Traditional methods of evaluating service quality in banking provide an incomplete and highly subjective view of your client service. Fortunately, natural language processing is providing client experience and operations leaders with the capabilities they need to understand, quantify and improve their services.
Client service is one of the most competitive and challenging areas of investment banking. Answering client questions, properly managing expectations and resolving client issues as quickly and efficiently as possible are key to winning and retaining business. But the question is, how do you know you’re doing enough? How do you compare your client service to the competition? And most importantly, how do you know what to improve?
Client service: A blindspot for banks
Client service is a complex web of different systems and communications channels. It’s difficult enough for operations to have visibility over the shared mailbox of a single service desk, nevermind managing hundreds or even thousands of mailboxes across a bank’s many service teams.
As a result, you have account and relationship managers struggling with poor visibility into client interactions until an event is escalated or a complaint raised. They lack the resources to get ahead of problems and proactively manage them. Nor do they have the data to demonstrate where a potential incident was avoided or turned into a positive outcome.
This environment makes the client experience little more than guesswork, and service improvement a process of trial and error. Operations and service leaders need the capability to quantify and track a holistic quality of service for the enterprise. Only then can they demonstrate improvements and get ahead of issues before they snowball into relationship-ending crises.
What is quality of service?
The first thing you must decide is what exactly you are looking for, and how you will measure it.
Firstly, how do you know when your client is having a good experience or a bad one? A great indicator is emotion. The tone of the conversation and the language being used can give a strong sense of whether the quality of your service satisfies the client’s needs. Yet sentiment alone isn’t perfect, nor is it enough. The language we use can be deceiving. When a broker sends your team a ‘friendly reminder’ they aren’t necessarily satisfied with the service you’re providing. To record it as such would result in a fundamentally flawed picture of your service.
An accurate representation of your quality of service needs context as well as sentiment. There are many sentiment analysis tools that attempt to analyse and extract the tone and emotional content of messages on a large scale. But such tools are typically based on keyword matching and language models that can’t be tailored to your specific context. We’ve already seen how language can be misleading, especially when viewed without domain knowledge of the interaction.
Only the subject matter expertise of your frontline agents can tell you whether a contact really is positive or negative. Yet you can’t ask your service desks to analyse each and every message in this way without massively increasing their workload or your total headcount. Not only do you need to extract this information across all channels and touchpoints, you then need to aggregate it in such a way that stakeholders are able to understand the service experience at a glance. Easier said than done.
Beyond automation: Scaling NLP
These obstacles have led to many banks falling back on broker reviews, client interviews and surveys to evaluate their quality of service. Yet these are blunt instruments at best, which can only ever give you a narrow cross-section of opinion. Furthermore, any findings are only useful in hindsight. You won’t be able to preempt emerging issues or measure the impact of any changes to your client service in real time. It’s akin to driving by only looking in the rear-view mirror.
However, recent advances in natural language processing (NLP) are fueling massive improvements in how leading banks quantify their quality of service. NLP models are now able to extract the intent and sentiment of communications between clients and agents, accurately and reliably. Through the model training process, agents are able to impart their subject matter expertise - including their understanding of context and specialist language - giving these models the context they need to interpret situations and client messages accurately.
Through NLP, banks are able to understand their client and employee communications at speed and on a massive scale. Machine learning models can then automatically estimate the impact of these communications on the bank’s quality of service, aggregating them into a central quality of service score. This opens up the bank’s communications - its service - to trend analysis and real-time monitoring.
Automated alerts can be set up that are triggered by specific events or deviations from acceptable quality of service levels. Service issues can be identified and resolved before they even have the chance to impact the client relationship. It paves the way for a client service that is real-time, responsive, constantly monitored and continuously improving.
Summary
To measure and monitor their quality of service, banks need to be able to extend analysis across all client messages and communications channels. Manual, analogue methods fail to achieve the coverage and responsiveness required.
NLP and machine learning techniques enable banks to analyse and monitor their client communications en masse and in real-time. Client service can be aggregated into an overall quality of service score, making it easier for operations and service leaders to keep track of and respond to changes in client satisfaction.
For the first time, investment banks have the tools they need to comprehensively quantify, monitor and improve their quality of service. They can now truly compete in client service.
