Determining Flight Costs with Storm and Hadoop (and YARN!)
Airline pricing has always been a mystery to me, a combination of art and science allowing the airline to make as much money as possible on each flight while providing the customer the options and flexibility they want. Under the covers I know there are complex models the airlines use to determine how many seats have been sold and how much they can get for the remaining seats. I didn’t realize how seriously complex the models were but more importantly, the opportunity available to the travel industry to become more customer-centric while staying competitive by harnessing the data now available to them.
Hortonworks systems integration partner Pactera has delivered a big data solution for a customer in the travel industry and anticipates the successful model will only get better.
The savvy consumer is constantly comparing and contrasting the options before they buy, looking for the best value and experience through promotions, asking friends, checking the social media sites for pros and cons of the offers available. All of this help the buyer make their final decision and often it is completed in one or two online sessions.
Meanwhile, behind the scenes, the buyer’s web activity is captured, including what web sites were visited and time spent on each page. Combine that customer behavior with competitive offers from other travel companies, statistics from business partners on activity like destination popularity, market pricing, past customer behaviors, flight times and flight availability (and there’s much more). All of these are inputs for the specially designed, in-memory solution developed by Pactera to deliver a flight price to the online buyer. The engine has to not only offer alternatives to the buyer while they are online, but has to deliver increased or decreased prices to other buyers viewing the same flights.
Since the inputs are continuously changing, the engine is constantly evaluating all the elements to create the most up-to-date flight price possible. The model also compares the price to “old” predictive algorithms to make sure the fare won’t result in a loss of revenue.
The results are fed back into Hadoop and RDBMS and made available to other areas of the business such as mobile applications as well as many reports and visualizations needed by marketing, finance, scientists, and clients.
Looking forward to YARN
The engine is made up of Storm for real time processing – not back to the consumer, but for downstream destinations. While this is solution is a successful use-case, Pactera is already planning to implement Hadoop 2.0 with YARN to create a more efficient engine.
Pactera: BI experts
Pactera is a leading provider of Business Intelligence, Solution Architecture, and enterprise performance management services. They have successfully provided BI and data warehouse solutions for many Fortune 500 and other prominent companies. Using global project management and quality control methods Pactera delivers timely and reliable services, professional project management, project execution, and business process optimization.
Applying advanced information technology to the entire enterprise not only enables the enterprise to obtain timely information, but also leverage the information for competitive advantage. Pactera helps customers build a business analysis platform that enhances business growth by providing information for decision-making, and ultimately leading to increased business value.
Webinar overview: Customer insight and marketplace predictions are a few of the profitable benefits found in big data technology. Leading companies are using the advanced analytics solution to find new revenue streams, increase customer satisfaction and optimize the supply chain.