A Framework for App Store Optimization
Artur Strzelecki1[0000-0003-3487-0971]
1
University of Economics in Katowice, Katowice 40-287, Poland
[email protected]
Abstract. In this paper a framework for app store optimization is proposed. The
framework is based on two main areas: developer dependent elements and user
dependent elements. Developer dependent elements are similar to factors in
search engine optimization. User dependent elements are similar to activities in
social media. The proposed framework is modelled after downloading sample
data from two leading app stores: Google Play and Apple iTunes. Results show
that developer dependent elements can be better optimized. Names and descrip-
tions of mobile apps are not fully utilized.
Keywords: App Store Optimization, Google Play, Apple iTunes, Mobile App
Store.
1 Introduction
Together with increasing use of mobile devices there is increased supply of applications
(apps) used on mobile devices. Software companies and developers create apps that
installed on mobile device can be useful for mobile device owners. Now in two most
popular app stores (iTunes and Google Play) are available millions of apps. Apps are
offered in these stores in different categories of software, however still, competition is
very high between apps in one category. Competition means that many apps offer sim-
ilar functions and user can browse and choose from variety of selection.
This creates a need to propose a framework of app store optimization. App store
optimization also can be described as app store marketing or mobile app search engine
optimization. The focus is to improve ranking of mobile app directly within app store.
Recently [1] noticed that developers make efforts to improve mobile app visibility in
app store, but some of such actions can be treated as fraud. Especially when it comes
to reviews and downloads, they can be pumped by developer. There is limited research
on factors which are taken into account when apps ranking is created. Since apps are
created by developers and used by users, ranking is influenced by these two groups.
App store optimization has its roots in search engine optimization, where different fac-
tors are taken into account, when it comes to create ranking of websites. Earning from
achievement in the field of search engine optimization author propose a framework for
app store optimization. There are identified similarities between these two areas.
The research question is what factors are used for creating ranking of mobile apps
directly in app store. Author in this paper propose a framework for ranking in app store
optimization. This paper is organized as follows. Section 2 contains a review of the
2
relevant literature on app stores and optimization in search engines area. In section 3
author describe the research method for choosing factors to propose framework. Sec-
tion 4 contains characteristics of the sample date which are collected from iTunes and
Google Play. In section 5 author highlight the contribution of the research and suggest
possible implications of results, analyze current limitations of the research, draw con-
clusions and present ideas as for their future research on app store optimization.
2 Literature review
In literature review two areas are explored. First is about efforts made by app develop-
ers to promote apps in app store and user’s which rate and review apps. Developer
creates app and deliver it to app store. After software is delivered it can be offered with
free app offers, continuous quality updates, investment in less popular (less competi-
tive) categories and price changes [2]. New releases are found to change user opinion
on app [3]. Number of rating, number of reviews, number of downloads are always
positively changed, since they only can rise. Some developers update their apps very
frequently, even once a week or twice a month. They are not too concerned about de-
tailing the content of new updates and users are not too concerned about such infor-
mation, whereas users highly rank frequently-updated apps instead of being annoyed
about the high update frequency [4]. App rating is assigned to an app over its entire life
time is aggregated into one rating that is displayed in the app store. However many apps
do increase their version-to-version rating, while the store-rating of an app is resilient
to fluctuations once an app has gathered a substantial number of raters [5]. The ap-
proach is proposed to assisting developers to select the proper release opportunity based
on the purpose of the update and current condition of the app [6]. Developer also can
do a shady moves to fraud app ranking. [1] defined two methods of fraud: inflate the
app number downloads and ratings in a very short time.
User can publish review about app. Reviews have a major influence on the user’s
purchase decision [7]. Average rating according to the star principle as well as the num-
ber of reviews given determine the buying decision of an app to a very large degree.
Review area is explored when it comes to see what the sentiment of the review is. Re-
view can have positive or negative sentiment [8]. Analyzing reviews is done by text
mining [9]. Most of the feedback is provided shortly after new releases, with a quickly
decreasing frequency over time. Reviews typically contain multiple topics, such as user
experience, bug reports, and feature requests [10]. Reviews are source for users’ feed-
back, requests for new features or reporting bugs. Reviews represent feature requests,
i.e. comments through which users either suggest new features for an app or express
preferences for the re-design of already existing features of an app [11]. Reviews can
be specific, as specific is an app. Analyzing feedbacks from a health and fitness-track-
ing app shows that the users of health and fitness-related apps are concerned about their
physical activity records and physiological records. The records include track, distance,
time, and calories burned during jogging or walking. App store reviews are used to
analyze different aspects of app development and evolution [12]. There are proposed
frameworks to acquire reviews in large number, extract informative user reviews by
3
filtering noisy and irrelevant ones, then group the informative reviews automatically
using topic modeling [13, 14]. There also systematic literature reviews where of opin-
ion mining studies from mobile app store user reviews, which describes and compares
the areas of research that have been explored thus far, drawing out common aspects, in
app store analysis [15, 16].
