“As a marketeer I want to promote our special passes to our frequent travelers via Meta (Facebook ads) and Google Ads in order to improve customer experience and revenue”.
In other words, the objective of Thalys is to enhance their customer experience by proposing a Thalys pass that is tailored for frequent users.
Thalys is a high-speed train service connecting major cities in France, Belgium, the Netherlands, and Germany. Thalys.com is the company’s largest sales channel today.Today the large majority of train tickets are sold through their website. As a business Thalys was facing different challenges:
Ambitious business objectives even though substantial infrastructural changes weren't possible due to the upcoming merger with Eurostar
Loads of customer data collected but underused
Customer data & technologies living in silos
The rigorous respect of customer data privacy
Our mission is to help Thalys identify, prioritize and put customer data use cases into production in order to drive business impact
Step 1: Identify users to promote the pass
As this pass has been designed for frequent travelers, we first needed to 1) have a clear definition of the term “frequent” and 2) confirm that we could identify the users that meet this criteria. In our case, Thalys had already implemented a process that highlights on a daily basis such users in their CRM (based on their frequency criteria) and made this customer list available in their SFTP.
Step 2: Ingest users data to Bigquery using Fivetran
The second step consists of ingesting this data into a data warehouse. There are several reasons we want to do this :
Retrieve the data in an accessible environment
Centralize data into a single source of truth
Transform and enrich the data from other sources. In our case , we will need to make the required matching between user_id and email addresses.
We used Bigquery as the data warehouse solution as we already implemented it in the context of another use case.
To ingest this data from the SFTP into Bigquery, we have chosen Fivetran as our ETL solution. Its role is to automate the transmission of data we want according to our preferred custom frequency. In our case, we extract the user_id.
Step 3 : Clean, transform and enrich the data
Once the data is accessible in Bigquery, the next step involves transforming and cleaning the data with the aim of building the audience. In this exercise, some factors need to be taken into account :
Given that our customer data only has pseudo ids, and considering our intention to synchronize this data with both Facebook and Google Ads platforms, we need to enrich this data with corresponding email addresses. These email addresses are retrievable from their database and the linkage is established using the user pseudo ids.
Since we are going to integrate this data with Facebook and Google Ads, email addresses are required to be securely hashed to uphold privacy standards. It is worth mentioning that Google Ads and Facebook automatically hash emails, but as an additional security measure, we hash them before transmitting the data.
Step 4: Sync Audiences with downstream destinations using Census
Now that the data is cleaned, it is time to create our audience and synchronize it with our designated destinations. Leveraging Census as our reverse ETL solution, we can directly build the audience through this tool using SQL.
A crucial aspect in this process is to exclude customers who have not provided their consent for targeted marketing, ensuring compliance with consent regulations.
Step 5 : create the campaign
The final stage is to confirm the message (tagline + visuals) for the media campaigns.
After approval, Thalys's CRM team set-up and launched the campaign. To reiterate, the objective of this use case was to promote the pass by targeting customers who showed potential interest in this product.
In order to measure the success of this data use case we decided to look at different KPI’s:
Direct transactions measured in Google Analytics’ attribution model
The amount of transactions taken from Thalys’ database for all users who were part of the defined audiences and received the email or notification push in a 24 h window
Results were striking with more than a 3x return on investments. Considering here that:
The investment is represented by the sum of Human37’s consulting fee and the licensing cost for the new technology we had to deploy (BigQuery, DBT & Census)
Revenue is measured as described above