Jurnal Nasional Teknik Elektro dan Teknologi Informasi EN-87

Volume 14 Number 2 May 2025

© Jurnal Nasional Teknik Elektro dan Teknologi Informasi

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License
DOI: 10.22146/jnteti.v14i2.18985

Comparative Analysis of MVVM and MVP Patterns

Performance on Android Dashboard System
Fajar Pradana1, Raziqa Izza Langundi1, Djoko Pramono1, Nur Ida Iriani2
1
Information System Department, Faculty of Computer Science, Universitas Brawijaya, Malang, Jawa Timur 65145, Indonesia
2
Management Program Study, Faculty of Economics, Universitas Tribhuwana Tunggadewi, Malang, Jawa Timur 65144, Indonesia

[Received: 23 January 2025, Revised: 7 March 2025, Accepted: 16 April 2025]

Corresponding Author: Fajar Pradana (email: fajar.p@ub.ac.id)

ABSTRACT — The rapid growth of the Android market in various developing countries has driven the demand for higher-
quality applications. Developing Android-based applications presents specific challenges, such as the need for responsive
designs and optimization for devices with diverse specifications. Design patterns like model-view-controller (MVC), model-
view-presenter (MVP), and model-view-viewmodel (MVVM) have become popular approaches to address these issues.
However, studies on the performance of design patterns in Android applications, especially in modern programming
languages like Kotlin, remain limited. This research aims to compare the performance of the MVP and MVVM design
patterns in an Android-based boarding house management application, KosGX. This application utilized Kotlin and featured
an interactive dashboard requiring significant device resources. Testing was conducted by measuring performance across
three key aspects: central processing unit (CPU) usage, memory usage, and system response time. The results of the study
showed that MVVM outperformed in CPU efficiency, with an average usage of 8.92% compared to 11.15% for MVP. In
terms of memory usage, MVVM was also slightly more efficient, with an average usage of 121.48 MB compared to 121.55
MB for MVP. However, MVP excelled in response time, averaging 236.88 ms, whereas MVVM reached 252.68 ms. This
study underscores that the choice of design pattern affects application performance. MVVM is more efficient in CPU and
memory usage, while MVP offers better response times. These findings provide valuable insights for developers in selecting
the optimal design pattern based on the specific needs of their applications.

KEYWORDS — Kotlin, CPU Efficiency, Response Time, MVP, MVVM, Memory Usage, Android Profiling.

I. INTRODUCTION application development is a relatively newer field, with many

The Android market is currently experiencing positive developers still in the beginner phase [11]. Novice developers
growth [1], driven by increasing demand in developing markets often face challenges that can impact application quality, such
and technological innovations such as the integration of as inadequate architectural planning, excessive memory usage,
artificial intelligence (AI) in mobile devices [2]. According to poor feature implementation, significant bugs, and difficulty in
a report by the International Data Corporation (IDC), global making improvements. Clear guidance in the form of software
smartphone shipments were projected to grow by 6.2% year- design pattern implementation is therefore necessary to
over-year (YoY) in 2024, reaching 1.24 billion units. Android’s produce higher-quality software that is reusable, maintainable,
rapid growth of 7.6% YoY is particularly evident in the Asia- and easier to evolve [12], [13].
Pacific region (excluding Japan), Latin America, the Middle Design patterns offer solutions to common problems in
East, Africa, and China, primarily in the lower-tier device software development [14]. By applying design patterns,
segment. Conversely, iOS was only expected to grow by 0.4% developers can accelerate their workflow, improve code quality,
in 2024. and create systems that are easier to maintain. Currently, the
The Android platform is utilized not only by mobile device most widely used design patterns on the Android platform are
users but also by software developers and manufacturers across model-view-controller (MVC), model-view-presenter (MVP),
various types of devices, including smart TVs, tablets, and model-view-viewmodel (MVVM) [15]. MVC is the most
wearables, or automobiles [3]. Reflecting this diversity, the frequently adopted design pattern due to its simplicity in the
official Android app store hosts over 2 million applications development process, but it has a notable drawback in the form
across 60 different categories. These applications range from of tight coupling between the controller and the view. MVP and
education [4], banking [5], games, medical [6], travelling, and MVVM, on the other hand, provide different approaches to
health monitoring [7]. managing data and interactions between application
Android application development differs significantly from components. In MVP, the presenter acts as the intermediary
desktop or web development. Applications must be optimized between the view and the model, enabling modifications to
for a wide variety of devices and specifications compared to both components. In contrast, MVVM’s viewmodel provides a
desktop or web, which typically involve more stable stream of data that the view can consume, eliminating the need
environments with greater resources [8]. Additionally, Android for the view to update data as in MVP [16]. By separating these
applications require responsive designs to accommodate components, these patterns reduce potential issues and enhance
varying screen sizes and resolutions, unlike the relatively the testability of the application. Although MVP and MVVM
uniform resolutions and screen sizes in desktop and web are claimed to be superior to MVC, limited research has
development [9], [10]. While desktop and web application examined the performance differences between these two
development have been established for a longer time, Android patterns. Previous studies on mobile device performance

Fajar Pradana: Comparative Analysis of MVVM … p-ISSN 2301–4156 | e-ISSN 2460–5719

