could adapt this design language and algorithm to incorporate digital fabrication building technologies into the production process, such as 3D printing with concrete or clay. In this scenario, the current model can serve as the basis for an innovative, seamless MC workflow from design to fabrication, with minimal human intervention.