Air Canada: Democratizing & Accelerating Data & AI Projects

Scaling AI Marie Merveilleux du Vignaux
air canada logo
The mighty Air Canada team has built and runs democratized processes across multiple functions, teams, and experts to achieve extraordinary ways to make work, marketing, and the customer experience better.

96%       

Improvement over prior marketing campaign analysis solution     

Hours    

Vs. weeks or months to build predictive models                    

Successful digital transformation and acceleration of AI projects hinges on the ability for easy collaboration, sharing, and understanding. These elements not only allow companies to build good data habits that permeate the organization but also to innovate for the future.

Air Canada has long said “goodbye” to the multiple centers of excellence and silos between technical- and business-oriented teams that often stand in the way of this success. With Dataiku, the mighty Air Canada team has built and runs democratized processes across multiple functions, teams, and experts to achieve extraordinary ways to make work, marketing and the customer experience better.

What’s great about [Dataiku] is it’s easy to look at the results in a quick way, make changes, and look at the results again. With business users that don’t understand really actually how it’s being done, but still understand the value, this was, really, really powerful and helped us to make a customer scoring model within two to three months that we’re going to now use on a day to day basis.

— Hervé Riboulet, Director Cargo Analytics and CRM at Air Canada

Post-Campaign Analysis Automation

After each marketing campaign, teams need to understand performance to repeat what works (and not repeat what doesn’t work). Air Canada’s post-campaign analysis (PCA) assesses campaign performance by looking at multiple datapoints across a range of channels, from email to website, paid media to conversion metrics. PCA was previously challenging because data was scattered and analysis was ad-hoc, taking so much time and resources that the team couldn’t actually use this method to evaluate all campaigns. 

Air Canada first solved for data centralization into one place with Snowflake. Then, they leveraged Dataiku’s visual flows and DataOps to automate and schedule the PCA process. 

The customer and loyalty analytics team generalized (parametrized) the processes so that marketing stakeholders would only need to provide the campaign parameters (e.g., booking dates, travel dates, origin and destination, campaign project number, promo codes, registration codes, etc.) that are required to run a PCA process. From there, thanks to Dataiku, PCAs are automated in batches instead of manual, bespoke, ad-hoc analyses. 

Once processes are run, all outputs are pushed to PowerBI in one single dashboard for consumption by marketing stakeholders. The result? Marketing stakeholders have quick access to PCA insights on all campaigns, meaning faster, better decisions to continuously increase performance over time:

  • Where it used to take two weeks to build one PCA, the team can now spin up 12 PCAs in 3.5 hours. They have fulfilled and eliminated ~90% of campaign PCA requests so that only deep dives on select campaigns remain. 
  • The team estimates an opportunity cost of 96% improvement over the prior unscalable manual solution. With more than 200 campaigns or potential PCAs per year, it would have cost over $400,000 in average data scientist salary to produce all these insights as opposed to a fixed cost of ~$15,000 with the new automated PCA process. 

Dataiku enabled the orchestration of the processes, allowing us to automate, create the flow, and schedule our processes as opposed to have to manually re-code and treat these projects as ad-hoc analysis, which has reduced analysts' efforts to a few hours one day a week. The seamless integration with Snowflake (to easily access the data) and PowerBI (to easily showcase the outputs) also helped streamline the whole process.

— The Customer & Loyalty Analytics Team, Air Canada (source)

Quick Modeling Solution to Increase Marketing Effectiveness

One of the customer and loyalty analytics team mandates is to produce signals through predictive models, customer segmentations, flags, or recommender systems, that can be leveraged in marketing campaigns.

By building a solution that leverages Dataiku and Snowflake, 80% of the total effort to build a predictive model (previously weeks/months) is now reduced to hours of marginal efforts. Air Canada successfully democratized the intelligence they build about their customers — now, any logic that a data scientist would run in their code every single time is being crunched once and everyone can benefit from it, instead of reinventing the wheel for every project.

The team created a Customer 360 solution in Snowflake, which consists in pre-crunching hundreds of features at the customer level, ready to use for analytics or business intelligence (BI) purposes. The team leveraged Dataiku’s visual flows, AutoML, and custom machine learning to connect Customer 360 to Dataiku so they can easily and quickly fetch multiple potential features and train new predictive models, customer segmentations, and recommender systems in hours instead of weeks or months.

Once models are created, all scores are fed back to the Customer 360 storage table, leveraging MLOps, then showcased as profiling variables in the standard profile and democratized to the marketing community.

This solution allows the team to quickly get to a minimum viable product (MVP), on which they can decide to invest more time (or not), thus increasing performance and accounting for diminishing returns.

image2-Sep-19-2025-09-42-19-9268-AM

GET TO KNOW AIR CANADA

TECH STACK
Dataiku + Snowflake + PowerBI

KEY CAPABILITIES
Machine Learning, AutoML, AI Engineering Ops, DataOps 

The following Q&A occurred during Everyday AI Conference Toronto.

Interview of Luc Gagnon, Director of Analytics, Travel Marketing & Loyalty Spend at Air Canada

 

You May Also Like

Agent Sprawl Is the New IT Sprawl, Here's How to Control It

Read More

The Business Case for MCP

Read More

Everything to Know: AI Agents for Supplier Risk Assessment

Read More

Avoid AI Agent Washing: 4 CEO Priorities for Signal Over Noise

Read More