Modeling Failure Dynamics in Time-Constrained Authentication Systems: Evidence of a Success Cliff in USSD Workflows

USSD-based financial services are vital in regions with limited smartphone access, yet they operate under strict session timers and sequential interaction requirements. While adding authentication steps (like SMS OTPs or multi-step PINs) improves security, it also increases the time burden on the user. This paper models the failure dynamics of these systems to determine when increased security measures render a transaction operationally unreliable.

The authors introduce the concept of the Success Cliff to describe a non-linear tipping point in system performance. Unlike gradual degradation, where each added security step causes a small, predictable drop in success, the Success Cliff represents a threshold where session completion rates plummet. The study demonstrates that this collapse occurs when the cumulative time required for user input, network round-trip delays, and external blocking delays (like waiting for an SMS) exceeds the system's fixed session timeout.

The researchers developed a simulation-based framework using a Keystroke-Level Model (KLM) to estimate user interaction times. They modeled three distinct abandonment behaviors—time-based, event-based, and combined—to account for how users might disengage from a session before the system officially times out. By running 50,000 simulations across varying network conditions and authentication configurations, the authors quantified the trade-off between security and usability.

As financial institutions move toward stronger authentication to combat fraud, they risk inadvertently locking out users in resource-constrained environments. This research provides a formal method for designers to evaluate whether a proposed authentication workflow will remain viable under real-world network conditions. It highlights that out-of-band delays, such as SMS delivery, are often the primary culprits behind session failure, suggesting that designers should prioritize low-latency authentication methods to avoid hitting the Success Cliff.