The art of data management

From the archive:  Just so you know, this article is more than 8 years old - links to external websites may no longer work.
The revelations made possible by big data may seem like the ultimate in technological triumph, but an artist’s touch lurks behind the genuine insights
Data as paint tubes

The plentiful data of the digital era has heralded heady announcements of the ultimate experience in personalized travel. Simply by crunching the numbers, an airline is supposedly assured of offering a customer the right deal at the right time through the right channel.

Indeed, there is much to suggest that big data will deliver on its promises. IBM, in its white paper, Big data and analytics in travel and transportation, suggests that “providers can embrace big data and analytics to more accurately model and optimize demand, capacity, schedules, pricing, customer sentiment, revenue, cost, and more.” The report concludes that “those organizations that leverage big data and analytics to drive improvements in marketing, sales, and operations will be well positioned with the insights and agility necessary for competitive advantage.”

Perhaps the most obvious example of leveraging big data is Google’s search engine, forever massaging search algorithms to ensure relevant content and a host of tempting offers, in response to the 500 million-plus searches made using the engine every day.

But that very same engine has also shown the challenges that big data can present. In 2009, a Google analysis of flu-related searches suggested a more precise picture of the spread of the illness than was available from respected medical authorities. But Google’s predictions of the number of flu cases, based on the number of flu-related searches, soon began to lose their accuracy. Earlier this year, Science reported that between August 2011 and September 2013. Google over-predicted flu cases in 100 out of 108 weeks. Such pitfalls need to be fully appreciated by airlines if they are to realize the potential of big data.

Ghosts in the machine

In Google’s case, one explanation for the faltering accuracy of its predictions was the changing nature of the search. Patterns detected at one point in time do not necessarily apply to data collected at a later stage, especially when the collection method has been influenced by those previous patterns.

So, Google’s recommended searches were throwing off the results. And some of the people who may have believed they had the flu probably didn’t have it, a form of false input.

Dealing with enormous data sets gathered from a variety of touch points can lead to a number of other difficulties. It is very easy to believe in correlations that are statistically relevant, for example. But in big data terms, those correlations could just be ghosts in the machine. The relationship may not be causal and any “meaning” cannot be inferred from the data alone. In other words, there are plenty of blind alleys to run into and drawing meaningful conclusions can be a complicated process.

Such technical complications mean the appearance of exactitude implied by big data can be misleading. If a company doesn’t ask the right questions of big data it cannot expect the right answers.

Sitting in silos

Airlines also face issues very specific to the industry. For an airline, big data is essentially every piece of data available, from operational information to customer tweets.

Naturally enough, much of this data is unstructured—meaning it is data from monitored sources, such as social media, rather than data that is purposefully captured, such as passenger name records. Unstructured data is estimated to be up to 85% of all data.

Airlines have to put the information into a format that can be understood by the powerful algorithms of analysis software, which is easier said than done given the speed of data acquisition and the volumes involved.

Another massive issue facing airlines is the fact that traveler information is stored in so many disparate databases, says Dave O’Flanagan, CEO of Boxever, a travel data agency.

“Each database has an incomplete picture of the traveler, and integrating these different sources to get a complete, 360-degree view of the customer is difficult,” he notes.

“Furthermore, in many cases the data is treated as ‘throwaway’—a traveler may come to the website, perform several searches, and ultimately book, but the next time they come to the site, they’re treated as a net-new visitor, having to start all over again. In most cases, only a very small group of travelers are given a personalized experience.”

While the retrieval and use of any meaningful data is made challenging by this silo framework, airlines are adapting. According to the SITA and Airline Business Airline IT Trends Survey, 97% of airlines recognize they need to do a better job sharing their data in-house and have plans to do so by 2015.

“We’re seeing a wave of passenger service systems contracts up for renewal over the coming two to three years, which will ignite the needs for better intelligence through use of big data,” says 15below Technical Director, John Clynes. “And let’s not forget, there are technology platforms out there today that can sit across these internal CMS, reservations, frequent flyer and other systems, capable of extracting the information needed to deliver a truly personal customer experience.”