EN-88 Jurnal Nasional Teknik Elektro dan Teknologi Informasi
Volume 14 Number 2 May 2025

regarding design pattern usage in Android applications have II. RELATED WORKS
reported that MVVM offers better CPU utilization and faster
A. DESIGN PATTERN
response times compared to MVP [17]. However, MVP
A design pattern is a solution to common problems
performs better in memory management. These tests were
encountered during system development, particularly those
conducted on a point of sale (PoS) application developed in
related to design, code organization, and system efficiency [14].
Java, leaving a research gap for performance evaluations in
In mobile applications, system development also requires
more modern programming languages.
design patterns to ensure architectural structures are more
This study introduced a new approach by evaluating the
organized and that the function and purpose of every line of
performance of an application named KosGX, built using
code written by the developer can be easily identified.
Kotlin, through a comparative analysis of the MVP and
Officially, Android recommends that applications consist of
MVVM design patterns. Unlike previous studies that focused
two layers: the presentation layer and the data layer, with an
solely on Java programming language with limited use cases,
additional layer acting as an intermediary to facilitate
this study leveraged Kotlin—a modern programming language
interaction between these two layers [15]. Design patterns are
officially supported by Android—to address the lack of
essential in Android system development to create efficient,
research evaluating performance in newer programming
maintainable applications with high scalability [21]. Various
environments. KosGX, a boarding house management
types of design patterns can be applied in Android development,
application with complex interactive elements, provides a
each with its unique characteristics, advantages, and
comprehensive testing platform to assess design pattern
disadvantages. For instance, applying the flyweight design
performance in real-world scenarios. The application serves as
pattern in Android has proven to enhance awareness of memory
a general example of a dashboard-based system, presenting
consumption during mobile application development [22]. One
data akin to typical dashboard systems. Memory usage reflects
study compared the flyweight pattern with traditional object-
the amount of RAM allocated to the application, making it
oriented programming, showing that the flyweight pattern does
critical for low-end devices where excessive RAM allocation
not negatively impact memory usage, enabling professional
can degrade performance [18]. System response time (SRT) is
software design without sacrificing efficiency. Another study
a key factor influencing user satisfaction [19].
discussed the importance of improving software quality and
This study offers three main contributions: empirical
reusability in Android systems [23]. As a result, a paper
evidence, practical implications, and actionable insights.
proposed PatRoid, a framework for automatically detecting the
Empirical evidence provides empirical evidence on the
presence of design patterns in source code. Preliminary
performance trade-offs between MVP and MVVM in Kotlin-
evaluations demonstrated that PatRoid successfully detected 23
based Android applications, addressing gaps in previous
gang of four (GoF) design patterns in Android applications.
research on contemporary software development. Meanwhile,
practical implications highlight the significance of choosing B. MODEL VIEW VIEWMODEL (MVVM) dan MODEL VIEW
design patterns for resource-constrained software, ensuring PRESENTER (MVP)
more optimal memory usage and shorter response times. The MVVM and MVP architectural patterns are widely
Actionable insights offer developers to improve application used in software development, particularly in Android
quality and user experience through informed architectural applications. Both patterns aim to decouple concerns,
decisions. improving maintainability and testability, but they differ in
From a societal perspective, these findings benefit both their implementation approaches and data handling. MVVM is
developers and end-users. For developers, the study equips a variation of the MVC architecture designed to achieve a
them with deeper insights into design pattern performance, complete separation between the model and the view
enabling them to create more efficient and maintainable components [24]. The model is a class containing data, the view
applications. For end-users, especially those in developing represents the application’s user interface (UI) and is
regions who rely on low-end Android devices, optimized responsible for displaying information, and the viewmodel
applications contribute to smoother user experiences and handles the application’s business logic, providing data streams
longer device lifespans. Furthermore, as digital solutions to the view component without being directly tied to it. In other
continue to address critical societal needs such as education, words, the viewmodel has no knowledge of the existence of a
View.
healthcare, and financial inclusion, developing high-
In contrast, MVP is an architecture similar to MVC but with
performance applications becomes a cornerstone for
some differences. The workflow begins with the view
technology-driven progress.
capturing user input, which is then passed to the presenter [25].
This research aims to identify the design pattern that The model contains the data to be displayed, the view
offered better data presentation and optimal resource utilization represents the application’s interface, and the presenter
for the KosGX application. Significant performance manages all interactions between the model and the view [26].
differences indicate that the choice of design pattern is an The presenter retrieves data from the model and delivers it to
essential factor in performance testing. Design pattern selection the view. MVP leverages interface classes, which are empty
impacts device performance when running applications, function definitions that can be extended by other classes.
making it essential for developers to choose the most suitable These interfaces are used in both the presenter and view
pattern to ensure a positive user experience. Application components to implement functions as required, whether to
performance plays a vital role in user satisfaction, and poor retrieve or send data [27]. Figure 1 illustrates the workflow of
performance can adversely affect users’ perceptions of the the MVP and MVVM architectures.
application [20]. Consequently, developers must consider MVVM excels in terms of modifiability, featuring the
system performance as an integral aspect of user experience lowest modification index, while MVP performs better in terms
design. of raw performance. Both architectures are part of the clean

p-ISSN 2301–4156 | e-ISSN 2460–5719 Fajar Pradana: Comparative Analysis of MVVM …