Second area in literature review is search engine optimization. Since app store rank-
ing is generated based on different factors, drawing from area of web search engines
can help to build framework for app store optimization. Web search engines use dif-
ferent factors identified on and off website to determine ranking of certain webpage in
search engines. In the beginning ranking factors were limited only to few elements
taken into account when search engine results pages were created [17, 18]. Nowadays
this topic is more explored and can be divided into onsite factors and offsite factors.
Onsite factors are domain-related, website-related and page related [19]. Offsite factors
are link-related [20], user-action-related [21], special-rules-related [22], brand-related
[23] and spam-related [24].
3 Framework model
App stores distribute apps through the app store and have additional information about
the app. A set of information is initially set by the developer. The app is delivered by
the app developer. The developer sets a name for the app and creates a description of
app features. The app is published with the new original url. The developer sets genre
content rating and define system requirements for an app. Apps can be distributed
through the app store for free or user needs to pay to download. Free apps can also offer
in-app purchases for additional features. Another set of information is created after app
is released. Users are downloading the app and make the number of downloads grow-
ing. User can also create reviews and rate an app in range of scale from 1 to 5.
Genre Developer Name
Requirements Short description Subtitle
URL Long description Content rating
What’s new Price Last update
Mobile Application
Downloads Average rating Reviews
4
Fig. 1. A framework for app store optimization
Data for analysis and further usage in framework for app store optimization were col-
lected from Google Play and Apple iTunes. Author used an automated software for
websites crawling - Screaming Frog SEO Spider, which retrieved data, divided into
several groups based on their type. Two different techniques for retrieving were used.
First technique applied regular expressions to match elements like number of down-
loads, last date of update, content rating, range of pricing and software requirements.
The second technique used CSSPath to match the next elements like developer, number
of reviews, category, name, average rating. The URL was also retrieved for each appli-
cation during the crawling process.
These two different techniques were needed, because some of the elements are writ-
ten down firmly into a website structure and are always placed in the same context.
These were retrieved by the CSSPath. The rest of elements can change their position in
website structure, due to the incomplete data provided by the developer. Some of the
apps in Google Play and Apple iTunes do not have all of the information usually dis-
played on app store. Regular expressions helped to collect the published in different
part of websites.
Framework proposition is based on two areas. One area depends on the developer.
Its content and settings are provided when the app is initially released in app store.
Second area depends on users. If the app is being popular among user, they are starting
to download and create reviews or rate the app in the app store.
3.1 Framework elements dependent on developer
Developer. The name of the developer is also a ranking factor, affecting directly the
position of the application itself. The positive history of the developer affects the better
evaluation of the application in the search ranking. Keywords in the developer's name
affects each of its applications.
Name. The name of the application is important both for the app store optimization and
for the experience. Often observed solution made by developers is to create a name by
combining the brand name and the most important key words for the app. Google Play
limits the name to 50 characters, whereas Apple iTunes limits the name to 30 characters.
Subtitle. (Only in Apple iTunes) Subtitle is placed right below the title and brings ad-
ditional information on mobile app. It complements the app name by communicating
the purpose and value of app in details and is limited to 30 characters.
Genre. Genre is a category for software. Currently Google Play offers 31 categories
for mobile apps and Apple iTunes offers 28 categories.
Description. App stores allows to prepare two types of description, short and long ver-
sion. The short version is only visible in mobile app store and its maximum length is
80 characters for Google Play and 170 characters for Apple iTunes. Long version is
also visible in a desktop version and the maximum size is 4000 characters. However,
only around 250 characters is visible after description is displayed, the rest is hidden. It
can be showed after clicking button “Read more”.
5
Content rating. A content rating rates the suitability of mobile applications for its au-
dience. It tells what age group is suitable to use mobile app.
Requirements. App store providers require that new apps target operating system ver-
sion. System version is set in requirements. For new apps requirement in Google Play
is at least Android 8.0 and in Apple App Store is at least iOS 12. Older apps can have
lower operating system requirements.
