A Study on Interaction Complexity and Time
Leonardo Germán Loza Bonora, Julián Grigera, Helmut Degen
TL;DR
This study investigates how the Big I notation for interaction complexity can be linked to execution time by deriving average per-step times from a controlled user study (n=100) of a movie-ticket Booking UI with two designs. By defining a consistent step-based model and calculating both quadratic and linear growth in interaction steps, the authors estimate a mean interaction speed around 1.05 IS/sec and relate Big I to practical time estimates via KLM benchmarks. The results demonstrate a path to translating early design complexity into time forecasts, revealing learning effects and cognitive-load differences across UI variants, and highlighting limits and future work for task-type generalization. The work provides a quantitative bridge between early UX concepts and time-based performance metrics, enabling more informed design choices in the lo-fi to production transition.
Abstract
Testing Web User Interfaces (UIs) requires considerable time and effort and resources, most notably participants for user testing. Additionally, the tests results may demand adjustments on the UI, taking further resources and testing. Early tests can make this process less costly with the help of low fidelity prototypes, but it is difficult to conduct user tests on them, and recruiting participants is still necessary. To tackle this issue, there are tools that can predict UI aspects like interaction time, as the well-known KLM model. Another aspect that can be predicted is complexity, and this was achieved by the Big I notation, which can be applied to early UX concepts like lo-fi wireframes. Big I assists developers in estimating the interaction complexity, specified as a function of user steps, which are composed of abstracted user actions. Interaction complexity is expressed in mathematical terms, making the comparison of interaction complexities for various UX concepts easy. However, big I is not able to predict execution time for user actions, which would be very helpful for early assessment of lo-fi prototypes. To address this shortcoming, in this paper we present a study in which we took measurements from real users (n=100) completing tasks in a fictitious website, in order to derive average times per interaction step. Using these results, we were able to study the relationship between interaction complexity and time and ultimately complement big I predictions with time estimates.