The IATA Data Science Lab transforms data into value and business impact through a combination of artificial intelligence and big data derived from the airline community. The Lab compiles important amounts of raw data into meaningful reports by developing and applying algorithms and techniques in areas such as social and information network analysis, machine learning, data mining and natural language processing. In the COVID-19 context, data science is a significant asset to understand and recommend solutions.

Data science applied to the airline industry

IATA supports its member airlines and the wider industry with advocacy, standards, and services. Within those three major areas, IATA has built data sharing communities in which data-driven services are provided for the industry. The variety and size of the data received from this community enable IATA to be in a unique position to transform the data into strong insights. Also, as a non-profit organization, IATA is able to serve the same community in a neutral and equal manner.

The IATA Data Science Lab provides data and analysis capabilities, allowing the airlines and other stakeholders to power-up their revenue strategies, existing business models and to create new revenue streams.

Initiatives supporting efforts against COVID-19 crisis

Here is a sneak preview of the current initiatives in the IATA Data Science Lab:




Provide intelligence on passengers and cargo traffic to provide IATA with a data-driven approach in their discussions with governments. Provide tangible data insights, comparative analysis and visualizations highlighting the impact on flights patterns due to COVID-19.
COVID-19 TRAFFIC IMPACT: PASSENGER FLIGHTS PREDICTIVE ANALYSIS Tracking travel demand is an important component of the restart plan. To help formulate a response to these challenges, a set of insights / indicators are created for the airline industry in order to answer the question "where and when demand is likely to rebound?" 
The solution consists of the following mains steps:
  •  Step 1: Identify signals / events that preceded the resumption of passenger flights.
  • Step 2: Build a historical dataset to capture signals / events identified in step 1.
  •  Step 3: Estimate passenger flights restart probability.

Create a global dashboard summarizing key indicators such as country restrictions, capacity trends, passenger sentiment analysis, shopping searches and ticketed sales for future travel.

Targeted at airlines in the area of commercial planning, network planning, pricing and revenue management, sales and distribution.


Past Achievements of the Data Science Lab

During 2019-2020, the Lab explored seven potential ideas integrating  artificial intelligence technology into business processes so that these ideas become an asset to the business.

Four out of the seven ideas delivered high accuracy allowing IATA to take action. 
* Accuracy is the percentage of correct predictions within the total number of input samples.

  Idea OBJECTIVE Accuracy
Transparency in Payments

Provide member airlines with transparency in payments: 

Provide classification of credit card payments to allow airlines to adjust their payment policy.

90% accuracy of predictions
Cargo iQ

Create the most possible reliable process time for any given set of shipment planning differentiators​.

Increase cargo revenue by improving reliability and predictability of the air cargo shipment execution.

83% accuracy of predictions

Cargo iS

Forecast shipments' tonnage and total charges per month, week and day (six-month horizon) and different levels of aggregation (world, region, country, city) 

Increase member airline revenues by optimizing the shipping forecast.

92% accuracy of predictions

Treasury Dashboard

Forecast airlines revenue based on past settlement.

Revenue management optimization.

60% accuracy of predictions

Agent Risk monitoring

Predict agent behaviors to prevent default and bustouts, and ultimately act prior to the default and protect airlines funds.

Secured airlines funds, by reducing BSP (Billing and Settlement Plan) losses.

85% accuracy of predictions

Short Term
Traffic Forecasting

Estimate past and future airline traffic of passengers (total passengers and total fares) daily with 3 months horizon.​

Use machine learning algorithms: 1. to complete the missing estimate of passengers 2.  to forecast future estimate of passenger flights

88% accuracy of predictions

Airport Operational Excellence (AOE)

There is a lack of an industry reference for airport performance. Hence, the need to build a global view of airport performance.

To help identify the top performing airports and assets at that airport.



Innovative solutions, tangible impact

Focus on Transparency In Payments (TIP) 


  • Problem statement: Virtual cards are used more frequently and are becoming the payment method of reference in certain types of transactions.
  • Objective:  Identify the type of credit cards used for an airline ticket payment to enable the airlines to take action to adjust their payment policy.
  • Output: By using artificial intelligence and machine learning, the Lab created a predictive model that identified the type of credit card used for a given transaction.
  • Added value: Due to the very high number of transactions received in the IATA Billing and Settlement Plan, manual classifcation of type of  credit cards is extremely complex. This automated solution has provided the capability to address this challenge.

Focus on the Data Science Lab

The Data Science Lab aims to improve the industry’s performance through the provision of data and analysis capabilities. The Lab’s services include a complete portfolio of data solutions, from industry to commercial products to bring efficiencies and ultimately generate revenues for the airline industry.
Projects must include a clear economic and/or industry outcome with a focus on the end benefits for the airline industry.

  • Rapid prototyping for digital transformation: A live-data prototype collects statistical evidence of the effectiveness of your ideas as a future final product/service. Furthermore, it explores the data without restriction to specific engineering practices or protocols.
  • Better knowledge of your data: Applied data science to your raw data to extract meaningful business insights. The output (e.g., statistical insights, predictive models) can be presented within a dashboard for more mature ideas or consumed as a micro-service API to build a more complex solution.
  • Pre-processed datasets: Most data-driven organizations include a team of data scientists. However, even these teams still need to collect, aggregate, clean and enrich datasets. IATA can help those organizations to accelerate their time to market by providing pre-processed datasets. It can take the form of existing datasets but can also be created as specific datasets to match exact needs.

The life cycle of an IATA data science product / service can be described in the following six steps:

  1. Ideation: Define industry and market needs
  2. Exploration: Explore the feasibility and data availability. Get subject matter expert commitment to support the Proof of Concept (steps 1 to 4)
  3. Design: Write business and technical specifications documents (Business Requirement Document , Software Requirements Specification)
  4. Build: Structure and clean data, train and test the machine learning model and get stakeholder validation.
  5. Product management and deployment: Define the pricing model, distribution channels and value proposition. Deploy automation and scalability.
  6. Feedback and support: Review product performance, get new enhancements and change requests, support customer service incidents and address bugs.

See full process (pdf)


  • The process can iterate through several cycles (based on new requests from customers).
  • The "Build" step requires stakeholder acceptances in order to continue. Otherwise the process may start again from the "Ideation" step with additional inputs (i.e., data, change request…).
  • 7 ideations explored
  • 82% average accuracy of predictions
  • 4 proofs of concepts to be promoted as Minimum Viable Products