URL. The URL can only be defined when the application is published, hence it is an
element worth refining.
Last update. (Only in Google Play) For the application store, security is important,
therefore the factor that has a ranking importance, but also affects the opinions about
the application, is the frequency of its updating. It is necessary to update the application
not only at the time of major changes to the functionality of the app, but also at every
subsequent update of the operating system.
What’s new. This element is intended to describe the updates introduced into the app.
Developers can change the contents of the section only after the new app version is
submitted to the store.
3.2 Framework elements dependent on users.
Downloads. (Only in Google Play) Downloads is the number of app installation on
devices worldwide. Google Play publishes the only threshold number that has been
reached.
Average rating. Average rating is a number with one decimal place in the scale from
1 to 5 and it is an aggregated rating from all ratings given by users.
Reviews. Apart from ratings users can write a review about an app. Reviews are a
source for users’ feedback, requests for new features or reporting bugs [25].
4 Data and Results
4.1 Data
In the process of app store mining, author has downloaded from Google Play about 50
000 apps, which belong to 31 different categories. Additionally, there is the main cate-
gory Games which includes 17 subcategories (action, adventure, arcade, board, card,
casino, casual, educational, music, puzzle, racing, role playing, simulation, sports, strat-
egy, trivia, word). In the downloaded sample games take 16,66% of total number of
applications, and other apps belong to 31 different categories. Figure 2 represents shares
of all categories in this sample.
6
Fig. 2. App categories in Google Play store
Table 1 shows the descriptive statistics of 49 990 sample apps downloaded from Google
Play and 6040 sample apps downloaded from Apple iTunes. Based on downloaded data
it contains main characteristics. Table 1 also contains framework elements dependent
on developers based on Google Play and Apple iTunes.
Table 1. Descriptive statistics of samples from Google Play and Apple iTunes
App Store Google Play Apple iTunes
Number of downloaded apps 49990 6040
Number of developers 16912 2786
Number of genres 31 25
Avg length of name 23 23
Avg length of subtitle na 26
Avg length of description 1338 1694
Median length of description 1052 1489
Apps with reviews 97% 79%
Average number of avg rating 4,29 4,35
Autor downloaded smaller sample from Apple iTunes because this app store prevents
against the massive download of data and after a few hundreds of downloaded apps, it
turns on the HTTP response code 403 on every page with an app. Table 2 contains
7
framework elements dependent on users based on Google Play like number of down-
loads, average number of reviews in download threshold and average from average
rating in download threshold.
Table 2. Framework elements dependent on users.
Download threshold Number of apps Avg of reviews Avg from avg rating
0 67 21 4,35
1 217 2 4,90
5 158 61 4,76
10 767 4 4,69
50 542 4 4,57
100 2324 8 4,43
500 1517 15 4,36
1000 5025 37 4,25
5000 2933 86 4,22
10000 8502 283 4,26
50000 4392 785 4,27
100000 10016 2757 4,27
500000 3753 9020 4,29
1000000 5833 31539 4,31
5000000 1520 102576 4,32
10000000 1813 343902 4,34
50000000 322 1113916 4,37
100000000 244 3541721 4,39
500000000 27 9739412 4,41
1000000000 17 26792154 4,27
5000000000 1 39797335 4,40
4.2 Results
Data in the Google Play store adapts to language settings of user’s browser and operat-
ing system. During initial screening the authors have seen different settings coming
from different languages. First, prices of apps itself or in-app purchases were displayed
in the currency set in the web browser. Second, the types of content rating were differ-
ent for different localization settings. Author checked three options of language &
country setting: Poland, Ukraine and US. For Poland and Ukraine, the rating is dis-
played in Pan European Game Information (PEGI) standard, while for US the rating
format comes from the Entertainment Software Rating Board (ESRB) [26]. Third, ex-
cept for recommending applications on the basis of language and country, Google Play
provides different search results for authorized and non-authorized users. If not logged
in into Google account, a user is suggested a set of applications that differs from the
one an authorized user will get (although the query in both situations is exactly the
same.
Framework elements depending on developers reveal, that they do not use fully all
of framework elements. Name of the app can be maximum 30 characters long, however
average name length is 23 characters. This suggest, that there is still space that can be
used by developers in name element. Subtitle is an optional element in Apple iTunes.
8
Not every app has set subtitle. In sample 41% of apps did not have subtitle. Subtitle can
be maximum 30 characters long and the average length is 26 characters. This suggest,
that if the subtitle is used, developers use it in its maximum capacity. Long description
can be maximum 4000 characters long, however average length is 1338 characters and
the median 1056 characters long for Google Play and average 1755 characters and me-
dian 1534 characters for Apple iTunes. This suggest, that there is still space that can be
used by developers in description element.
