Chatbot Usability and Responsiveness as Proposed Digital Financial Literacy Proxies: Customer Satisfaction and Purchase Decision Evidence from Rural Indonesian E-Commerce

Authors

  • Shinta Maharani Trivena Politeknik Negeri Malang, Indonesia Author
  • Amma Fazizah Universitas Yudharta Pasuruan, Indonesia Author
  • Nurul Afiqah Binti Haji Amin University of Brunei Darussalam, Brunei Darussalam Author
  • Rena Feri Wijayanti Politeknik Negeri Malang, Indonesia Author
  • Saiful Ghozi Politeknik Negeri Balikpapan, Indonesia Author

DOI:

https://doi.org/10.33795/7ma01f82

Keywords:

chatbot usability, chatbot responsiveness, digital financial literacy, ceiling effect, robustness analysis, rural Indonesia

Abstract

This paper examines how chatbot service quality relates to customer satisfaction and purchase decisions among rural e-commerce users in a low digital financial literacy (DFL) context, and proposes the chatbot-as-DFL-proxy concept. A structured questionnaire yielded responses from 150 Shopee users in Tutur District, East Java. Ordinary least squares path analysis produced significant effects of usability (β = 0.284, p = 0.003) and responsiveness (β = 0.321, p = 0.001) on satisfaction (R² = 0.309), and of usability (β = 0.379, p < 0.001) and responsiveness (β = 0.240, p = 0.011) on purchase decision (R² = 0.378), but no effect of satisfaction on purchase decision (β = 0.075, p = 0.338). Mediation was therefore not supported for either path (Sobel z = 0.92 in both cases; bootstrap intervals include zero). Robustness testing revealed a severe ceiling effect: 96.7% of respondents scored within a narrow high band, at a per-item mean of 4.17 of 5.00. Rank correlations were near zero for four of five paths, two respondents were highly influential, and excluding ten flagged cases (7% of the sample) removed every effect. The associations therefore depend on a small group of low raters and should be treated as fragile. Because DFL was not measured at the individual level, the proxy claim remains a theoretical proposition that still needs validation. The study offers a methodological caution for measuring satisfaction in homogeneous populations and recommends response formats that discriminate better.

Downloads

Download data is not yet available.

Downloads

Published

2026-10-31