20 October 2016
Daniel Cohen

Big data and Banking

Daniel Cohen - DataStax

2Posts 9,307Views 1Comments

Building a 360 degree view of the customer with data is a two-way process

18 October 2016  |  2865 views  |  1

The banking sector in the UK has become much more competitive – new challenger banks have increased savings rates to attract customers while new service offerings aimed at underserved markets are also coming to market. The banking sector has seen growth, mergers and splits over the past few years as well, creating new entities that have to attract customers in new and smarter ways.

All this change adds up to increased demand for marketing strategies that can engage customers alongside new service channels that can keep them over time. However, these evolutions have led to increased pressure on banks’ IT infrastructure, particularly around the use of data.

Marketing and analytics in banking – making better use of all available data

Today, banks and financial services companies engage with customers across multiple channels from physical branches and telephone support through to online services and new mobile applications. These channels have to support a wide range of financial products. This results in a wealth of customer information that is often stored in isolated data silos due to scaling limitations imposed by existing IT infrastructures.

The challenge here is not that banks don’t have this data – indeed, many banks have been building data lakes for years. Instead, it is difficult for them to get this data into a useful form that can be used at the “point of customer value.” This refers to any interaction with the customer where there is an opportunity to either improve their opinion of banking services or where new sales opportunities for financial products might start.

Using analytics, marketers can see which products are really of interest to each group of customers and then target them with directed messaging to increase conversion. The important insight here is where purchase behaviour fits against predictions and models. For example, targeting new parents with saving account information for their children can help encourage them to open new accounts, while offering insurance products to newlyweds is a market that has been developed through use of analytics.

Another example is looking at targeted products or personalised recommendations for high value customers, where each set of targets would have different products associated with them based on previous customer status and current life status.

However, data on customers is spread far and wide across the organisation, making it more difficult to work in this way with data. The deluge of data that customers generate over time finds its way into separate islands of information rather than being one consistent, coherent whole. Because of this, it is very difficult for the organisation to accurately understand all the ways that it might engage with a customer and then model campaigns in the right way.

This complexity makes it extremely difficult to conduct effective analytics campaigns that could provide real value back to the bank, such as auditing behaviour and preferences to spot new opportunities to upsell high margin products. To combat this, it’s important to rethink how data is created and stored over time. Consolidating data lets banks make more use of customer information as they can store data longer, join it with other data sets, and better identify phases of the customer lifecycle. Using this complete overview, customer analytics programmes can be deployed to help increase sales, improve engagement and focus retention efforts on the best customers.

How you store data makes a difference

Storing data involves using both a physical storage medium and a database to organise that data over time. Many database solutions attempt to provide banks with the ability to manage huge volumes of data over time, but they often run into limitations of scale, throughput, or latency.

For example, solutions built upon data warehouse products such as Hadoop can often store or analyse large amounts of data, but Hadoop-based systems can’t handle a large volume of requests at low latency as this is not what they have been built for. Because data warehousing products aren’t designed to deal with real time user traffic, the recommendations or insights provided are built using batch operations. As these run periodically rather than in real time, this limits their relevancy of these insights at the point when the customer is completing a transaction.  

Additionally, data warehouse products often can’t deal with the sheer volume of user requests generated by online or mobile banking applications today. For example, in the 1970s a user might get their account balance as part of a paper statement. In the 80s, users could check that balance via an ATM during the day. Today, online and mobile banking means that balance data can be checked as often as a webpage can be refreshed or an app opened. At the same time, these new channels offer more functionality than just providing one balance total; more services and interactivity drives up the number of times that someone might use their banking app. While this interaction increases customer satisfaction and retention, it also means that banks have often had to employ caching layers to cope with all these transactions. In turn, this increases operational complexity and cost to deliver services.

One alternative here is to look at new databases and data storage methods that can help banks cope with the increasing demand for online services. However, any new approach has to cope with the volume of data that banks create. This means that throughput of data is a necessary consideration for the bank to use and consume the data provided by those online or mobile applications. Without the right level of throughput support, data can be lost. Alternatively, data may have to be thrown away or collection limited, which defeats the point of consolidation. Limiting the amount of data collected - or the granularity of data that is created and stored over time - limits the insights that can be created.

