Name: Kelvin Ye, Team Lead, Data Science
Country: Canada
Organization: Plusgrade
Plusgrade powers the global travel industry with its portfolio of leading ancillary revenue solutions. Over 200 airline, hospitality, cruise, passenger rail, and financial services companies trust Plusgrade to create new, meaningful revenue streams through incredible customer experiences. As an ancillary revenue powerhouse, Plusgrade has generated billions of dollars in new revenue opportunities across its platform for its partners, while creating enhanced travel experiences for millions of their passengers and guests. Plusgrade was founded in 2009 with headquarters in Montreal and has offices around the world.
Awards Categories:
- Best Moonshot Use Case
- Best MLOps Use Case
- Best ROI Story
Business Challenge:
As the global leader in loyalty commerce, Points, Plusgrade’s loyalty business unit, paves the way for implementing data-driven marketing strategies at scale. Plusgrade powers the global travel industry with its portfolio of leading ancillary revenue solutions. Over 200 airline, hospitality, cruise, passenger rail, and financial services companies trust Plusgrade to create new, meaningful revenue streams through incredible customer experiences. As an ancillary revenue powerhouse, Plusgrade has generated billions of dollars in new revenue opportunities across its platform for its partners, while creating enhanced travel experiences for millions of their passengers and guests. Plusgrade was founded in 2009 with headquarters in Montreal and offices around the world.
At Plusgrade, our Data and Analytics teams sit in Marketing, which allows us to work closely with our Partner Marketing, Product Marketing, and Performance Marketing teams. The Analytics team found that, despite their broad experience and immense data set (or perhaps because of it!), we were struggling to efficiently use all the data at hand, which spanned across multiple customer touch points and tools. This involved a lot of raw data, Excel, and CSV files which were tricky to manage and analyze.
Automation was also a major challenge, as it was difficult to obtain an up-to-date, recurring view of key performance indicators. Before automation, the Data Science team needed to manually process all data from CSV files and score relevant data and export into csv for campaigns.
Business Solution:
The Analytics team adopted Dataiku in 2019. This had three immediate positive effects:
- Building a central data warehouse: Dataiku makes it very efficient to handle larger volumes of data by connecting to any databases, vs. in-memory computing from our previous tools.
- Points started using Dataiku with a database that’s less commonly used in the industry. The team was still able to drastically improve efficiency thanks to the ease of setting up the connection of the database with Dataiku.
- Facilitating data access: The team built clean repositories of SQL and Python code templates to make them easier to access and scale to future models.
- Automation: As the first quick win, the team was able to automate all the scenarios, which were previously run manually at a frequent cadence. This freed up resources for the team to focus on the most added value parts, as well as reducing errors from manual processes.
The team was able to build the following use cases:
- Campaign optimization: Leveraging data from different sources to build ML models to identify customers for campaign targets and optimize the offer to maximize revenue.
- We started automating our ML scoring process starting 2019, but were interrupted due the rapidly changing market conditions of the industry caused by COVID-19. We started to resume ML scoring automation in 2022 as the COVID-19 restrictions eased, and now have automated most of our ML scoring. In 2023 we are also working on automating offer optimization to identify the most suitable offer for our customers.
- Customer segmentation: Getting a more granular understanding of customers to tailor product offerings.
- Product analysis: To understand any feature overlap or market cannibalization between products and enhance our offering.
- Performance tracking: Measuring how promotions impact and key objectives of the company, such as customer acquisition, leads, etc.
Day-to-day Change:
Having the ML process automated changed a lot of our day-to-day processes:
- To speed up the process significantly by removing manual inserts and configurations to score the latest data. This also helped avoid room for error due to manual processes.
- Automation also inspired and helped stimulate the transformation of our data system. We have built a comprehensive data warehouse to work better for automation.
- We also made our solutions more scalable with Dataiku; we are now much better prepared to add models for new partners.
Business Area Enhanced: Marketing/Sales/Customer Relationship Management
Use Case Stage: In Production
Value Generated:
Since starting with Dataiku in 2019, Plusgrade has increased our data models eight-fold. These new models and algorithms are not only the backbone of ensuring promotion efficiency, but are also used to understand the different member behaviors to help customize creatives to maximize member engagement.
We have a case study that showed a close to 40% gross sales increase in a promotion thanks to the customized promotion driven by the models. We continuously improve our models and customize the promotions to boost member engagement and partner revenue.
Value Brought by Dataiku:
The team automated the basic models in several months with support from the Dataiku team before 2020. However, the automation was interrupted due to COVID-19. Starting 2022, we resumed working on automating our more sophisticated suite of models.
During the project, we started to intentionally develop our models and flows with more code-based recipes so that it’s more scalable across different partners and projects. The team also developed a more robust data warehouse solution for better data management for analytics and ML models and migrated most models in Snowflake for better scalability and model performance. Dataiku also showed great ability to adapt to the new data changes.
Value Type:
- Improve customer/employee satisfaction
- Increase revenue
- Reduce cost
- Save time
- Increase trust
