Analysing Tea Auction Trends for Beneficial Seasonal Tea Production in Sri Lanka

Abstract

Tea is an age old industry in Sri Lanka which has been highly regulated via Ceylon Tea Board. Green Leaves are plucked from different parts of the country named after ‘Elevations’ by the geographical height form the sea level, categorized into High, Medium and Low. Consumable tea is then manufactured by factories, labelled by its ‘Factory Code’ and a ‘Selling Mark’. Colour and appearance of tea could be identified by the ‘Tea Grade’ of manufactured tea; it does not indicate the purity of tea. Quality of tea exposed to the market is guaranteed by an auctioning process which takes place weekly in the heart of Colombo. Eight registered tea brokers facilitate the tea auction and about 15000 tea lots are weekly being auctioned by the end of year 2013. Factory produced tea is guaranteed to be tested by the brokers in this process and samples re-tested by potential buyers before the auction. The auctioning process does not guarantee a fixed price for a given tea lot for the factory. Brokers drive their business from the brokerage they earn from successfully selling each tea lot at the auction. Better prices for tea guarantee a better brokerage due a percentage of proceeds being earned by the broker. Brokers inevitably have long established relationships with factories to mediate the auction process. Hence, they naturally perform the role of advisors to the factories in terms of predicting demand of different varieties of tea. However these predictions are not fool proof and is run by the instinct of the brokers. To make it more complex, there is a gap of 3 to 4 weeks until manufactured tea is actually auctioned due to a long cataloguing, sampling and testing process. Though abstract level statistics are obtained and published, Sri Lankan tea industry has not gone to the extent of analysing past auction data to determine seasonal demands for different varieties of tea. This research is focused on extracting seasonal demands from tea auction history records to better facilitate the brokers predict demand of upcoming sales. Initial analysis on the dataset provides better insights to information. This knowledge, combined with the domain knowledge of the industry is then applied on the sample dataset using association rule mining technique. Outcomes reveal information hidden in the tea auction dataset which could not be extracted using basic analysis. Outcome of this research display promising results. Findings suggest that manufacturing factories display similarities in some sale months in terms of the price range of the auctioned teas providing clues to possible seasonal demands in the dataset. Most of all, it unveils the tea auction dataset for future possibilities of further pattern discovery.