Data-centric prognostics use historical and current data statistically and probabilistically to predict an asset's remaining useful life. In the era of Industry 4.0, technological advances in new sensors and rapid advances in computing offer an obvious motivation to develop and apply data-centric prognostics. The present paper endeavours to thoroughly examine the challenges at hand. The exclusive challenges addressed encompass explainability, prescriptive analytics, and affordability in data-centric prognostics. For instance, the lack of transparency in data-centric prognostics undermines trust and accountability, hindering informed decision-making and potentially leading to misguided actions. Likewise, developing robust methodologies that predict outcomes and provide actionable insights and recommendations for optimal decision-making is critical. Lastly, demand for affordable solutions can enhance SMEs’ operational efficiency and competitiveness. Moreover, common challenges discussed include the availability of sufficient data, uncertainty, time-variant life cycle loading, etc. Another main contribution of this paper is that it offers insightful guidance on future directions. Such as the prospect of federated learning to offset the drawbacks of a data-centric approach is deliberated. Two other promising directions for future research include prototype-based models and Turing's type-B random machines. The highlighted challenges and opportunities lead to a better understanding for the researcher on further developing and successfully implementing data-centric prognostics in Industry 4.0 systems.