This is an old revision of the document!


A large-scale comparative study of beta testers and standard users [CACM 2018]

The study of online security software explains the importance of knowing who the beta testers are.

  • Authors: Vlasta Stavova, Lenka Dedkova, Martin Ukrop and Vashek Matyas
  • Primary contact: Vlasta Stavova vlasta.stavova@mail.muni.cz
  • Journal: Communications of the ACM, issue XX
  • Author pre-print: To be added later
  • Citation: BibTeX

  • Use data you can collect to find out who your users and beta testers are. The fewer testers you have, the pickier you should be about their selection.
  • Testers should be representative of standard users. Keep checking that this is the case – or remediate with additional analyses and more carefully reached conclusions.

Beta testers are the first end users outside a software company to use its product. They have been used for decades and are rightly credited not only with finding and reporting bugs, but also with improving general product usability through their feedback and/or the ways they use the product.

In this paper, we investigate whether beta testers represent standard users well enough to allow for the extrapolation of testing data to standard users. We have investigated records of beta testers and standard users of home security solution developed by the IT security software provider ESET. With more than 600 000 participants from more than 180 countries, we present what we believe to be the first large-scale comparison between standard users and beta testers.

We compared several aspects of both populations, such as hardware, operating system, country of origin and EULA reading time, all taken from system data. Other attributes, such as age, gender, privacy perception and computer proficiency self-evaluation, were available thanks to a user questionnaire.

We conclude that - at least in our study - beta users represent standard users well in terms of hardware and operating system in large scale beta testing. However, populations differ significantly in the distribution of users and testers between countries. This may cause a problem when a testing includes localization and usability issues that may be influenced by regional differences.

  • Diploma thesis based on localization data from this project (in czech only). It compares free available online GeoIP databases on a large sample of available IP adresses.

We thank Masaryk University (MUNI/M/1052/2013) and Miroslav Bartosek for support, and to anonymous reviewers and Vit Bukac for valuable feedback.