Maintenance spending is the wrong yardstick: a pre-registered negative result for a satellite measure of road condition across 850 Nigerian highway corridors

Can free satellite imagery tell a road agency where maintenance money should go? This paper reports a pre-registered test of that question on 850 named Nigerian federal highway corridors, 18,419 km over 22 months at 10 m, using a verge-minus-road vegetation contrast corrected for corridor level and season, and validated against whether a maintenance payment appears in the open procurement record. The bar and the direction were fixed before the run: an area under the curve of at least 0.60 with the lower bound above 0.50, and corridors with a recorded payment scoring as more degraded. The result was 0.3901, interval 0.3196 to 0.4671, wholly below 0.50. The ranking is inverted: funded corridors score as less degraded than matched unfunded ones, and a hundred label permutations place the observed value outside the null. The explanation offered, and not proved, is that the label is the problem: a procurement record evidences money allocated, not road failure. The durable contribution is methodological — a doubly-centred score has no level, and a first attempt that averaged one over the same corridor's months ranked missingness rather than condition. Free 10 m imagery is not the obstacle, since published work classifies roughness from it nationally. The paper leaves behind a complete open measurement, a pipeline that runs at zero cost, and a method that caught its own error.