Jurnal Nasional Teknik Elektro dan Teknologi Informasi EN-89
Volume 14 Number 2 May 2025

commonly used to develop new products or systems through

research, design, development, and evaluation processes [31].
Research methodology was designed to support a comparative
performance analysis of the MVVM and MVP design patterns
in an Android dashboard system. The study began with the
software development phase, where an application
implementing both MVVM and MVP design patterns was
developed using the Kotlin programming language. Once the
application was completed, the designing test case phase was
conducted to create testing scenarios covering various
(a)
performance parameters, such as CPU usage, memory usage,
and response time. The subsequent phase, experiment, involved
executing the test scenarios on the application to collect
performance data. The data obtained from the experiments
were gathered during the data collection phase to ensure their
completeness and validity. Subsequently, the data were
analyzed during the data analysis phase using relevant
(b)
statistical methods, such as the independent sample t-test, in
Figure 1. Illustration of how pattern architecture works, (a) MVP and (b) MVVM. order to identify significant differences between the two design
patterns.
architecture paradigm, which enhances Android application
development by ensuring minimal dependencies and improved A. SOFTWARE DEVELOPMENT
usability [26]. The first step in this research was the development of a
A study has reported that MVVM is the best in dashboard-based application named KosGX, built using Kotlin,
modifiability, boasting the lowest modification index, while a modern programming language officially supported by
MVP outperforms in performance and model-view-intent Google for Android development. KosGX is a boarding house
(MVI) excels in test coverage [28]. Due to the limited research management application with a dashboard feature that includes
exploring the comparative analysis of MVP and MVVM in various visual components such as text, images, graphs, and
native Android platforms, particularly in systems developed interactive elements requiring significant CPU and memory
using Kotlin, this study aimed to address the existing gap. resources.
The application was developed using MVP and MVVM
C. SYSTEM PERFORMANCE design patterns. The application interfaces were made identical,
Modern smartphone users are no longer limited to displaying data such as payment visual diagrams, the number
preinstalled applications provided by device manufacturers. of residents, rooms, a list of residents, a list of needs for the
They can also access third-party applications downloaded boarding house, and a summary of payment income. The data
through various app distribution platforms [21]. In this were stored in a local database using Room, a library built on
ecosystem, the use of third-party applications often leads to SQLite for local data storage on Android. Figure 2 illustrates
significant memory consumption, which can result in the UI of the KosGX application, which serves as the
insufficient available memory to run these applications experimental object in the performance comparison study of
optimally. This issue is particularly common in lower-tier the MVVM and MVP design patterns on Android applications.
mobile devices with limited memory capacity, making memory The application is designed to assist in managing boarding
shortages a more frequent problem [29]. houses with key features such as monitoring payment status,
Application performance reflects how quickly an tracking room availability, revenue management, and a to-do
application runs, how fast it loads data, and its overall list.
connectivity with various operations. Several key aspects need On the left side of the interface, the main dashboard
to be considered to evaluate the performance of Android displays a summary of important data, including payment
applications. These include the central processing unit (CPU) status represented as a pie chart, room availability with total
usage, execution time, and memory consumption, which are the capacity, total revenue, and a high-priority task list for facility
primary metrics for measuring application performance [30]. maintenance. Meanwhile, on the right side, a detailed view of
The demand for high-performance applications on the Android room occupants is shown, featuring information such as
platform continues to grow with advances in technology and occupant names, professions, payment amounts, and payment
rising user expectations. Thus, this study aimed to compare statuses (paid or unpaid), marked with visual icons.
design patterns to identify the one with the best performance on This application was developed to support data collection
the Android platform. related to application performance in real-world usage
In this research, Android Profiler was utilized to measure scenarios. The UI components are designed to accommodate
performance. During the testing process, the device was the needs of boarding house managers, focusing on efficiency,
connected to Android Profiler to record CPU and memory clear information presentation, and simple interactions. In the
usage while opening the dashboard. For response time, logs experiment, this application is used to compare CPU
from Android Studio were analyzed to determine the time taken performance, memory usage, and response time between the
by the application to display data. implementations of the MVVM and MVP design patterns.
III. METHODOLOGY B. DESIGNING TEST CASE
This study employed a methodology inspired by the The test scenarios in this research focus on the application’s
research and development (R&D) methodology, which is performance, which includes evaluating responsiveness,

Fajar Pradana: Comparative Analysis of MVVM … p-ISSN 2301–4156 | e-ISSN 2460–5719

EN-90 Jurnal Nasional Teknik Elektro dan Teknologi Informasi
Volume 14 Number 2 May 2025

