Abstract

High-frequency mobile enterprise applications — field-service management, claims processing, retail operations, inspection workflows — place state management under pressure that desktop and web equivalents rarely face. Devices operate on flaky networks, in moving vehicles, in basements, behind firewalls, and across time zones; users capture data in bursts and then disappear for hours; business rules demand that a record updated at 09:14 in one office must be visible at 09:14 in another without round-trip coordination overhead. We argue that the conventional “server holds truth, client renders it” model is structurally inadequate for these workloads, and that the right level of abstraction is the workflow instance, not the database row. The article makes three contributions. First, it defines five canonical states that every mobile workflow instance passes through on a device — Unknown, Prefetched, In-Flight, Divergent, Reconciled — and shows that most production failures are failures of state modelling rather than failures of networking code. Second, it surveys four architectural patterns for managing state across device and server — Pull-Only with optimistic local state, Push-Only with event-sourced log, Hybrid with materialized view and reconciliation queue, Edge-Coordinated with peer-to-peer state — and compares them against online staleness, offline support, conflict handling behaviour, and operational cost. Third, it presents a decision framework that maps workload properties to architecture choice, draws trade-off boundaries explicitly, and closes with a worked scenario from a 600-technician field-service deployment. The intended reader is the enterprise architect who must decide whether a mobile application needs a custom state platform or can be served by an off-the-shelf mobile BaaS, and the engineering lead who must choose between consistency, completeness, and the operational budget required to maintain each