Breaking Down Big Data

Analyzing Big Data For More Precise Commercial Planning

Today, airlines can take large amounts of data, or Big Data, break it down through data analytics and apply it to network planning and scheduling functions to produce more optimal results than ever before.

One of the hottest trends in the technology world is Big Data. While the term is cemented in the lexicon, it seems to have multiple interpretations. For the purposes of this article, Big Data is described as:

  • Gathering and processing extremely large amounts of data,
  • Using data analytics to glean more insight into patterns and trends at a very detailed level.

In the past, data needed to be summarized and stored in silos because of the limitations of data processing and storage. Today, the step-function in data storage and computer processing capabilities make it easier to acquire and analyze large amounts of detailed data. With the ability to store larger volumes of data, powerful analytical tools were developed to turn data into information. Airline planning departments are consumers of large amounts of data.

Can harnessing Big Data have a profound and positive impact on network planning and scheduling? The answer is yes. The more information a network planner has on customer travel behaviors, the better airlines can use that information to build a more profitable route network.

The History Of Network Planning

Thirty years ago, network planning was a laborious task. Some airlines created econometric models and government aviation data to estimate market sizes and market share. That all changed when Marketing Information Data Tapes (MIDT) data became available. In regions, such as North America, the United Kingdom and Western Europe, where global distribution systems (GDSs) had an extremely high share of industry bookings, there was rich data for network planning. The MIDT data provided a good proxy for O&D market sizes and market shares. Airlines would then use this data in their forecasting systems to build their network.

This model worked well where there was high GDS booking penetration, but changes in the industry challenged this model. The first risk was the growth of low-cost carriers (LCCs). Many LCCs accepted bookings only through their reservations system. In effect, they dropped off the grid, which meant all that demand was completely blind to the traditional carriers using MIDT. The second risk was the Internet. Airlines saw an opportunity to reduce distribution costs by funneling bookings to their websites. This also created lower GDS share of bookings, creating more uncertainty about market demand.

To some extent, this uncertainty remains today. Some airlines are examining ways to utilize all relevant data available to make better forecasts. The challenge is to use these disparate data sources to analyze and identify if there are any correlations to help better forecast market sizes.

For example, when working with airlines in network planning, we use the Sabre® Global Demand Database (GDD), which incorporates more than 40 sources of data to estimate market sizes. By calibrating the data from multiple sources, we enhance the existing MIDT data to create more accurate market forecasts and reduce the risk of being dependent on one data source.

Big Data And Network Planning

Big Data can drive positive change for network planning in two ways:

  • Going deeper into the current data,
  • Utilizing non-traditional data that might correlate with the airline industry.

Historically, network planning departments used MIDT data at the O&D and/or segment levels. Given the state of the technology at the time, that was as deep into the data as they could get. However, Big Data enables airlines to span well beyond these levels.

For example, Big Data supports micro-targeting, a marketing strategy that uses consumer data and demographics to identify individuals or small groups of like-minded individuals and influence their thoughts or actions. This is a common practice in politics. Previously, the method to determine which candidate would win was done through random polling, with a margin of error. Now, campaigns use Big Data (voter registration, party affiliation and precinct) to get a more precise idea of the make-up of the voting district. Rather than going door-to-door for votes, they go to their own party members’ doors to learn if they will vote or not, and this data then goes back to the party for analysis and action.

Micro-targeting can be used in network planning by drilling deeper into the MIDT data. Instead of using the market, or even cabin-level data, it is possible to go down to the Reservation/Booking Designator (RBD) and postal-code levels.

Here is an example of how more granular data can improve market forecasting. An airline was ready to take delivery of a new aircraft and wanted to know where best to fly it from its hub city. Using the traditional segment-level MIDT data, the airline ran through its normal processes to identify the best market for the new aircraft. Using the results of the analysis, the airline determined the market would be a marginal performer at best. After a few months of service, the new market was exceeding expectations; the actual demand was far higher than the initial forecast.