This consolidation of data can be used to reduce customer churn and intervene before the customer becomes a flight risk. If a customer demonstrates low engagement – for example, by not using Internet banking or mobile services when they are available, say – this can be investigated further. It may be that the audience is very engaged on other channels, or they may be starting to consider moving to another bank.

Because all interactions are captured, it’s possible to determine if the customer has either an incomplete level of interaction with the bank or has a negative opinion currently. It’s then possible to target these disenchanted individuals with individualised messages to increase the adoption of differentiating features or products, allowing staff to actively intervene and prevent churn.

Spotting this kind of signal in the noise of all the transactions that are taking place is difficult. New database models can help capture transactions into one place so that each team within the bank can make use of this data. Alongside this, new visualisation and data management techniques can help organise that customer data so people can spot patterns more easily. For example, graph databases can help banks organise data to emphasise the relationships between data items.

This can be positive for customers - for example, spotting patterns that exist across existing customers to help in recommending future products that future customers in those same circumstances will buy. Alternatively, data and graph results can be used for internal productivity improvements. For example, fraud detection can be improved through spotting the patterns that exist across different groups and then applying these patterns to other, similar individuals or organisations. This can therefore help improve security for customers and help stop fraudulent transactions.

Building a 360° view of the customer typically requires consuming data from multiple data sources, each with a different schema, and a different set of attributes. To consolidate this requires a platform that can cope with ingesting all this data and then index it in real time. By bringing data together in this way, it’s possible to focus on marketing in smarter ways that can prevent churn and improve the customer experience too.



TagsMobile & onlineInnovation

Comments: (2)

Melvin Haskins
Melvin Haskins - Haston International Limited - | 19 October, 2016, 04:06

Surely you are assuming that customers have all of their business with one bank. I do not think that I am alone in having my principal credit card from a different supplier than the bank that I hold my current account with. I also took out a mortgage with a different bank altogether. I have Individual Savings Accounts with both my principal bank and yet another bank. How do you build a 360 degree view of people like me?

Be the first to give this comment the thumbs up 0 thumb ups! (Log in to thumb up)
Daniel Cohen
Daniel Cohen - DataStax - London | 19 October, 2016, 09:28

Hey Melvin, to accomplish what you describe for multibanked customers such as yourself, we will have to wait for PSD II. The revised Payments Services Directive from the European Commission creates a framework for third-party access to customer accounts (XS2A), which in turn opens doors for Account Information Service Providers (AISP). Stay tuned.

Until then, indeed, a bank cannot aggregate information it does not have. Now that said, there is plenty to learn from information banks do have, even from a single customer account. This may inform product offers targeting that customer and in turn help prevent customer churn. Start with the basics: how old is the customer? Where does he live? What is his income? His credit history? Now, is he recently married? Perhaps he needs a home loan. Is he starting a family? Offer him life insurance.

With newer graph technologies, we can explore further dimensions in the connections among customer data. For instance, do several customers share an address? A phone number? These connections may be critical in detecting and preventing fraud.

Be the first to give this comment the thumbs up 0 thumb ups! (Log in to thumb up)
Comment on this story (membership required)

Latest posts from Daniel

Building a 360 degree view of the customer with data is a two-way process

18 October 2016  |  2865 views  |  1 comments | recomends Recommends 0 TagsMobile & onlineInnovation

Simplifying Market Tick Data Management

12 August 2016  |  6443 views  |  0 comments | recomends Recommends 0 TagsRisk & regulationPost-trade & ops

Daniel's profile

job title Solutions Engineer
location London
member since 2016
Summary profile See full profile »
Daniel Cohen is a Solutions Engineer at DataStax, the leading provider of database software for cloud applications. Over the last fifteen years, he has led development on large scale applications for...

Daniel's expertise

What Daniel reads
Daniel writes about
Daniel's blog archive
October 2016 (1)August 2016 (1)

Who's commenting on Daniel's posts

Melvin Haskins