TABLE I
TEST CASE MVVM

Response
Data Range CPU Memory
Time
TC-VM- TC-VM- TC-VM-
10,000– 19,000
CPU-01 MEM-01 RT-01
TC-VM- TC-VM- TC-VM-
20,000– 29,000
CPU-02 MEM-02 RT-02
TC-VM- TC-VM- TC-VM-
30,000– 39,000
CPU-03 MEM-03 RT-03
TC-VM- TC-VM- TC-VM-
40,000– 49,000
CPU-04 MEM-04 RT-04
TABLE II
Figure 2. Example of KosGX application UI. TEST CASE MVP
.
Response
scalability, stability, and resource usage [32]. Dummy data, or Data Range CPU Memory
Time
synthetic data, are displayed on the application’s dashboard for TC-P-MEM- TC-P-RT-
10,000– 19,000 TC-P-CPU-01
testing. Each test case is identified using a specific code format: 01 01
“TC-X-Y-Z,” where X represents the design pattern, Y 20,000– 29,000 TC-P-CPU-02
TC-P-MEM- TC-P-RT-
represents the test type, and Z indicates the test case number. 02 02
Data for each test case consists of 10 data points. Tables I and TC-P-MEM- TC-P-RT-
30,00 – 39,000 TC-P-CPU-03
II show the test cases for each design pattern. 03 03
TC-P-MEM- TC-P-RT-
40,000– 49,000 TC-P-CPU-04
C. EXPERIMENT 04 04
The testing involves running the application populated with
dummy data according to the designated test cases. During
testing, the device is connected to Android Profiler to record irregularities, they may lack the statistical power of parametric
CPU and memory usage while opening the dashboard. For tests, meaning that detecting significant effects might require
response time, logs from Android Studio are used to measure larger sample sizes or more pronounced differences.
the time the application takes to display data. Before testing In this study, hypotheses were formulated to examine
begins, it is ensured that no other applications are running on whether there are significant differences in application
the device. The testing was conducted on an Android device, performance metrics—CPU usage, memory usage, and
specifically a Samsung Galaxy A52 with a Snapdragon 778G response time—between the MVVM and MVP design patterns.
processor, 8GB of RAM, and Android 14 operating system. These hypotheses were tested using appropriate statistical
techniques based on the processed research data. For example,
D. DATA COLLECTION an independent sample t-test was applied if the data met
CPU and memory performance were recorded using parametric criteria, allowing for accurate comparisons of mean
Android Profiler, while dashboard response times were differences between the two design patterns. Conversely, if
collected using Android activity’s built-in methods. Data assumptions were violated, the Mann-Whitney U test was used
collection occurs exclusively during the dashboard activity. to compare the distributions of performance metrics without
When the dashboard was opened, CPU and memory usage relying on normality.
appeared in the Android Profiler. CPU usage data were The completion of the data analysis phase provided critical
summed and averaged, while memory usage was measured at insights into the performance characteristics of each design
the start, peak, and end of the activity, then averaged. pattern. The results determined whether observed differences
E. DATA ANALYSIS
in CPU usage, memory consumption, or response time were
Quantitative data were analyzed using parametric or statistically significant or merely due to random variation.
nonparametric statistical techniques, depending on the These findings not only validated the hypotheses but also
characteristics of the data. Parametric statistics are typically offered actionable conclusions regarding the suitability of
used when the data meet certain assumptions, including normal MVVM and MVP for various application scenarios,
contributing to a deeper understanding of their performance
distribution, linearity, and homogeneity of variance. These
trade-offs in Android development.
assumptions are crucial as they ensure the reliability and
By employing a rigorous and adaptive statistical approach,
validity of tests such as t-tests or ANOVA, which rely on
this study ensured that the analysis was scientifically robust and
precise mathematical models to evaluate differences or
capable of adapting to the nuances of the data, providing
relationships between variables. For instance, the normality of
reliable evidence to support the conclusions drawn. This
data distribution is often assessed using tools such as the
meticulous methodology reinforced the validity of the research
Shapiro-Wilk test, while homogeneity of variance can be
findings and emphasized the importance of selecting
evaluated using Levene’s test.
appropriate statistical techniques tailored to the characteristics
If the data fail to meet these assumptions, nonparametric of the data.
statistical techniques are used as an alternative. Methods such
as the Mann-Whitney U test or Kruskal-Wallis test do not rely F. RESULTS AND DISCUSSION
on strict assumptions about the underlying data distribution, After statistical analysis is completed, the results will yield
making them more robust for analyzing skewed or nonlinear conclusions about whether there is a significant difference
data. While nonparametric tests are less sensitive to outliers and between the MVVM and MVP design patterns in displaying

p-ISSN 2301–4156 | e-ISSN 2460–5719 Fajar Pradana: Comparative Analysis of MVVM …

Jurnal Nasional Teknik Elektro dan Teknologi Informasi EN-91
Volume 14 Number 2 May 2025

data on an Android application’s dashboard. The following 14

criteria must be satisfied to determine the optimal design

CPU Usage (in Percentage)

pattern. 12
1. The design pattern is considered better if it has a lower
10
average CPU usage.
2. The design pattern is considered better if it has a lower 8
average memory usage.
3. The design pattern is considered better if it demonstrates 6
faster response times. 4
MVVM
IV. RESULTS AND DISCUSSION 2
MVP
A. RESULTS
0
The testing was conducted on an Android device, TC1 TC2 TC3 TC4
specifically a Samsung Galaxy A52 with a Snapdragon 778G
processor, 8GB of RAM, and Android 14 operating system. Test Case
The tests were divided into four test cases (TC) based on the
Figure 3. CPU performance comparison.
range of data used in each test. For TC-1, which used a data
range of 10,000–19,000 entries in the application’s database, 16
the average CPU usage was 7.24% for MVVM and 9.54% for
MVP. In TC-2 (20,000–29,000 data), average CPU 14
usage was 8.49% for MVVM and 10.90% for MVP. In TC-
3 (30,000–39,000 data), it was 9.37% for MVVM and 11.76%

CPU Usage (in Percentage)

12
for MVP. Finally, in TC-4 (40,000–49,000 data), CPU usage
averaged 10.57% for MVVM and 12.39% for MVP. Figure 3
10
illustrates the rising trend in CPU usage for each design pattern,
with MVVM represented by a blue line and MVP by a red line.
8 MVVM
In the CPU category, MVVM consistently outperformed MVP
by using less CPU in all test cases. MVP
6
The distribution of CPU usage data can be visualized using
a boxplot in Figure 4. The lowest CPU usage is represented by
the bottommost point, at 6% for MVVM and 7.43% for MVP. 4
The highest CPU usage is represented by the topmost point, at
11.86% for MVVM and 14% for MVP. The box itself 2
represents the interquartile range (IQR), which is the difference
between the first and third quartiles, illustrating the variability 0
of the data around the median. The median CPU usage is Design Pattern
marked by the line inside the box, at 9% for MVVM and 11%
for MVP. Visually, MVVM appears to be more efficient in Figure 4. Boxplot diagram of CPU usage.
CPU usage compared to MVP.
In TC-1, with a data range of 10,000–19,000, the average 135
memory usage was 113.03 MB for MVVM and 113.35 MB for
Memory Usage (in Percentage)

