CLV: Gaming insights are helping banks pinpoint key long-term customers

Q – What learnings from your experience in the gaming business have you not been able to apply in the banking sector?

ZA - What I learned and developed at Electronic Arts (EA) was similar to what we are now using in banking. There’s maybe a little bit of difference in its application. 

Video games are quite easy to download, and you can soon start playing them. At EA we established a really tight feedback loop between the marketing and CLV practices and customer acquisition. We were using active feedback from customer segments and marketing strategies to drive the acquisition of valuable customers. 

The banking industry doesn't operate that way. There's more friction when switching accounts in banking than there is with video games. So, we tend not to run our marketing quite as tightly. We have done some minor experiments but we’re not doing anything like that systematically.

At EA we were running CLV models for 500 million customers a day. Customers were downloading millions of games a day and we used CLV to sort out which marketing campaigns and targeting packages were working best. We then fed information back to the marketers who would adjust their campaigns and bidding strategies. That's a very tight feedback loop. 

We don't do that at NatWest yet.

Q - What are your thoughts about using personalization to maximize CLV?

ZA - I like the concept of earning the right to sell somebody a product. We are using engagement together with personalization to provide customers with benefits in their financial services. Customers don’t buy many financial services. A few times in their lives they might get a mortgage or pick up a credit card or switch to a new current account. But our relationship with customers lasts a long time. So, we are using engagement and personalization to give customers great service around all those points of contact. Hopefully we can surprise and delight them. If we can do that, then we’ll be in the running when those customers consider buying another financial product.

So, personalized engagement is a mechanism to exchange long term value between us and our customers. But it also provides us with a way to earn the right to be a considered candidate to provide our customers with their next financial product.

Q - How will AI transform the computational processes that drive CLV?

ZA – We can build models that not just help estimate CLV but also provide lots of diagnostics data and then information that we can act on.  This was not possible eight or 10 years ago. 

When I was at EA, we were running very large models. But to get the scale we needed, we had to simplify the covariates and the models to get estimates out. We were running models with very large numbers of customers but the actual number of indicators was pretty low. Now with the advent of cloud computing and other capabilities it's so much easier to do that work. At NatWest we are partnering with Amazon and use its SageMaker platform.

Our CVL models are still some of the most complicated and most complex machine learning models that we run in the bank. Almost everything else we do is simpler. Even a calculation for a risk model requires a simpler set of inputs and data than a CLV model.

However, the data needed for CLV modelling can be relatively simple. You can play around and get pretty good insights even with a small number of customers. That's how I started at EA. Before we began calculating big high-frequency models, we were employing CLV using Buy Till You Die (BTYD) models and Excel spreadsheets. CLV can get complicated but doesn’t have to be too complicated. 

Q - CLV is often seen as a tool suited to long term predictions but not short-term considerations. What do you believe?

ZA - To get to a long-term prediction you must do short-term predictions too. Anybody who's just doing lifetime value modelling without looking at what happens in the next week, next month, or next year is missing a trick. How else do you verify a lifetime value if you're not tracking your short-term expectations against your long-term models?  

Let's be honest. If you look at some banks’ one-year value projections for their customers, all they’ve done is take last month's revenue and forecasted to the end of the year in a straight line with assumptions about NIMs (net interest margins) or whatever else is going on in the background. 

We have quite a naive forecasting process and in a lot of banks. CLV is a much more sophisticated forecasting tool.

Q – What did you discover in gaming that you would like to see emerge in the banking industry?

ZA - I'd like to change people's relationship with money. One of the biggest challenges we face in promoting financial health is time consistency.  People often make immediate choices to spend when they should be making long term decisions to save. This is similar to how people make food choices when you're trying to lose weight or exercise choices when they want to get fit. All those kinds of time consistency problems benefit from shortening the intervals, making day-to-day decisions, getting fast feedback about customer behaviour and then nudging them in the way that helps them accomplish their goals. Such CLV thinking was common during my time in the gaming industry.

If all banks operated that way and we really improved people's financial knowledge, we could have a pretty big impact. It's a super opportunity.

Q – In the past 10 years banks have been trying to mitigate the decline in interest rates by, among other things, reducing their branch networks. What is the key to strengthening customer relationships amid the decline in branch networks?

ZA – There's been a sectorial shift. Banks aren't closing branches just because they want to. In our retail bank something like 96% of interactions with customers are through our mobile app. It’s similar in our sales and service activities. The reality is that branches are getting empty.

I think it's an exciting time to be in banking; to think about what your branches can do, how they could support digital, how digital services might support the branches, and how they should share feedback with each other?

The same with telephony. Telephony is still a pretty good channel, particularly inbound telephony for service calls.  How should we think about the intersection of branch, telephony and digital? In many businesses, data is often separated within those different channels.

To give customers a great experience, banks have to bring the channels together. So the telephony agent knows if the customer is in a branch or online and the online agent or the relationship manager in the branch also knows what's going on. You must think about the whole population of customers you're serving. Heterogeneity matters. Some customers might not be mass affluent customers but they’re still valuable.

Q- What impact do you think GenAI will have on the banking sector?

ZA - We're still in the early days of Gen AI, not the early days of AI, but GenAI. Most banks are looking at using GenAI to achieve operational efficiencies.

But using GenAI to generate marketing content personalization is a really exciting opportunity. I think generating content is a special power of GenAI. No one chooses an advertising agency because it produces neat material just as you want it. We choose advertising agencies because their creatives come up with crazy ideas that sometimes resonate with customers. Using GenAI to generate content marketing content is similar. GenAI hallucinations, as long as you have good checks to protect customers, can generate ideas that are really creative. They might not be what a typical banker would have thought of. And that’s actually valuable.

Q – In the past few months, major retailers, airlines and telecos have announced data collaboration partnerships. How do you expect this trend to affect the banking industry?

ZA - It's a logical move. It’s similar to what's happened in the grocery business. A lot of grocers now have ad tech platforms and are allowing partners to target customers using their CRM and loyalty data. It's interesting that a big banks like JP Morgan is moving into data collaboration because it has a big retail and a big commercial business. They're connecting both sides of their customer base. It's quite an exciting market model.

In the UK we have unique sharing opportunities because of the PSD (Payment Services Directive) and PSD2 (Revised Payment Services Directive) and the sharing rules around the CMA9 (the nine largest banks in the UK as determined by the Competition and Markets Authority) that allow us to share data with customer consent. I also run the open banking group at NatWest. We're doing things, for example, with identity sharing and identity verification where we've generated APIs that retailers or commercial businesses you can use. If their customers want to opt in, we'll provide checkout data as well as account, address and age verification. We get a little bit of value from providing that information and the customer gets a cleaner, better checkout process. But customers must always give consent at each transaction.