Although we live in an information overwhelmed era, in many applications it is still difficult to collect meaningful data due to data scarcity issues, time constraints and the cost in getting the data available. In such scenarios, we need to make better use of the scarce data available so that it can be utilized for performing further analysis. Existing approaches use available data for performing data analytics only if the estimation accuracy of the whole dataset satisfies a defined threshold. However, this approach is not beneficial when the data is scarce and the overall estimation accuracy is below the given threshold. To address this issue, we develop a heuristic approach for getting the most benefit out of the available data. We classify the existing data into classes of different errors and identify the usable data from the available data so it can be used by decision makers for performing further data analytics.