Ai Editorial: To count on AI, ML and DL, bank on astute data strategy

First Published on 15th February, 2018

Ai Editorial: For one to reap maximum benefits from artificial intelligence (AI), machine learning (ML) and deep learning (DL), there has to be a meticulous acquisition of data, making it worthy of use, further propelled by “training” and “re-training” of models to deliver, writes Ai’s Ritesh Gupta


It’s pulsating to assess how the ever-improving statistical, data-driven systems and algorithms to interpret patterns in data sets can shape the future of every interaction, every experience with travellers.

Considering the fact AI and machine learning demand substantial time and proficiency to get going, airlines need to lay a solid foundation for their respective data strategies. But that doesn’t tend to be the case with this industry.

“Airlines (over the years) have largely focused on safety and operations, running their operations. Even as certain carriers have made progress with their data strategy to refine their merchandising, customer service etc., typically in case of mid-size airlines they still “don’t know who their customers are”,” points out Binay Warrier, Head of Business Development, Loyalty & CRM, IBS Software Services.

And among those who initiate data-driven decision-making, “some operate a basic CRM to support their marketing function”. “One issue that has been with airlines is usage of different systems for booking data, at the airport (such as departure control system) etc. So, if interactions happen at the airport or transactions take place, or even data is available based on the in-flight experience, all of this is disjointed, doesn’t give one picture of the traveller. If one were to assess the airline purchase journey, there is shopping around not just fares – but also destination, scheduling, air- and non-ancillaries etc. The booking window varies considering leisure and business travel. Airlines are in a unique position where they hold a relationship in a transaction for the lengthiest amount of time,” said Warrier. As it turns out, there is an opportunity to communicate, right from dreaming to booking to consumption of the product to the post-travel era.

And this can be meaningful only if data is captured, pooled, processed, analyzed, visualized and action is taken in methodical way, and for that a data platform is must.  

The ultimate goal is to bring together multiple identifiers and connect them to work out a unique traveller profile.

Key aspects are as follows:

Preparation: Readiness is an issue, as availability of data doesn’t mean that it would result in insights. “In case of carriers that have been around for a while, it is not easy to embark on a data strategy by leveraging an available platform. The way they (airlines) are handling data today it is not easy to take information out of that, process it, store it in a way for serving a future purpose/ running analytics on that. That’s an important aspect, plus can be a costly affair too. So that’s where data cleansing, deduplication, integration, pool management etc. comes into play.  Also, a gap analysis or missing pieces of action needs to be factored in,” mentioned Warrier. It is also asserted that an organization needs to be ready for issues around data quality, metadata management, access, sharing, ownership, security etc. This also needs to be considered along with data strategy plan, change management, execution, testing and learning, and measurement metrics etc. Critical questions that need to be answered include data ownership, ascertaining the quality of data, refining of data architecture etc.

Capturing data: The sum of all interactions/ requests can only be incorporated in a profile if data is captured. An airline needs to unify data from several sources – CRM, from staff/ call centre interactions, booking flow metrics, purchase data, from social applications etc. “Profile data, transactional data, mobile and social data – these are basic requirements for setting up a profile,” mentioned Warrier.

Today’s data platforms are capable of storing data from any internal or external source, as well as unstructured data. Overall, airlines should be able to integrate, cleanse, standardize and dedupe data to create a single view of the traveller by merging all of the available online and offline data.

As we highlighted in our article earlier this week, JetBlue today is able to aggregate a single view of the customer (on service channels). So all interactions (say featuring a JetBlue account on Whatsapp, Facebook Messenger, Instagram etc. or an interaction at the airport or with a call centre executive) are captured and aggregated into a single conversational view of a customer. If a conversation takes place between passenger at the airport and the airline staff, then how about analyzing the same without any privacy issues? Today, as IBM asserts, technology is in place to “capture the audio of the conversations and run speech-to-text and tone analysis directly on the device, thus completely avoiding sending any sensitive data to the cloud”. Essentially what is being done is “tone analysis” and this can be equated with attributes like happiness, sadness or anger. Such insights can go a long way in coming up with a relevant recommendation when the passenger interacts or attempts a booking next time.

Working on data: Cleanliness of the available data would decide the efficacy of the whole exercise, with key to same being cutting down on the size of the data set and paving way for algorithms to examine the same. Other areas include refining the integrity of the data as well as doing away with irrelevant/ pointless information, which is part of transformation. “De-duping, cleansing is fundamental to identify the unique customer (find missing and inaccurate parts of the information) are the main starting steps…you may run algorithms or even incentivize the customer to share correct information,” mentioned Warrier. And then curation is equally important. So it could be about blending – demographic information, online behavior (website visit, purchase etc.), in-app behavior (sessions, last opened etc.), email interactions, social media activity and offline interactions.

A core objective is to assess the flow of travellers and that data across their journey. This way an airline can evaluate where internal or technological silos may exist that affect the passenger experience.

Also, data platforms work in conjunction with marketing software, plus pull or push results to 3rd party reporting and visualization tools. Plus these data platforms, blended with AI and ML, result in identification of patterns and the likelihood of certain behavior. After working out profiles and evaluating identities across devices, travel companies can:

  • Deploy algorithmic audience modeling capabilities to capitalize on what an organization knows about their most valuable audience segments and rely on the same to discover new audiences with the same potential value.
  • Use data-driven marketing tools like cross-channel analytics, predictive analytics, and advanced personalization to shape up desired customer experiences.


Hear from experts about data strategy and making the most of AI and ML at the upcoming Ancillary Merchandising Conference, to be held in Edinburgh, Scotland this year (9-11 April, 2018).

For more info, click here


For Ai’s 2018 Events, check -

Follow Ai on Twitter: @Ai_Connects_Us