Formerly known as Renault Crédit International, France-based RCI Bank focuses on automotive financing and insurance for Renault, Nissan, and Mitsubishi. The firm handles several million service contracts yearly and was struggling to interpret customer behavior and activity due to data silos.
To address these issues, the bank launched a Big Data initiative in 2016. The strategic objective, according to Jean-Charles Gemin, Head of Big Data at RCI Bank, was to enhance the bank’s existing risk capabilities with better marketing tools:
We have a very long history in data mining and risk scoring since we are a bank and need to underwrite financing. So we mastered our risk management, but when it comes to customer knowledge and marketing, we were a bit late to the game.
Historically, RCI only sold its products and services at dealer shops, but now also approaches customers directly. But that change of direction makes the bank all the more dependent on effective marketing campaigns, a problem when there’s not enough insight on offer, admits Gemin:
The change in our commercial strategy meant we needed to know our customers to properly cater to their needs and do business with them. So we had to invest in Big Data and data processing to meet marketing KPIs.
Devising the analytics plan
According to Gemin, RCI tackled this in a “very classical manner” by listing use cases, understanding what it wanted to do with the data, then defining the architecture and what would be the primary systems to serve its needs. This resulted at the end of 2016 in an analytics proof of concept and architecture tests.
RCI built its architecture based on two key systems: Oracle’s Big Data Appliance enables the firm to ingest large volumes of data from disparate sources, while a customer Master Data Management (MDM) master data platform from DataStax Enterprise (DSE) provides access to online sales tools and external data from stakeholders including insurers, builders and surveyors.
The DataStax Enterprise tech also serves up the bank’s customer data to different systems, including R/Python systems developed by its data scientists to build machine learning and deep learning models. DSE is then used to store the calculated propensity to buy or potential churn. It can also be used in batch or in real time for lead management, customer care or as part of a more comprehensive marketing campaign.
However, changing business needs meant that the team had to stop work in early 2017 due to uncertainty over whether the plan for the architecture was sound enough. Gemin explains:
We chose to stop the project as we started to wonder whether our architecture was solid or not. We then asked Deloitte to come in and review what we had done and planned by that point – and thankfully it was all OK.
Deloitte worked alongside RCI between May and September 2017, and at the end of the review project, the bank continued with the customer master data management project, which went live in April. So far RCI integrated around 6 million client clusters in France and another 700,000 from its UK subsidiary. Germany followed, adding another 700,000 clients to the system.
Looking back and despite the initial challenges, Gemin says the project has gone according to plan and met expectations. He says:
The main challenge was the installation of all the infrastructures, of the right technological stacks, to have everything operational. At that point, the technical teams at both DataStax and Oracle encountered problems but solved them within short timeframes.
We were looking for a robust architecture and database that was available 24/7, and that’s what we have today. We can see that the solution is compelling for real-time purposes, which is what we wanted as we need to not only feed data from marketing campaign tools that could be loaded in batches, but we also wanted to be able to address real-time issues.
The bank badly needed an efficient search mechanism, which the Solr engine built into the DataStax offering provides, delivering results from queries about its millions of customers by entering small bits of information, such as surnames and postcodes, in a matter of seconds. Flexibility was also a consideration, adds Gemin:
We were also looking for a powerful database since we ingest several sources of data across various countries. But now we are looking into more sources to ingest from and types of information, which means the database needed to be flexible so models could be easily evaluated.
Before selecting DataStax, , the bank looked at off-the-shelf data management tools including Informatica and Talend, while for the NoSQL database, it looked at Hbase and Marklogic. According to Gemin, there were more basic options in the market, but the firm wanted a more technically-rich offering:
There are a lot of technical details to master, but we were able to build this project in France. So other companies elsewhere should also go ahead and try this as well. Projects like this can turn into success stories and be the start of a great analytics journey.
With the first major projects out of the way, Gemin is planning for further evolution in analytics at RCI. As well as building more sophisticated propensity models, other areas of focus in advanced analytics are around machine learning for process optimization.
Evolving the strategy
Even though it’s still early days, the bank can already see the improvements and opportunities around customer experience, marketing, and sales that the projects can bring. Gemin says:
It is now possible to only contact customers once, as we now have a single client view. Given that we also have a consolidated view of all our products, we’re able to make offers at the right moment to each customer.
RCI has 40 subsidiaries worldwide, and the system has already attracted the attention of other offices and projects are underway to roll the Big Data set -up in Germany, Russia, Spain, and Brazil, says Gemin:
Since it is a real success story, the subsidiaries are very interested. Plus, we reduced our implementation times significantly: we initially took six months to roll the project out and can now deliver it in one month.
Even though Gemin could not specify the actual impact the project is having to the business bottom line, the company remains positive about the need to invest in developing its analytics set-up:
Since the project had only gone live a couple of months ago, it’s hard to talk about precise gains. But we need to believe that we will be much more efficient: we will reduce the cost of our campaigns, and we now have a solution to store these model solutions to enable subsidiaries to manage their campaigns properly. While we don’t have figures in dollars or euros that we can disclose, we have to work on [the analytics projects] because it’s an expectation from our management.”
The RCI Bank story illustrates the need to think carefully about decisions related to architecture before jumping on the data analytics bandwagon. To adapt to new business demands, it is inevitable that data architecture must change, but combining the world of legacy data warehousing with new tools can create an environment of opportunity but also great complexity.
Just as in data preparation, enhancements to architecture before starting data analytics initiatives are vital to accuracy of models and also save time, effort and money. This is especially true in cases where analytics experiments go well in the proof of concept stage, but fail when they hit the production stage.
Image credit - Twitter
Disclosure - At time of writing, Oracle is a premier partner of diginomica