Animal ecologists often use animal sign as a surrogate for direct observation of organisms, especially when species are secretive or difficult to observe. Spatial heterogeneity in arid environments makes it challenging to consistently detect and precisely characterize animal sign, which can bias estimates of animal abundance or habitat use. Piute ground squirrels (Urocitellus mollis) and Owyhee harvester ants (Pogonomyrmex salinus) live in arid environments and are fossorial, which can make them difficult to observe directly. Their relative abundance can be assessed using sign (i.e., burrows and nests). We implemented an over-sampling framework (i.e., recorded an excessive amount of information) with two observers to 1) identify a sampling intensity that balanced precision with our resource constraints, and 2) assess classification and detection of squirrel burrows and ant nests across vegetation conditions. We sampled 20 1-ha plots for ground squirrel burrows and ant nests using six 4 m × 100 m belt transects. Analyses of precision and sampling effort indicated that three belt transects covering 1200 m² per ha provided sufficient precision, while minimizing effort. Regardless of vegetation conditions, counts by two observers were strongly correlated for ground squirrel burrows (r = 0.99, P < 0.001, df = 18; slope = 0.92) and harvester ant nests (r = 0.99, P < 0.001, df = 18; slope = 1.01) indicating observer consistency and perhaps high detection probability. These findings illustrate an approach for evaluating sampling designs in many ecological contexts.