In the age of technology and startups, the financial services industry has undoubtedly gone through shake-ups and changes, causing a bit of unease and uncertainty. Yet, during my time in the industry, I have been lucky enough to see it mature and raise the analytical stakes.
Just one or two years ago, most of my customers’ concerns were about the quick launch of a few proof-of-concepts to experiment with what was possible with behavioral segmentation. Now, we have moved on to talking about how to deploy these initiatives at a larger scale and how to sustain the first data products coming from these segmentations.
Here are some common questions and thoughts I repeatedly hear from and discuss with big industry players looking to get ahead of the start-up wave as well as insights and actionable next steps on these topics.
What Can I Learn About My Customers With Big Data?
TL;DR: A lot.
Data science and big data have recently opened the doors to certain types of data that couldn’t be handled by traditional business intelligence (BI) teams. By setting up an unconstrained data lake with a vast array of data types and formats, teams can experiment freely and are not limited to restrictive, siloed data sets that might not completely answer the business questions at hand. This shared resource and self-serve model encourages collaboration habits between IT teams, analysts, and marketing or business departments.
As an example of something that would be difficult with traditional BI teams, take the rich customer data analysis from the web and from mobile apps, which help give a more complete view of customer journeys. This type of data provides a new level of dynamic segmentation that goes far beyond traditional customer segmentation for financial services.
You could take this analysis a step further by enriching the customer journey with data from other silos, such as metadata from call centers and emails. Ultimately, the goal is to combine these silos with transactional data to produce a very rich and dynamic behavioral customer segmentation.
With these combined data sets and the experimental data lake, teams have the capacity to set up experiments outside of production constraints relatively easily (compared to classic BI projects where the initial results can take several years). From there, timely insights can efficiently enrich the existing Customer Relationship Management (CRM) platform and lead to further machine learning experiments that augment or improve customers’ experiences.
The Primary Big Data Challenge Is Organizational, Not Technical
A representative from a key French retail bank shared feedback with me recently about her experience delivering her first big data project. It took 18 months from first contact to deployment to deliver that project.
The biggest slowdowns in this case (and in most cases I see, so this is certainly not unique) were organizational and procedural challenges. Not technical ones. This is good news, as this problem is easily solved by addressing and optimizing the ways in which the company gathers BI and mines, collects, and provisions data.
The key to solving for these organizational challenges is to empower members of different teams involved with data to work more closely together, including working in the same open space or using solutions that simplify processes to collaborate early and often on projects. With those barriers removed and the right technology in place, creating models, testing, deploying, iterating, rinsing, and repeating is easy. Removing the organizational hurdle allows businesses to achieve their first results from a big data experimentation project within just six months.
From Behavioral Customer Segmentation to Concrete Value
Once CRM teams validate their capability to make relevant behavioral segmentations through their new big data platform, the next step is to actually use those new customer segmentations to send the right message to the right person at the right moment.
The problem here tends to be that there are so many ways to achieve behavioral segmentation — so many paths to the end goal —that to find concrete first use cases is not so easy. In addition, unclear objectives of an IT department working in parallel on the big data ready-for-production infrastructure combined with security, availability, and scalability concerns makes the perfect recipe for ending up in the experimentation stage forever without ever deploying a solution.
This is a major concern. Improvements can only happen when teams take what they learn by experimenting, find the best solution, and then actually make that solution a reality. This is where the value of big data lies.
To get a good start on providing concrete value, it’s crucial to activate data initiatives with operational teams in an efficient and visible way. Here are two examples of how to involve operational teams quickly and share concrete feedback:
- Give call centers a list of customers every week to call first based on their communication channel preference and/or a particular event detected.
- Enrich marketing campaign tools with new labels and scores (updated daily) that offer new, advanced variables with which to define campaign targets.
With such actions, operational teams can put concrete value behind the big data buzzword, which will further support more advanced big data use. While these principles derive from what we learned from customer segmentation in banking, they can be applied across a myriad of industries. The bottom line: When implementing your data science strategy, ensure you first have concrete expectations for deployment and industrialization.
Have questions about big data in financial services that I didn’t answer? Reach out.