Multi-scale architectures have shown effectiveness in a variety of tasks including single image denoising, thanks to appealing cross-scale complementarity. However, existing methods treat different scale features equally without considering their scale-specific characteristics, i.e., the within-scale characteristics are ignored. In this paper, we reveal this missing piece for multi-scale architecture design and accordingly propose a novel Multi-Scale Adaptive Network (MSANet) for single image denoising. To be specific, MSANet simultaneously embraces the within-scale characteristics and the cross-scale complementarity thanks to three novel neural blocks, i.e., adaptive feature block (AFeB), adaptive multi-scale block (AMB), and adaptive fusion block (AFuB). In brief, AFeB is designed to adaptively select details and filter noises, which is highly expected for fine-grained features. AMB could enlarge the receptive field and aggregate the multi-scale information, which is designed to satisfy the demands of both fine- and coarse-grained features. AFuB devotes to adaptively sampling and transferring the features from one scale to another scale, which is used to fuse the features with varying characteristics from coarse to fine. Extensive experiments on both three real and six synthetic noisy image datasets show the superiority of MSANet compared with 12 methods.