In 2018, Spotify teamed up with Ancestry.com so it can personalise playlists for music lovers based on DNA data, CB Insights reports. In other news, L’Oreal is looking into how they could use shoppers’ smartphone cameras to gather information and blend cosmetics that are just the right shade for customers’ individual skin tone or hair colour. With industry after industry embracing the age of hyper-personalisation, banks are still having a hard time following suit. To give you a sneak peek into our upcoming webinar, we sat down with Sára Hanniker, W.UP’s head of data science, to talk about the power of data in banking, sales and winning (back) the heart of customers.
Why are artificial intelligence, machine learning and data analytics essential for banks today?
We believe that banks need to connect with customers through messages that are relevant and useful to them, grab their attention and give them guidance on their finances. And generate sales, of course. But if you can’t segment your customers, the only thing you can do is put everyone in the same bucket and interact with each and every one of them the same way. And you don’t have endless opportunities to engage them, either. You can’t overwhelm customers and spam their inboxes with messages that have no relevance to them whatsoever. It’s crucial to get to know them as well as possible, and in the 21st century, the best way to do that is through data. There’s a black box between communications and data, and this is exactly what we offer to banks. The ability to turn data into meaningful interactions.
By data I mean traditional customer information, such as online and mobile banking activity, and transaction history that shows how much customers spend and where they spend it. If we have the customer’s consent, we can collect geolocation data and track their routes. Thanks to PSD2, we can draw transactional data from other banks, too. Messages are created based on all these different types of information. Of course, you don’t need all this for every single customer interaction. There are campaigns that are triggered by very basic business rules. Let’s say you see that a customer’s balance has dropped and there’s an upcoming direct debit order for their electricity bill that they won’t be able to cover. You don’t need machine learning technology to spot something like that. But for more complex campaigns you do. And you also need to make sure that all the rules you set up work smoothly together.
Are banks starting to recognise the importance of these new technologies?
I think they’re starting to understand why they are important, yes. There are innovators at banks, too. The problem is that more often than not, they’re part of a dated organisation with dated technologies. For example, sharing transactional data with us in real time is a huge challenge for many financial institutions. They don’t have the necessary solutions so they’d first need to do some development work. And that’s usually where the problems begin. Cultural barriers don’t exactly help, either. Most banks have their own way of developing products and they often refuse to break away from what they know, and remain siloed. To deploy these technologies, however, you need to set up 360-degree customer views and know precisely which message is the most relevant to them at any given moment.
Are there other reasons why banks still lag behind?
Well, besides technology hurdles, another problem is that the data that they have is not organised properly and there are a lot of things they don’t even measure. For example, not every bank will track if you click on ‘Savings’ in your mobile app to make a term deposit but don’t complete the transaction. Or if you start a loan application but abandon it for whatever reason. Most of the time they don’t even have or store this information, or the data is not stored at the right level of granularity. Yet another issue is that banks have tons of internal and statutory reporting and monitoring obligations, and they simply can’t dedicate ten data scientists to internal data mining, let alone figuring out how to make the most of what they find. Especially in a way that they can use findings beyond a single project and implement an entire platform to handle multiple rules in parallel.
But for the sake of argument, let’s say that a bank has a data scientist team of six. They start looking into data to come up with a solution to a problem brought up by the sales team. For example, to drive mortgage sales, they have to find out what life event or events are tell-tale signs that customers might need a new flat in the near future. Once they get the results, how do they put them to use? If they start building machine learning models, these models must be regularly updated and maintained. Do they have a platform or system with ready-to-use data so such models can be easily implemented and kept running? Rarely. This is where our solution works wonders. It gathers, prunes and organises heaps of data so that banks can see and understand the customer behind that data. Once you have that, you can easily build and maintain a variety of machine learning models.
What I also see quite often is that many of them don’t know what to do with data in the first place. Even if banks see that a customer likes sports, shops at Ikea every weekend, has a knack for interior design, and loves eating out and travelling but they don’t have a firm grasp of how they could use this information to their advantage. So we advise them on how to go deeper and find ways to predict what products and services customers will probably need. In other words, we help them adopt a mindset that drives companies like Revolut, for example. What Revolut does isn’t the result of some impossibly sophisticated data analytics and machine learning sorcery. But it still tells me where I’ve been travelling, what shops I’ve dropped by, how much I’ve spent there and what have you. So what they’re big on is customer experience.
Could you give a few examples how machine learning can be used in banking?
Credit evaluation is an area where scoring models have been widely used for the longest time. There are great models for marketing, too, like when you predict who’s likely to sign up for a credit card or how much a customer fits the profile of those who have already signed up. Or what’s the likelihood of them closing all their accounts and churn. But the solution we’re offering goes much, much deeper than this. Because these models simply aren’t advanced enough to allow for proper customer segmentation or life event detection.
One of the things we’ll be talking about at the webinar is how to get from basic applications to more sophisticated ones through the example of card-linked offers, or CLOs.
If used right, card-linked marketing can be a main driver behind customer loyalty. However, what most banks do is send the same offers from the same retailers to their entire customer database. Now, I for one never find anything relevant there. They offer discounts, cashback or loyalty points at stores I never shop at or haven’t even heard about. We’re going to show how banks can boost card usage and make it possible for card holders to shop where they want with the help of data analytics and machine learning. It’s a win-win.
Ready for more? Join our webinar with Sára on 11 June and hear her speak about how AI can drive digital banking sales and customer engagement. Our guest speaker, Tom Peace, Business Development Leader of Financial Services and Retail at Collinson, will dive into how card-linked offers can be used for customer journeys and can provide contextual use cases for banks. More details this way.