130
MVP. In TC-2, with a range of 20,000–29,000 data, the average
memory usage was 119.84 MB for MVVM and 120.39 MB for 125
MVP. For TC-3, with 30,000–39,000 data, the average memory
usage was 124.10 MB for MVVM and 123.37 MB for MVP. 120
Finally, in TC-4, with 40,000–49,000 data, the average
115
memory usage was 128.96 MB for MVVM and 129.10 MB for
MVP. The visualization of the increasing average memory 110
usage is shown in Figure 5, with MVVM represented by a blue MVVM
line and MVP by a red line. In the memory performance 105
MVP
category, MVVM and MVP were closely matched, with MVP 100
performing better in TC-1, TC-2, and TC-4, while MVVM TC1 TC2 TC3 TC4
outperformed in TC-3.
The distribution of memory usage data can be visualized Test Case
using a boxplot, as shown in Figure 6, with MVVM represented Figure 5. Memory performance comparison.
by a blue color and MVP by a red color. The bottommost point
represents the lowest memory usage at 107.37 MB for MVVM
and 108.63 MB for MVP. The highest memory usage is exhibits a wider data spread and the lowest memory usage,
represented by the topmost point at 137.83 MB for MVVM and while the MVP design pattern has a narrower data spread.
133.97 MB for MVP. The median memory usage is indicated In TC-1, with a data range of 10,000–19,000, the average
by the line within the box, at 122.70 MB for MVVM and response time was 144.79 ms for MVVM and 134.19 ms for
122.23 MB for MVP. Visually, the MVVM design pattern MVP. In TC-2, with a data range of 20,000–29,000, the average

Fajar Pradana: Comparative Analysis of MVVM … p-ISSN 2301–4156 | e-ISSN 2460–5719

EN-92 Jurnal Nasional Teknik Elektro dan Teknologi Informasi
Volume 14 Number 2 May 2025

140 400

Response Time (in Percentage)

350
135
300
Memory Usage (in Percentage)

130
250

125 200

150
120 MVVM
100
MVP MVVM
115
50 MVP

110 0
TC1 TC2 TC3 TC4

105 Test Case

100 Figure 7. Response time comparison.

Design Pattern 450

Figure 6. Boxplot diagram of the memory usage.

400
Response Time (in Percentage)

response time was 217.16 ms for MVVM and 202.28 ms for 350
MVP. In TC-3, with a data range of 30,000–39,000, the average
response time was 285.17 ms for MVVM and 270.49 ms for 300
MVP. Finally, in TC-4, with a data range of 40,000–49,000, the
average response time was 363.61 ms for MVVM and 340.57 250 MVVM
ms for MVP. The visualization of the increasing average
response time for each design pattern is shown in Figure 7, with MVP
200
MVVM represented by a blue line and MVP by a red line.
MVVM outperformed in TC-1, while MVP performed better in 150
TC-2, TC-3, and TC-4, with a relatively close gap between the
two design patterns. Graphically, MVP is faster at displaying
100
data compared to MVVM, though the difference in speed
between the two design patterns is quite narrow.
50
The distribution of response time data can be visualized
using a boxplot, as shown in Figure 8, with MVVM represented Design Pattern
in blue and MVP in red. The lowest response time is at the
bottommost point, 116.30 ms for MVVM and 107.70 ms for Figure 8. Boxplot diagram response time.
MVP. The highest response time is at the topmost point, 407.50 TABLE III
ms for MVVM and 390.80 ms for MVP. The median response NORMALITY TESTING
time is indicated by the line within the box, at 256.50 ms for
Design Response
MVVM and 231.65 ms for MVP. Visually, both design patterns CPU Memory
Pattern Time
have boxes of the same size, but MVP is slightly more optimal MVVM 0.986 0.674 0.1768
than MVVM. MVP 0.8933 0.1883 0.1809
Once the experimental data were collected, the next step
was to analyze it. The first analysis involved testing the
normality of the data, which determines the type of difference summarizes the Shapiro-Wilk normality test results for each
test to use. If the data are normally distributed, a parametric test performance category.
is applied. Otherwise, a nonparametric test is used. Homogeneity testing was conducted to determine whether
For CPU usage with MVVM, the p-value was 0.986, which the variance between groups is equal or homogeneous, which
is greater than 0.05, indicating a normal distribution. Similarly, is required for an independent t-test. For CPU, the p-value was
CPU usage for MVP had a p-value of 0.8933, which is also 0.5931, greater than 0.05, indicating homogeneous variance.
greater than 0.05, confirming normal distribution. For the For memory, the p-value was 0.9689, also greater than 0.05,
memory usage, MVVM had a p-value of 0.674, while and MVP confirming homogeneity. For response time, the p-value was
had a p-value of 0.1883. These results indicate normal 0.756, greater than 0.05, indicating homogeneous variance.
distribution. For response time, MVVM and MVP had p-values Since all data categories had homogeneous variances, an
of 0.1768 and 0.1809, respectively, both greater than 0.05, independent t-test was conducted.
confirming normality. With the prerequisites fulfilled, the final testing was
Thus, all data categories were found to be normally conducted to determine whether there was a significant
distributed, allowing for parametric testing. Table III difference between the MVVM and MVP design patterns in

p-ISSN 2301–4156 | e-ISSN 2460–5719 Fajar Pradana: Comparative Analysis of MVVM …

