Stretchable strain sensors, which convert mechanical stimuli into electrical signals, largely fuel the growth of wearable bioelectronics due to the ubiquitous, health-related strain in biological systems. In contrast to rigid conventional strain sensors, stretchable strain sensors present advantages of conformality and stretchability, solving the mechanical mismatch between electronics and the human body. However, the great challenge of stretchable strain sensors lies in achieving high sensitivity, which is required for both signal fidelity and cost considerations. Recent advances to solve this sensitivity challenge have focused on material optimization, in search of the optimum combination of conductive active materials and elastomer substrates among a myriad of artificial or natural materials. However, high sensitivity with a gauge factor larger than 50 remains a grand challenge, especially within large-strain regions. Here we present heterogeneous strain distribution of elastomer substrates as a powerful strategy to significantly enhance the sensitivity of stretchable strain sensors. The theoretical foundation of this strategy is mathematically proven on the basis of Ohm's law in electrics and mechanics of materials. First, the extent of the sensitivity enhancement is proved to be determined by the local strain in resistance-testing segments of heterogeneous strain sensors. Next, the local strain is proved to be quantitatively decided by material properties such as section area and Young's modulus. Thus, the necessary and sufficient condition to achieve high sensitivity in heterogeneous strain sensors is that the Young's modulus reciprocal or section area reciprocal in the resistance-testing segment is larger than the mean value. This provides a theoretical design guideline to achieve high sensitivity via heterogeneous strain distribution. On the basis of this guideline, we systematically summarize concrete instances of heterogeneity-induced sensitivity improvement in stretchable strain sensors, in sequence of increasing dimensionality. A typical example of a one-dimensional heterogeneous strain sensor is a structured fiber with microbeads, where the varied section area along the fiber axis results in heterogeneous strain and sensitivity improvement. Two-dimensional heterogeneous sensors in the form of thin films contain thickness gradient sensors and auxetic mechanical metamaterial sensors. The former exhibit heterogeneous section area via the self-pinning method, while the latter show heterogeneity in both the strain direction and amplitude, leading to a 24-fold improvement in sensitivity. Three-dimensional strain sensors include rationally structured sensors for out-of-plane force detection and asymmetric active materials in electronic whiskers. The resultant enhanced sensitivity in these heterogeneous strain sensors is beneficial for applications such as continuous health monitoring, biomedical diagnostics, and replacement prosthetics, taking advantage of augmented detection accuracy and declined device cost. Finally, we discuss possible future work in exploiting heterogeneous strain distributions, involving extended methodology to achieve heterogeneity, employing suppressed strain for stretchable electrodes, cyclic durability for long-term applications, and multifunctional system-level integration. We believe that this strategy of using heterogeneous strain distribution to enhance sensitivity can strongly promote the development of stretchable strain sensors for both practical and theoretical requirements.