Zero-Shot Stance Detection (ZSSD) identifies whether a text supports, opposes, or is neutral towards an unseen topic. While existing methods like the teacher-student framework improve generalisation, they are often resource-heavy. A simplified, lightweight and efficient data augmentation approach for ZSSD is proposed in this work, named as Direct Stance Assignment (DSA). DSA augments the training data through stance label assignment based on cosine similarity in a shared semantic space. Further, a classifier that integrates gMLP blocks into BART encoder is developed to leverage improved token interaction, leading to the proposed methodology, gDSA = gMLP + DSA. This reduces training time and parameters by nearly half of the SOTA models, while maintaining the similar performance. We also propose a much simpler data augmentation technique: Target-Aware Polarity Swapping (TAPS), that does not need the expensive keyphrase generation. Further, the role of commonsense is investigated in the teacher-student setting to evaluate the importance of external knowledge in the framework. Experiments on benchmark data show that the proposed method performs close to state-of-the-art models, with significantly lower computational cost.