Jurnal Nasional Teknik Elektro dan Teknologi Informasi EN-93
Volume 14 Number 2 May 2025

each performance category. The hypotheses for the where the Presenter actively controls UI updates and might
independent t-test were as follows: keep unnecessary UI-bound computations within the main
1. null hypothesis (H0): If p > 0.05, no significant thread.
difference exists; Conversely, MVP demonstrated superior response times,
2. alternative hypothesis (H1): If p ≤ 0.05, a significant likely due to its simpler data flow architecture, where the
difference exists. presenter acts as a direct channel between the model and the
For CPU usage, the p-value was 1.866e-10 (less than 0.05), view. This aligns with [28], which observed that minimizing
indicating a significant difference between the design patterns. the number of intermediary layers in the data flow can result in
For memory usage, the p-value was 0.9608, greater than 0.05, faster response times. However, this advantage comes at the
indicating no significant difference. For response time, the p- cost of higher CPU usage, as the presenter requires more
value was 0.756, also greater than 0.05, indicating no frequent interactions with the View and Model components,
significant difference. Thus, a significant difference was especially in complex applications.
observed in CPU usage between the design patterns, while no Interestingly, MVVM’s slight advantage in memory usage
significant difference was found for memory usage and compared to MVP contradicts previous findings [17], where
response time. Additionally, the analysis results indicate that in MVP was reported to excel in this category. A possible
terms of CPU usage, MVVM is more efficient than MVP, with explanation is MVVM’s more efficient handling of lifecycle-
Cohen’s d = -1.64, representing a very large and significant aware components, which reduces the likelihood of memory
effect. Conversely, in response time, MVP demonstrates better leaks—a common issue in MVP when managing long-lived
performance than MVVM, with Cohen’s d = 1.07, which is also Presenters. This aligns with [20], which emphasizes the
a large effect. Meanwhile, in memory usage, the difference importance of lifecycle-aware components in effectively
between the two design patterns is not significant (Cohen’s d = managing memory consumption.
-0.03) with a high p-value, indicating that memory usage Overall, these findings underscore the nuanced trade-offs
between MVVM and MVP is relatively the same. Regarding between MVVM and MVP, particularly when applied to
potential biases, the results of the normality test (Shapiro-Wilk Android development using Kotlin. While MVVM
test) and homogeneity of variance test (Levene’s test) confirm demonstrates superior CPU efficiency and marginally better
that the statistical assumptions are met, ensuring that the t-test memory usage, MVP offers a more responsive user experience.
results are reliable and free from significant bias. The choice between the two should consider the application’s
performance priorities and complexity, as well as the
B. DISCUSSION
development team’s familiarity with each design pattern.
Among the three performance categories—CPU, memory,
Future research could further explore these trade-offs by
and response time—only CPU usage showed a significant
including more complex scenarios or integrating additional
difference between the MVVM and MVP design patterns.
performance metrics, such as energy consumption or
MVVM was superior in CPU usage, with an average usage of
maintainability.
8.02%, compared to 11.15% for MVP. For memory usage,
This study provides valuable insights into the comparative
MVVM slightly outperformed MVP with an average of 121.48
performance of the MVVM and MVP design patterns in
MB versus 121.55 MB. However, MVP excelled in response
Android applications developed with Kotlin. One of the main
time, with an average of 236.88 ms compared to 252.68 ms for
advantages of this research is its empirical evaluation of real-
MVVM. This finding differs from previous research [17]
world application scenarios, ensuring that the findings are
which reported that MVVM was superior in both CPU usage
relevant and applicable to modern Android development.
and response time, while MVP excelled in memory usage. This
Additionally, the use of statistical analysis strengthens the
discrepancy could be attributed to differences in testing
validity of the conclusions, offering developers concrete data
environments, application complexity, or the specific use of
to support design pattern selection based on specific
Kotlin in this study, which may influence performance metrics.
performance needs. However, the study also has limitations.
Additionally, variations in how the UI rendering pipeline is
The testing was conducted on a single device model, which
managed between different versions of Android could also
may not fully capture variations in hardware performance
contribute to these differences, as certain optimizations in UI
across different Android devices. Furthermore, the study
thread execution may favor one architecture over the other.
primarily focuses on CPU usage, memory consumption, and
Based on the results of the independent sample t-test, only
response time, while other important factors such as energy
the CPU category showed a significant difference in
efficiency, maintainability, and scalability were not explored in
application performance between the design patterns for
depth. Future research could address these limitations by
Android-based applications developed using Kotlin. This
expanding the scope to include a broader range of devices and
difference is assumed to arise from the distinct ways each
additional performance metrics to provide a more
design pattern manages data flow and presents data to the user.
comprehensive evaluation of these design patterns.
The key distinction lies in the intermediary component used to
handle requests to the model and deliver data to the view: the V. CONCLUSION
viewmodel in MVVM and the presenter in MVP. The This study compared the performance of MVVM and MVP
viewmodel facilitates a reactive data-binding mechanism that design patterns in an Android application built with Kotlin,
reduces the overhead of frequent updates, thereby optimizing focusing on CPU usage, memory consumption, and response
CPU usage. This finding is consistent with [16], which time. The results indicate that MVVM outperforms MVP in
highlighted the efficiency of reactive programming paradigms CPU efficiency, with an average usage of 8.92% compared to
in reducing CPU load. Furthermore, since MVVM leverages 11.15%. MVVM also demonstrated slightly better memory
LiveData and flows in Kotlin, it offloads computation-heavy efficiency (121.48 MB vs. 121.55 MB), though the difference
operations to background threads more efficiently than MVP, was not statistically significant. In contrast, MVP exhibited

Fajar Pradana: Comparative Analysis of MVVM … p-ISSN 2301–4156 | e-ISSN 2460–5719

