Cost–Accuracy Trade-Offs in Unpaved Rural Road Condition Assessment: Visual Indices and Smartphone-Based IRI in the Ecuadorian Andes

More than 90% of the road network of Loja, Ecuador, is unpaved and lacks an official condition metric. Low-cost measures relate to the IRI on pavements, but not demonstrably on unpaved roads. A 15 km Andean corridor was surveyed in one dry season with four visual–manual indices (URCI, MTC, PASER, ICNP), a Class III profilometer, and an uncalibrated smartphone application, over 150 stations harmonized from 60 native index observations, 30 for the MTC. A single profilometer run was moderately reliable (ICC = 0.631, 95% CI 0.559–0.699); four runs averaged 0.872, so repeated runs are a quantified requirement, not a convention. The application erred in level, not in spatial pattern (standardized ICC = 0.900), placing 88.7% of stations within one quartile of the reference: its role is segment-scale screening, not absolute IRI. Under leave-one-block-out spatial validation, the index–IRI equations retain an R2 of 0.32–0.40; MTC was the only index whose association depended on the measuring instrument (p = 0.010). Surveying on foot cost 4.45–12.36 USD/km against 23.68 for reference profilometry, whose largest item is the vehicle. A tiered cost–accuracy architecture is proposed, with URCI and ICNP as preliminary candidates for official adoption, subject to validation based on further corridors and in wet condition.