One of the phrases on the buzz these days is big data, which claims to transform banking and insurance industries. Before embarking on such a journey, large organisations need to first understand that big data isn’t a plug-and-play initiative that would automatically make them “insight rich” (or super rich, come to that). From my experience, financial institutions are still insight poor, but this isn’t because of a lack of data. Rather, it stems from an inability to utilise existing data sets efficiently for various reasons: multiple data sources, disconnected platforms, lack of capacity and/or expertise on data analytics, are some of them. If you’re not already in a position to handle existing data sets efficiently, are you aware of the problems that big data can impose on your organisation?
According to the Joint Committee of the European Supervisory Authorities, big data is defined by the three ‘Vs’ (volume, velocity and variety). When it comes to variety, I would add a ‘U’ (for unstructured). Still, trying to capture the sense of big data with one definition might be difficult. Looking at other descriptions, one may also add to the three Vs and one U, according to Brewster Knowlton, that big data require non-traditional database systems and business intelligence to analyse them (listen also to A:360 Podcast by Brewster Knowlton on ‘What is big data?’).
The above definition should already be pointing to the direction of why big data isn’t a solution by itself (or at least not an easy solution) and why big data shouldn’t be viewed as a plug-and-play initiative. At the risk of being accused of sounding completely unoriginal, I will advocate that the biggest obstacle for utilising big data effectively is organisational culture, not legacy systems. Here’s exactly where Chris Skinner’s Fintech Bullshit Detector (FiBD) starts beeping loudly. So let’s get practical.
Internal communication gap
I’m still taken by surprise when clients ask me, “Can we make money out of big data?”. To which the answer is always, no you cannot. You can make money out of servicing your clients. Big data is like a toolset, it can help you or break you. The elements to look at for achieving the former are many, hence the referral to the organisational culture.
Look, for example, at the internal communication gap between the central analytics team and retail business units (whether you’re operating in a centralised or decentralised data ownership model). BUs are still taking decisions on “existing client segmentation”, which is static and provides no insights to clients’ lifestyles, life events or everyday moments. On the other hand, big data initiatives from the central analytics teams have yet to influence, or find out how to influence, the way business is taking decisions.
One reason for the above is the frequency of analysis – it’s still slow and can’t keep up with the real-time characteristic of big data. Another reason is the way BUs are operating, largely influenced by wrong-KPIs mentality. Take a look at existing KPIs – sales teams are typically measured by “new accounts”, which could be very far for becoming new customers, let alone profitable customers. Departments are measured by next quarter’s revenue, not by customer attrition. So if your churn propensity models show a customer population that shouldn’t be onboarded because the cost will be higher than the predicted revenue, are you going to ‘disincentivise’ clients from opening an account with you?
As much as performance indicators can be a good management tool, wrong KPIs may incentivise the wrong behaviour. This should come as no surprise, though – people manage their careers, not their jobs, meaning they will pursue whatever target you give them. How certain are you that current KPIs reflect your strategic goals? If big data analytics require change of KPIs and relevant bonus schemes, how prepared is your organisation to make such big changes?
Developing the ability to analyse big data effectively across the organisation is one challenge. To create meaningful messages is another. Here I’m referring to the misconception that n = all. Analysing quintillion bytes every day across different dimensions doesn’t make the understanding of (the causality) of correlations any easier. Sample error and sample bias are still relevant, and the need to understand causality is still important. If you don’t understand causation, you won’t know what may break the correlation.
If you believe your organisation is insight poor, you need to identify the drivers before you look at the data. Look at elements such as platforms, models, algorithms, people’s skills, capabilities, capacity, frequency of analysis, and so on. If you don’t manage to sort out the drivers of the problem, big data can only make your problems bigger.
READ NEXT: Becoming a data-driven organisation
Image: garriphoto, Shutterstock.com