Curious, the airline wanted to understand why the market outperformed the forecast. The first analysis was to go to the postal-code level. What the airline realized was that it had underestimated the size of the catchment area. Because, historically, there weren’t non-stop flights from the airline out of its hub, customers on the outskirts of the hub city were driving to other airports to get non-stop flights. Once the airline added the new non-stop route, the potential customers migrated back to the airline’s hub city. Airlines that have operations in high-density metropolitan areas need to incorporate the use of micro-targeting at the postal-code level to understand the true size of their catchment area.

Enhancing Traditional Forecasting Data

To map long-distance calls and volume within the United States to understand interactions between communities, MIT used anonymous mobile-phone data. The map illustrates the paths of the communication, while the thickness of the lines measures the volume. Through Big Data analytics tools, this data can be used to establish if there is a connection between mobile-phone data and air-travel demand to enrich the traditional forecasting data.

Using Non-Traditional Data Sources For Network Planning

We have seen that traditional airline data can be enhanced through Big Data and data analytics. With the explosion of data and the ability to crunch large volumes of information, in the future, there may be other types of data that are not traditionally used in airline forecasting. Here are some potential ways non-airline data might be used enhance market forecasting.

The Sensible City Lab at Massachusetts Institute of Technology has been using Big Data and analytical tools to “understand the interface between people, technologies and the city. For example, they have worked on using digital data to understand the flow of taxis in a city, and using that data to optimize travel patterns.

In another example, MIT used anonymous mobile-phone data to map long-distance calls and volume within the United States to understand interactions between communities. The result of the visualization of the data is shown in Figure 1.The map shows the paths of the communication, and the thickness of the lines measures the volume. The striking thing is how similar this mobile-phone map looks like an airline route map. Visually, it looks like there is a potential correlation. Using Big Data analytical tools, the data can be evaluated to determine if there is a correlation between mobile-phone data and air-travel demand to enhance the traditional forecasting data.

Another potential use of non-traditional data for network planning is census data. Census data has very rich and detailed information about statistical cities and regions, including population and household wealth. This data can be correlated between statistical regions to model growth and trends within and between metropolitan areas, and how that drives the demand for air travel.

Commercial activities and the flow of global travel can be used to create more accurate demand forecasts. Determining demand for air travel in developing countries can sometimes be challenging. Because these countries are developing, the true demand in MIDT is understated. One potential leading indicator of growth is the investment of infrastructure. Growth of materials and equipment can indicate the growth of an emerging economy.

Big Data And Scheduling

While Big Data is promising for network planning and forecasting, there are also more detailed data that can improve the way an airline builds and optimizes its schedules. One application is close-in re-fleeting (CIRF). Some airlines have been doing CIRF for quite some time. There is revenue benefit in matching capacity with demand for aircraft with common cockpits. To perform CIRF, most airlines take a feed of bookings forecasts from the revenue management system and apply an average fare for the market. It is usually at a high level of detail and performed manually.

As a form of micro-targeting, airlines with O&D revenue management systems have forecast information at the highly detailed O&D level. For those carriers, it is possible to provide real-time O&D forecasts, bookings and fares. Having information at such a detailed level provides rich information to the scheduling department so it can more precisely determine which aircraft to swap.

Another Big Data area of benefit for scheduling is incorporating new data sources to build more reliable schedules. Historically, airlines generally had two sources of delay data: their internal on-time data (for their airline only) and from government statistics. The website Flightstats.com has on-time performance data for 85 percent of the industry, in almost real time. Because the Flightstats.com data takes into account industry performance, this data can be used to complement existing data, providing information to schedulers on potential schedule adjustments to build better schedule reliability and on-time performance, which, in turn, saves delay-related costs.

Some people may be a bit skeptical about whether or not airlines can really use infrastructure investment to forecast future demand for air travel. However, based on the fact that the availability of data is growing at a stupendous rate, it’s quite possible. The winners in this race will be the airlines that mine that data and use Big Data analytics tools to discover correlations within the data to extract a competitive advantage over their competitors.