Airline Food Wastage
DATEFeb 8, 2019
Technology UsedAmazon AWS, QuBole, Java, Python, Jupyter, Talend, Apache Spark
A Europe based travel giant handles airline and cruise liner travels. It faced a major challenge of revenue loss due to overstocking of food inventory resulting in a wastage of the perishable items. They needed assistance to help them optimize food stocks so that there is no wastage of food on the carrier. Along with this, they also wanted to ensure that sufficient food is available as per the demand without getting into the issues of under-stocking of food supplies on the flight.
Since the manual process was prone to errors, they needed a tech solution that can automatically recommend the right inventory of perishable food on a given flight path without resulting in a shortage of food. Since this resulted in cost overheads in an extremely competitive industry, the client needed a comprehensive solution to address this issue and come up with a tech-backed solution without the need for manual intervention.
At AdFolks we first started out by accumulating historical data about food consumption and disposal patterns across different routes and travel segments. The data came from multiple sources that gave a lot of information around the passengers’ food consumption patterns. The privately available data for airlines gave tremendous information on the type of food, seasonal trends, consumption levels and level of wastage. This was augmented by the inventory data at any given time or given flight route.
After requirement gathering and assessment of various possibilities, we narrowed down to the use of data science to optimize food disposal, eliminate inventory shortage, and generate an automated process that showed how much food would need to be stocked depending on parameters like sales data and route data.
In order to leverage the predictive modeling potential of data science we first accumulated the data and created a data lake. The data was in the form of tables with flight data, barset info, sales data, and route data loaded on to Qubole cloud native data platform.
Now we had the necessary historical data on which we can apply algorithmic programming. We used Jupyter Notebook to program a cost function. The goal for designing this cost function was clear – keep it as low as possible to reduce the food wastage happening on the airline carriers. The cost function accounted for the cost of the product, disposal, storage, insurance, missed opportunity and impact on reputation.
The cost function helped us design the probability density function. This function analyzed the historical data based on time, seasonality etc. for each product.
The solution enables the airline to look at historical trends into food consumption and disposal. This is the foundation of the data science solution that delivered unmatched advantages as below
1 – Better insight to optimize food consumption and disposal
Our data lake correlates a huge array of information like flight data, barset info, sales data, seasonal demand drivers, segments data, and route data with the amount of food consumed or thrown away. We are able to analyze at a granular level in order to provide correct insight on each type of food product served during the flight.
2 - Automated Inventory management framework
Airlines are now able to quickly and accurately determine the amount to be stocked for different product categories taking into consideration storage cost, perishable item cost, opportunity cost, credit line cost etc). The framework also factors in the need to avoid additional bottlenecks. It suggests the amount of inventory under each product for a given set of total budget and total storage space. It also factors in additional constraints on cost of perishable items as well as lost revenue due to unavailability of stock.
3 – Quicker decision making
The data science solution provides airlines the ability to take data driven decision under uncertainty, with an impressive degree of accuracy and without bias.