South Africa's R200 billion road-maintenance backlog is not primarily a funding problem. It is a data problem. No municipality in the country has a current, network-wide condition map. The technology to generate one costs less than a smartphone and a windshield mount, and is being deployed on fleet vehicles that already drive the network every day.
In the weeks since Helen Zille's viral swim through a Douglasdale pothole, South Africa's road crisis has consumed the national conversation. Opposition parties are staging pothole-filling photo ops. Operation Fix Joburg is a weekend activity. The ANC is deploying Jetpatcher technology in Tshwane. Every party, it seems, has a pothole strategy.
What none of them has, according to the engineers actually running municipal road programs, is a pothole map.
The paradox of pothole politics
South Africa's relationship with its roads has always been political. Road quality is one of the most visible indicators of governance. When Cape Town points to its 72-hour repair turnaround, or when commentators mock the "Pothole Alliance," they are weaponising a failure every voter can see and feel.
But the political discourse obscures a more structural problem. South Africa's road-management systems, across every party and every tier of government, remain overwhelmingly manual. Visual inspections conducted by engineers driving routes at low speed. Complaints logged through call centres and social media. Repair schedules built on intuition and political pressure rather than systematic assessment.
The Johannesburg Roads Agency illustrates the consequences. JRA receives roughly 3,900 pothole reports per month, yet between July and December 2025 it logged only 23,572 verified reports, suggesting that even its own complaint system captures barely more than the surface. With the agency now insolvent after a R300 million budget shortfall, and an estimated R81.5 billion needed over the next decade, the prospect of managing repairs without comprehensive network data seems almost absurd.
Enter the smartphone
The underlying proposition is simple. Mount an Android smartphone on the windshield of a vehicle that already drives the road network: a municipal refuse truck, a water-services van, a traffic-enforcement vehicle. The phone's camera and inertial sensors capture continuous data. A cloud pipeline running five production vision models, trained on more than 74,000 annotated images, processes the data and returns a comprehensive condition assessment.
The system doesn't just find potholes. It classifies them by severity. It detects alligator cracking, longitudinal and transverse cracks, wheel-track deterioration, ravelling and subsidence. It identifies road assets, including signs, markings, guardrails and manholes, and rates sign condition on a 1 to 5 scale. It generates standardised Pavement Condition Index (PCI) scores aligned to ASTM D6433 and International Roughness Index (IRI) estimates banded to World Bank thresholds.
Every detection is pinned to GPS coordinates, timestamped, and accompanied by photographic evidence, creating what amounts to a living, continuously updated digital twin of the road network.
The economics of ignorance
The financial argument for continuous monitoring is compelling even before accounting for lives saved.
Traditional road-condition surveys using specialised vehicles and trained assessors cost between R1,400 and R5,200 per kilometre. At those rates, a comprehensive survey of Johannesburg's road network is a multi-hundred-million-rand exercise, which is precisely why it rarely happens. Most municipalities survey their networks every three to five years at best, meaning by the time data informs a maintenance plan, the conditions have already changed.
Smartphone-based platforms operate at R140 to R700 per kilometre per year, enabling continuous rather than periodic assessment. For context, Santam reported that pothole-related insurance claims surged by 40%, with average payouts of R20,000 to R25,000 per claim. A municipality that spends R500 per kilometre annually on AI-powered monitoring to prevent even a modest percentage of those claims, and the litigation, vehicle damage and human injury behind them, achieves immediate return on investment.
The irony is that the municipalities with the smallest budgets are the ones who most need this data. When you only have R149 million for road resurfacing and you need R81.5 billion worth of work, every rand has to go to the highest-priority defect. Without data, you're guessing. With continuous, calibrated condition scores, you're making evidence-based decisions.
Beyond the pothole: a platform approach
What makes a modern road-intelligence platform more than a detection tool is the full workflow. Detected defects flow into a work-order system where maintenance teams receive assignments, priorities and deadlines. A planning layer enables budget forecasting and deterioration modelling, allowing engineers to project where failures will occur and plan pre-emptive interventions.
The platform exports data in GIS-standard formats (Shapefile, GeoJSON, KML, CSV), making it immediately compatible with the ArcGIS and QGIS systems that most engineering departments already use. A read-only REST API enables integration with existing municipal asset-management platforms.
Detection without action is just surveillance. The point isn't to build a catalogue of every pothole in South Africa. The point is to give engineers and decision-makers a tool that connects what's happening on the road to what needs to happen next: the work order, the budget allocation, the five-year maintenance plan.
A message to municipal leaders
Every party is promising to fix potholes. The DA says 72 hours. ActionSA is crowdsourcing locations. The ANC is deploying Jetpatchers. These are all fine responses. But they're all reactive. They all start with the pothole already existing and someone already complaining about it.
The paradigm shift is proactive maintenance: detecting deterioration before it becomes a pothole, prioritising roads on objective condition data rather than volume of complaints on social media, and planning budgets on engineering projections rather than political pressure.