<p>Time series forecasting is a challenging domain and data limitations further add up to the challenges. The boom in generative AI has led to the emergence of powerful foundation models with extensive prior training, allowing them to achieve remarkable performance across a wide range of text, image, and video processing tasks. Time series foundational models which are pre-trained on time series data, face unique challenge with data availability from different domains during their training process and thus their efficiency may vary with domain variance. Indian card industry on the other hand is a steep challenging domain in terms of prediction due to the influence of various factors and data limitations, making forecasting all the more difficult in this domain, while a powerful model in this domain would benefit a lot of associated stakeholders including policymakers, economic watchdogs, investors etc. Our research paper investigates the efficacy of various Time Series Foundation Models (TSFMs), comparing them against each other, a benchmark, and a set of non-foundation models in the data-limited environment of the Indian card industry. It also tries to infer if further fine-tuning of time series foundational models can lead to performance gains. Our work presents a wide gamut of metrics of Indian card industry and performs a comprehensive performance evaluation which convincingly establishes performance advantages for usage of TSFMs vis-à-vis other non-foundational models in performing time series forecasting.</p>