Framework elements depending on users reveal that the larger download threshold
is achieved, the more average reviews apps have. It is linear relationship, which is ex-
pected, since more downloads can result in more reviews. However, the average rating
for apps with lowest download threshold is the highest and then for next thresholds is
decreasing. Average rating is decreasing to 4,22 with threshold of 5000 downloads and
then is increasing to threshold of 500 million downloads. Users’ framework is built on
data divided according to download threshold. It shows, that if app is more and more
downloaded, the overall rating from user is growing. Usually developers are taking into
account requests made in reviews and update apps with new and requested features.
5 Conclusion and discussion
In this paper a framework for app store optimization is proposed. The framework is
based on two areas. One area is dependent on developer. The developer sets the initial
setting for each app, which is distributed to the app store. After the app is being distrib-
uted, users’ engagement in app distribution is second area. Users create number of
downloads, reviews and overall rating. A framework is proposed for both leading app
stores: Google Play and Apple iTunes.
Results show that developers are not fully using elements which are dependent on
them. The lowest use is with description, in Google Play, developers on average use
33% of its capacity, and in Apple iTunes developers on average us 41% of its capacity.
Second element which still has some space to use is the name of an app. In both app
stores, on average developers use 76% of its full capacity.
This paper is a first attempt to create framework, which will explain, what elements
are taken into account, when the ranking is created in app stores. Framework is created
on data downloaded from two leading app stores: Google Play and Apple iTunes. This
framework reveals that this ranking is depending on more than dozen factors. Some of
these factors are identified in this paper. Author divided them into two groups, where
one is depending on app developers and the second is depending on the users’ engage-
ment.
Proposed framework has some limitations. First is that the framework is proposed
only on data that is publicly visible and accessible. Perhaps there are some other ele-
ments taken into this framework, not visible for users. This could be number of app
uninstalls from device, number of app removals from the app store. Second, it takes
only data from two app store. There are other app stores like Windows Phone or Black-
Berry World, which were not taken into building this framework. This could future
9
direction of research, to take also data from these stores and enhance proposed frame-
work.
References
1. Zhu, H., Xiong, H., Ge, Y., Chen, E.: Discovery of Ranking Fraud for Mobile
Apps. IEEE Trans. Knowl. Data Eng. 27, 74–87 (2015).
https://doi.org/10.1109/TKDE.2014.2320733.
2. Lee, G., Raghu, T.S.: Determinants of Mobile Apps’ Success: Evidence from
the App Store Market. J. Manag. Inf. Syst. 31, 133–170 (2014).
https://doi.org/10.2753/MIS0742-1222310206.
3. Martin, W., Sarro, F., Harman, M.: Causal impact analysis for app releases in
google play. In: Proceedings of the 2016 24th ACM SIGSOFT International
Symposium on Foundations of Software Engineering - FSE 2016. pp. 435–446.
ACM, New York, USA (2016). https://doi.org/10.1145/2950290.2950320.
4. McIlroy, S., Ali, N., Hassan, A.E.: Fresh apps: an empirical study of frequently-
updated mobile apps in the Google play store. Empir. Softw. Eng. 21, 1346–
1370 (2016). https://doi.org/10.1007/s10664-015-9388-2.
5. Mojica Ruiz, I.J., Nagappan, M., Adams, B., Berger, T., Dienst, S., Hassan,
A.E.: Examining the Rating System Used in Mobile-App Stores. IEEE Softw.
33, 86–92 (2016). https://doi.org/10.1109/MS.2015.56.
6. Shen, S., Lu, X., Hu, Z., Liu, X.: Towards Release Strategy Optimization for
Apps in Google Play. In: Proceedings of the 9th Asia-Pacific Symposium on
Internetware - Internetware’17. pp. 1–10. ACM Press, New York, New York,
USA (2017). https://doi.org/10.1145/3131704.3131710.
7. Böhm, S., Schreiber, S.: Mobile App Marketing : A Conjoint-based Analysis
on the Importance of App Store Elements. 7–14 (2014).
8. Fu, B., Lin, J., Li, L., Faloutsos, C., Hong, J., Sadeh, N.: Why people hate your
app. In: Proceedings of the 19th ACM SIGKDD international conference on
Knowledge discovery and data mining - KDD ’13. p. 1276. ACM Press, New
York, New York, USA (2013). https://doi.org/10.1145/2487575.2488202.