EN-94 Jurnal Nasional Teknik Elektro dan Teknologi Informasi
Volume 14 Number 2 May 2025

faster response times, averaging 236.88 ms compared to 252.68 design,” Sustainability, vol. 14, no. 15, pp. 1–26, Aug. 2022, doi:
10.3390/su14159131.
ms for MVVM.
[10] D. Amalfitano, M. Júnior, A.R. Fasolino, and M. Delamaro, “A GUI-
These findings suggest that the choice of design pattern based metamorphic testing technique for detecting authentication
should be based on the application’s performance priorities. vulnerabilities in Android mobile apps,” J. Syst. Softw., vol. 224, pp. 1–
MVVM is more suitable for applications requiring optimized 17, Jun. 2025, doi: 10.1016/j.jss.2025.112364.
CPU and memory management, particularly those with [11] N. Hoshieah, S. Zein, N. Salleh, and J. Grundy, “A static analysis of
complex data-binding scenarios or resource constraints. MVP, Android source code for lifecycle development usage patterns,” J.
Comput. Sci., vol. 15, no. 1, pp. 92–107, Jan. 2019, doi:
on the other hand, is preferable for applications demanding 10.3844/jcssp.2019.92.107.
real-time responsiveness with minimal latency. [12] B.S. Panca, S. Mardiyanto, and B. Hendradjaya, “Evaluation of software
Future research could expand on these findings by design pattern on mobile application based service development related
incorporating additional metrics such as energy consumption, to the value of maintainability and modularity,” in 2016 Int. Conf. Data
maintainability, and scalability. Furthermore, testing on a wider Softw. Eng. (ICoDSE), 2016, pp. 1–5, doi:
10.1109/ICODSE.2016.7936132.
range of devices and real-world scenarios could provide deeper
[13] B.B. Mayvan, A. Rasoolzadegan, and Z.G. Yazdi, “The state of the art on
insights into the performance of these design patterns under design patterns: A systematic mapping of the literature,” J. Syst. Softw.,
various conditions, ensuring broader applicability of the vol. 125, pp. 93–118, Mar. 2017, doi: 10.1016/j.jss.2016.11.030.
conclusions drawn from this study. [14] A. Naghdipour, S.M.H. Hasheminejad, and M.R. Keyvanpour, “DPSA:
A brief review for design pattern selection approaches,” in 2021 26th Int.
CONFLICTS OF INTEREST Comput. Conf. Comput. Soc. Iran (CSICC), 2021, pp. 1–6, doi:
The authors declare that there is no conflict of interest in the 10.1109/CSICC52343.2021.9420629.
research and preparation of this paper. [15] D. Panchal, “Comparative study on Android design patterns,” Int. Res. J.
Eng. Technol., vol. 7, no. 9, pp. 833–840, Sep. 2020.
AUTHORS’ CONTRIBUTIONS [16] R.L.B. Baptista, “Framedrop-Mobile Client,” M.S. thesis, University of
Conceptualization, Fajar Pradana and Raziqa Izza Coimbra, Coimbra, Portugal, 2023.
Langundi; methodology, Fajar Pradana; software, Raziqa Izza [17] B. Wisnuadhi, G. Munawar, and U. Wahyu, “Performance comparison of
native Android application on MVP and MVVM,” in Proc. Int. Semin.
Langundi; validation, Djoko Pramono; formal analysis, Nur Ida Sci. Appl. Technol. (ISSAT 2020), 2020, pp. 276–282, doi:
Iriani; investigation, Fajar Pradana; resources, Raziqa Izza 10.2991/aer.k.201221.047.
Langundi; data curation, Raziqa Izza Langundi; writing— [18] M. Willocx, J. Vossaert, and V. Naessens, “Comparing performance
original draft preparation, Fajar Pradana; writing—reviewing parameters of mobile app development strategies,” in MOBILESoft '16,
Proc. Int. Conf. Mob. Softw. Eng. Syst., 2016, pp. 38–47, doi:
and editing, Fajar Pradana; visualization, Raziqa Izza Langundi;
10.1145/2897073.2897092.
supervision, Fajar Pradana; project administration, Nur Ida [19] R.A. Doherty and P. Sorenson, “Keeping users in the flow: Mapping
Iriani; funding acquisition, Nur Ida Iriani. system responsiveness with user experience,” Procedia Manuf., vol. 3,
pp. 4384–4391, 2015, doi: 10.1016/j.promfg.2015.07.436.
ACKNOWLEDGMENT [20] F. Rösler, A. Nitze, and A. Schmietendorf, “Towards a mobile application
This research was funded by the Faculty of Computer performance benchmark,” in ICIW 2014, 9th Int. Conf. Internet Web Appl.
Science, Universitas Brawijaya, Malang. Serv., 2014, pp. 55–59.
[21] G. Lim, C. Min, and Y.I. Eom, “Enhancing application performance by
REFERENCES memory partitioning in Android platforms,” in 2013 IEEE Int. Conf.
[1] International Data Corporation (IDC) “Worldwide smartphone market Consum. Electron. (ICCE), 2021, pp. 649–650, doi:
forecast to grow 6.2% in 2024, fueled by Robust growth for Android in 10.1109/ICCE.2013.6487055.