"Successfully combining the many disparate data sets to achieve that goal may necessitate airlines changing their business processes. “

Big data can generate a lot of insights into sales and operational aspects of an airline,” says Kevin O’Sullivan, Lead Engineer, SITA Lab. “However, acting on these predictions may require a degree of flexibility not normally available in airlines. For example, is an e-commerce engine dynamic enough to provide real-time offers based on the shopper profile? Is the staff rostering system dynamic enough to adjust the schedule in future hours? Airlines will need to address existing business processes to introduce this degree of flexibility to take advantage of what big data can offer.”

The attempts at integration are complicated by third-party providers that play an important role in the complete journey experience, from home or office to destination hotel. The trend among providers is to vie for exclusive ownership of that experience, rather than facilitating seamless handoffs between the various touch points. Airlines are cross-selling airport transfers, recommending hotels and rental cars, for example.

“Ultimately, however, the customer will decide what type of user experience they want and expect—and thus how much information they’re willing to share and what their expectation is about what of their information is shared between suppliers,” believes O’Flanagan.

Again, the SITA and Airline Business survey underlines the signs of hope. Some 48%–64% of airlines plan to start sharing data with partners such as ground service providers, airport operators and government organizations.

Art for science’s sake

Perhaps the greatest challenge in big data, however, is that it is ultimately an art and not a science. “Speed of insight and speed of action are becoming core differentiators,” says Marty Salfen, General Manager, IBM Global Travel and Transportation Industry. “The tools are now readily available to any airline. The art is knowing the right questions to ask.”

Having access to valuable data doesn’t necessarily mean making the best use of that data. It takes human insight to know which questions to ask and to interpret the answers given within the context of the airline domain.

“Certainly technology can help take away some of the heavy lifting associated with collecting, collating, and analyzing information,” says O’Flanagan. “It can even be employed to learn from the way customers respond to specific messages or offers, and automatically promote or demote those appropriately. And this is incredibly important to enable airlines to respond nimbly to changing consumer behaviors, available inventory, and external factors. But ultimately, it’s important to have human checks and balances—having staff who can review the insights and actions to make sure they match the objectives of the business, and then can tweak the technology to reflect the same.”

Perhaps the greatest challenge in big data is that it is an art and not a science

It’s not enough to just collect and store a lot of information; it’s what you do with the insights and learnings from that data that have real, measurable impacts on a business.

“The important part about actionable insights is that you don’t have to take them on all at once—even using a small set of insights can help boost conversion, revenue, and loyalty,” O’Flanagan continues. “For example, one of our large online travel agency clients used insights about web visitors who were abandoning their online shopping carts before completing their purchase to automatically trigger an email 30 minutes after the session closed to remind the visitor they didn’t complete the process. In one market on one web property, just sending this email and giving customers a quick and easy way to complete their purchase netted almost $1 million per week in additional revenue. These learnings can be parleyed into other properties and other markets, and then the next insight can be actioned.”

If airlines can get it right, big data could be an airline’s single greatest competitive advantage and a building block of sustainable profitability. The companies that combine domain expertise, mathematics, and agile business processes will  be those who benefit the most.

And for that, data artists are just as invaluable as data scientists.

{C}{C}The big data era

Areas of focus

Airlines are addressing primary areas with big data analytics. IATA’s Global Aviation Data Management (GADM) program, for example, has evolved from the Global Safety Information Center and has the goal of integrating all sources of operational data received from a multitude of channels into a common and interlinked database structure.

GADM will support a proactive data-driven approach for advanced trend analysis and predictive risk mitigation. Other areas of focus include:

• The customer experience: Big data and analytics can help create a comprehensive view of the customer, dramatically improving customer interaction at every touch point across the end-to-end journey experience

• Predictive maintenance: In an asset-intensive industry, success depends on the reliable performance of assets. By capturing and analyzing more operational data, analytics can help manage and maintain assets to improve up-time, equipment life, and total cost of ownership

• Fuel optimization: The ability to analyze more variables and more historical information with higher frequency—even in near real time—allows for smarter fuel optimization techniques, which can generate immediate cost savings.