9. Martin, W., Harman, M., Jia, Y., Sarro, F., Zhang, Y.: The App Sampling
Problem for App Store Mining. In: 2015 IEEE/ACM 12th Working Conference
on Mining Software Repositories. pp. 123–133. IEEE (2015).
https://doi.org/10.1109/MSR.2015.19.
10. Pagano, D., Maalej, W.: User feedback in the appstore: An empirical study. In:
2013 21st IEEE International Requirements Engineering Conference (RE). pp.
125–134. IEEE (2013). https://doi.org/10.1109/RE.2013.6636712.
11. Iacob, C., Harrison, R.: Retrieving and analyzing mobile apps feature requests
from online reviews. In: 2013 10th Working Conference on Mining Software
Repositories (MSR). pp. 41–44. IEEE (2013).
https://doi.org/10.1109/MSR.2013.6624001.
12. Lai, Y.-H., Huang, F.-F., Chiou, P.-Y.: Analysis of User Feedback in The
Mobile App Store Using Text Mining: A Case Study of Google Fit. In: 2017
IEEE 8th International Conference on Awareness Science and Technology
10
(ICAST). pp. 50–54 (2017).
13. Chen, N., Lin, J., Hoi, S.C.H., Xiao, X., Zhang, B., Chen, N., Xiao, X., Zhang,
B.: AR-miner: mining informative reviews for developers from mobile app
marketplace. In: Proceedings of the 36th International Conference on Software
Engineering - ICSE 2014. pp. 767–778. ACM Press, New York, New York,
USA (2014). https://doi.org/10.1145/2568225.2568263.
14. Nayebi, M., Cho, H., Ruhe, G.: App store mining is not enough for app
improvement. Empir. Softw. Eng. 23, 2764–2794 (2018).
https://doi.org/10.1007/s10664-018-9601-1.
15. Genc-Nayebi, N., Abran, A.: A systematic literature review: Opinion mining
studies from mobile app store user reviews. J. Syst. Softw. 125, 207–219
(2017). https://doi.org/10.1016/j.jss.2016.11.027.
16. Martin, W., Sarro, F., Jia, Y., Zhang, Y., Harman, M.: A Survey of App Store
Analysis for Software Engineering. IEEE Trans. Softw. Eng. 43, 817–847
(2017). https://doi.org/10.1109/TSE.2016.2630689.
17. Page, L., Brin, S., Motwani, R., Winograd, T.: The PageRank Citation Ranking.
World Wide Web Internet Web Inf. Syst. (1998).
https://doi.org/10.1.1.31.1768.
18. Kleinberg, J.M.: Authoritative sources in a hyperlinked environment. J. ACM.
(1999). https://doi.org/10.1145/324133.324140.
19. Evans, M.P.: Analysing Google rankings through search engine optimization
data. Internet Res. 17, 21–37 (2007).
https://doi.org/10.1108/10662240710730470.
20. Bifet, A., Castillo, C., Chirita, P.-A., Weber, I.: An analysis of factors used in
search engine ranking. In: Proceedings of the 4th International World Wide
Web Conference (2005).
21. Agichtein, E., Brill, E., Dumais, S.: Improving Web Search Ranking by
Incorporating User Behavior Information. ACM SIGIR Forum. 52, 11–18
(2019). https://doi.org/10.1145/3308774.3308778.
22. Egri, G., Bayrak, C.: The role of search engine optimization on keeping the user
on the site. In: Procedia Computer Science (2014).
https://doi.org/10.1016/j.procs.2014.09.102.
23. Dotson, J.P., Fan, R.R., Feit, E.M., Oldham, J.D., Yeh, Y.-H.: Brand Attitudes
and Search Engine Queries. J. Interact. Mark. 37, 105–116 (2017).
https://doi.org/10.1016/j.intmar.2016.10.002.
24. Gyöngyi, Z., Garcia-Molina, H., Pedersen, J.: Combating Web Spam with
TrustRank. In: Proceedings 2004 VLDB Conference. pp. 576–587. Elsevier
(2004). https://doi.org/10.1016/B978-012088469-8.50052-8.
25. Rizun, M., Strzelecki, A.: Knowledge Graph Development for App Store Data
Modeling, arXiv preprint arXiv:1903.07182 (2019).
26. Felini, D.: Beyond Today’s Video Game Rating Systems. Games Cult. 10, 106–
122 (2015). https://doi.org/10.1177/1555412014560192.