emerging markets and China, according to IDC.” Access date: 23-Jan- [22] W. Ngaogate, “Applying the Flyweight design pattern to Android
2025. [Online]. Available: application development,” ASEAN J. Sci. Technol. Rep. (AJSTR), vol. 26,
https://www.idc.com/getdoc.jsp?containerId=prUS52757624 no. 2, pp. 49–57, Apr.-Jun. 2023, doi: 10.55164/ajstr.v26i2.247607.
[2] A. Karapantelakis et al., “Generative AI in mobile networks: A survey,” [23] D. Rimawi and S. Zein, “A model based approach for Android design
Ann. Telecommun., vol. 79, no. 1–2, pp. 15–33, Feb. 2024, doi: patterns detection,” in 2019 3rd Int. Symp. Multidiscip. Stud. Innov.
10.1007/s12243-023-00980-9. Technol. (ISMSIT), 2019, pp. 1–10, doi: 10.1109/ISMSIT.2019.8932921.
[3] D. Rimawi and S. Zein, “A static analysis of Android source code for [24] R.F. García, “MVVM: Model–view–viewmodel,” in iOS Architecture
design patterns usage,” Int. J. Adv. Trends Comput. Sci. Eng., vol. 9, no. Patterns. Berkeley, CA, USA: Apress, 2023, pp. 145–224.
2, pp. 2178–2186, Mar./Apr. 2020, doi: [25] X. Li, S. Wang, Z. Liu, and G. Wu, “Design and implementation of
10.30534/ijatcse/2020/194922020. enterprise web application common framework based on model-view-
[4] S. Papadakis, M. Kalogiannakis, and N. Zaranis, “Educational apps from viewmodel architecture,” in 5th Int. Conf. Mechatron. Comput. Technol.
the Android Google Play for Greek preschoolers: A systematic review,” Eng. (MCTE 2022), 2022, pp. 1-4, doi: 10.1117/12.2661040.
Comput. Educ., vol. 116, pp. 139–160, Jan. 2018, doi: [26] M.I. Alfathar et al., “Penerapan MVVM (model view viewmodel) pada
10.1016/j.compedu.2017.09.007. pengembangan aplikasi bank sampah digital,” J. Ris. Apl. Mhs. Inform.
[5] J.B. Jorgensen et al., “Variability handling for mobile banking apps on (JRAMI), vol. 5, no. 2, pp. 406–414, Apr. 2024, doi:
iOS and Android,” in 2016 13th Work. IEEE/IFIP Conf. Softw. Archit. 10.30998/jrami.v5i2.11071.
(WICSA), 2016, pp. 283–286, doi: 10.1109/WICSA.2016.29. [27] C.J. Sampayo-Rodríguez, R. González-Ambriz, B.A. Gonzalez-Martinez,
[6] F.M. Kundi, A. Habib, A. Habib, and M.Z. Asghar, “Android-based and J. Aldana-Herrera, “Processor and memory performance with design
health care management system,” Int. J. Comput. Sci. Inf. Secur. (IJCSIS), patterns in a native Android application,” J. Appl. Comput., vol. 6, no. 18,
vol. 14, no. 7, pp. 77–87, Jul. 2016. pp. 53–61, Jun. 2022, doi: 10.35429/JCA.2022.18.6.53.61.
[7] M. Prakash, U. Gowshika, and T. Ravichandran, “A smart device [28] F.F. Anhar, M.H.P. Swari, and F.P. Aditiawan, “Analisis perbandingan
integrated with an Android for alerting a person’s health condition: implementasi clean architecture menggunakan MVP, MVI, dan MVVM
Internet of things,” Indian J. Sci. Technol., vol. 9, no. 6, pp. 1–6, Feb. pada pengembangan aplikasi Android native,” Jupiter, Publ. Ilmu
2016, doi: 10.17485/ijst/2016/v9i6/69545. Keteknikan Ind. Tek. Elekt. Inform., vol. 2, no. 2, pp. 181–191, Mar. 2024,
[8] G.H. Prakash et al., “Development and validation of Android mobile doi: 10.61132/jupiter.v2i2.155.
application in the management of mental health,” Clin. Epidemiol. Glob. [29] A. Moreno-Azze, D. López-Plaza, F. Alacid, and D. Falcón-Miguel,
Health, vol. 31, pp. 1–7, Jan./Feb. 2025, doi: “Validity and reliability of an iOS mobile application for measuring
10.1016/j.cegh.2024.101894. change of direction across health, performance, and school sports
[9] W. Li, Y. Zhou, S. Luo, and Y. Dong, “Design factors to improve the contexts,” Appl. Sci., vol. 15, no. 4, pp. 1–11, Feb. 2025, doi:
consistency and sustainable user experience of responsive interface 10.3390/app15041891.

p-ISSN 2301–4156 | e-ISSN 2460–5719 Fajar Pradana: Comparative Analysis of MVVM …

Jurnal Nasional Teknik Elektro dan Teknologi Informasi EN-95
Volume 14 Number 2 May 2025

[30] L. Corral, A. Sillitti, and G. Succi, “Mobile multiplatform development: Technol. (iJIM), vol. 17, no. 3, pp. 21–38, Feb. 2023, doi:
An experiment for performance analysis,” Procedia Comput. Sci., vol. 10, 10.3991/ijim.v17i03.36957.
pp. 736–743, 2012, doi: 10.1016/j.procs.2012.06.094. [32] S. Pargaonkar, “A comprehensive review of performance testing
[31] F. Pradana, P. Setyosari, S. Ulfa, and T. Hirashima, “Development of methodologies and best practices: Software quality engineering,” Int. J.
gamification-based e-learning on web design topic,” Int. J. Interact. Mob. Sci. Res. (IJSR), vol. 12, no. 8, pp. 2008–2014, Aug. 2023, doi:
10.21275/SR23822111402.

Fajar Pradana: Comparative Analysis of MVVM … p-ISSN 2301–4156 | e-ISSN 2460–5719