Big data initiatives are launched with the goal of obtaining valuable insights that can boost business performance, yet organizations struggle to understand the different capabilities of analytics that they can leverage, and when to use them. The progression in using applied business analytics has been diagrammed as follows. They begin with the collection of basic data and standard reporting, and move through various levels of sophistication through to optimization programs.
Conceptually, each level of sophistication builds on the previous. Airlines, like all data-oriented organizations, have likewise moved from the “basic data reporting” stage through various stages to get to the level “optimization,” which is the deemed the most sophisticated. Thus, airline revenue management is a great example of this highest tier of big data analytics as it optimizes seat allocation across fares and passenger itineraries on a daily basis. In fact, airlines have been at the highest tier – optimization” – for decades.
Despite this, airline revenue management has actually continued to evolve, even after reaching this level since ‘80’s, and there continue to be even more opportunities for improvement. Based on the evolution of airline revenue management over the past decades – and in place today – there are five major areas where any airline applying big data analytics can soar past the “highest tier”.
1) Models and model assumptions need to be continually challenged
Firstly, all models are just that, models. They represent simplifications of the world, hopefully focused on the most important factors. But they all are dependent on certain basic assumptions. For airlines, the initial assumptions were, for example, that connect traffic could be treated generically (aggregate “flow”) for the purposes of optimization (the leg-based revenue management model). Only later did models add the granularity of different O&’s (the O&D model). Also of course, the traditional airline revenue optimization model assumes fares are categorized properly in “buckets,” an assumption that continues to be refined.
2) The bigger the data pool, the better the information gleaned through analytics
Airline revenue management introduced links to competitive fare availability tracking in the past decade or so, whereby fares are automatically adjusted when competitive airlines increase low fare availability. Airlines are working to incorporate various qualitative factors such as social media mentions into their historically quantitative databases.
3) Forecasts can continually improve as big data enables new refinements
“Predictive intelligence” is the phase before optimization and essential for the development of optimization models. However, such forecasting has different levels of sophistication and accuracy. Airlines have improved forecasting over time. For example, estimates of elasticity now further refine underlying demand for lower fare classes. The president of United Airlines, an airline that has practiced revenue management at the highest level for decades, recently promised investors a dramatic improvement in forecasting accuracy as a result of enhancements in their decades-old airline revenue management system.
4) The objective of “what” we can optimize is currently under review
Airline optimization has been built around network revenue from basic fares – but is that what airlines should be optimizing? With the growth in ancillary revenue – over 40% of total revenue for some airlines – the focus on base fares is no longer optimal. Also, with increases personalization and e-merchandising, potentially we should pursue “customer optimization” instead of just “network optimization.”
5) Airline revenue management isn’t a silo and more functions need to be included in the model
Revenue management should work closely with other airlines commercial and operations functions including pricing, marketing, and schedules. For example, inventory control optimization models have historically taken pricing as a given, where there is no companion optimization model for pricing. Similarly, airline marketing campaigns need to be better integrated with revenue management, initiating highly-targeted promotional campaigns for interested customers with flights and fares. Demand-driven dispatch is an example of how inventory control can work with scheduling to correspond the aircraft size dynamically with revenue management demand forecasts.
In fact, once an airline, or any big data analytics application, reaches the highest level of “optimization,” it needs to continue to review each preceding level for further refinements and enhancements. Within this context, there is no “highest level”. Big data analytics requires perpetual learning and improvement of each of the defined “stages”.