How Gamification Affects Physical Activity: Large-scale Analysis of Walking Challenges in a Mobile Application

Ali Shameli, Tim Althoff, Amin Saberi, Jure Leskovec

Abstract

Physical inactivity is a major global health challenge. This paper presents a large-scale quantitative study of how mobile walking competitions influence physical activity, analyzing over 800,000 person-days of objective tracking data from the Azumio Argus app. We find that competitions boost average physical activity by 23%, with the largest gains observed in previously inactive users. We also demonstrate that the "fairness" of participant matching and gender diversity are critical drivers of engagement.

Introduction

Physical inactivity contributes to millions of deaths and significant economic costs annually. While gamification—the use of game-design elements in non-game contexts—is a popular strategy to encourage exercise, there is a lack of large-scale, objective data on how specific mechanics like head-to-head competitions shape behavior. Most existing research relies on small-scale studies or biased self-reports. This study addresses this gap by using objective accelerometer data to determine the efficacy of competitions and the factors that drive their success.

Methods and Dataset

The researchers analyzed data from the Azumio Argus smartphone app, focusing on 7-day walking challenges. The dataset includes:

  • Scale: 2,432 seven-day competitions involving 3,637 unique users.
  • Volume: Over 800,000 person-days of activity tracking and 535 million total steps.
  • Demographics: Median age 34, 51% female, including a significant range of BMI categories.

The study employed a within-subject comparative analysis, measuring user activity during competitions against their own individual "baseline" activity. Additionally, the authors built predictive models (Logistic Regression and Gradient Boosted Trees) to classify competitions based on total activity increase ($\Delta$ACTIVITY), the final gap between participants (FIRST-LAST), and the frequency of rank changes (RANK SWAPS).

Key Results: The Impact of Competition

The primary finding is that participation in a competition leads to a significant increase in physical activity. On average, users increased their daily step count by 23% during a challenge compared to their baseline.

Average daily steps significantly increase during competition periods compared to baseline activity.

Winning a typical 7-day challenge requires approximately 25% more steps per day than one's baseline. The predictive models achieved an ROC AUC of 0.72 for predicting activity increases, suggesting that the success of a competition can be forecasted using pre-competition data.

Insights into Engagement and Demographics

The study revealed several nuanced drivers of physical activity:

  1. The "Inactive" Advantage: Contrary to expectations, the largest behavioral changes were not among fitness enthusiasts. Users who were previously the least active saw the largest absolute gains, increasing their daily count by more than 2,500 steps.
  2. The "Fairness" Threshold: Competition quality is highly sensitive to matchmaking. If the baseline daily step difference between participants exceeds ~5,000 steps, the frequency of leaderboard "swaps" drops, and the overall activity boost fades.
  3. Demographic Diversity: Mixed-gender competitions resulted in higher overall activity levels than single-gender groups, despite men being generally more active than women in the dataset.
Activity increases broken down by (a) Gender, (b) Age, (c) BMI, and (d) Previous Activity levels.

Discussion and Limitations

While the results are promising, the study acknowledges a slight selection bias, as users who join competitions are already slightly more active than those who do not. Furthermore, the focus on a 7-day window leaves questions about long-term habit formation unanswered. However, the findings suggest that algorithmic matchmaking could be leveraged to pair users in a way that maximizes "close" and high-activity races, potentially turning short-term gamified bursts into sustainable health improvements.

Summary updated

BibTeX
@inproceedings{shameli2017how,
  author = {Ali Shameli and Tim Althoff and Amin Saberi and Jure Leskovec},
  title = {{How Gamification Affects Physical Activity: Large-scale Analysis of Walking Challenges in a Mobile Application}},
  booktitle = {The Web Conference},
  year = {2017},
  url = {https://openreview.net/forum?id=rJNUGZWuWr},
}
Figure