Improving Customer Service and Security with Data Analytics

Improving Customer Service and Security with Data Analytics

The advantages of analytics to customer service have already been shown. Now the question becomes: How can analytics be used to improve security?

Organizations are collecting more and more data. And while rich data allows personalized service, detailed data about real people (rightly) often raises concerns. Just as this data is increasingly valuable to organizations, it can be valuable to criminals as well, leading to an ever-escalating series of data breaches. Data analytics exacerbates trade-offs between security and service; the analytical processes on data can, at a minimum, raise privacy concerns for individuals because much of marketing analytics tries to learn as much as possible about potential customers. These analytics processes are becoming increasingly powerful at de-anonymizing people from their trace data.

However, these de-anonymization techniques are an example of a way that analytics offers at least a partial solution to the problems it has exacerbated.

Consider, for example, placing a call to your bank for help after losing your debit card. The core problem is that, before providing customer service, the bank must authenticate that you are who you say you are. This authentication process must begin with the assumption that the caller is a malefactor impersonating the real customer — guilty until proven innocent. The bank will help the caller only after being convinced of the caller’s identity.

While this process is annoying when we’re customers seeking help, we actually want and need this level of security. It is in our best interests that the bank will verify that we are who we say we are before continuing to assist us. After all, we don’t want the bank to hand out our money (or our new debit card) willy-nilly to just anyone.

Historically, this telephone authentication process involves answering a set of questions. What is your account number? What is your personal identification number (PIN)? What is your Social Security number? Can you verify the last three transactions in the account? What is your prior address? The process continues, potentially escalating to security challenge questions based on shared secrets, until the bank is convinced of our identity.

This process is adversarial by design. Even the name “security challenge question” evokes a combative stance, a challenge. The initiator of the call is not trusted until passing through a gauntlet. For banks, it is unfortunate that so many initial interactions with a customer are adversarial in nature.

But data and machine learning, specifically speech processing, offer a great example of an invisible way that analytics can simultaneously help improve security and service. The technology itself isn’t that new, but speech processing has progressed to the point now where financial services companies can match a caller’s voice to their prior calls, allowing the authentication process to occur behind the scenes as the customer service conversation progresses.

Fidelity Investments, for example, encourages the use of voiceprints to confirm identity within the first moments of a conversation. HSBC is beginning to do this not just for premier clients, but at scale for retail clients as well. And the change doesn’t just help the customers avoid yet another password or secret question: Barclays notes a 20-second reduction in time to authenticate — and those 20 seconds add up quickly to considerable savings in employee time for the bank.

The convenience and savings may be the initial drivers of this change. However, perhaps a bigger effect, more elusive to quantify, is the change in orientation. Data and machine learning can ensure that the customer interaction begins by focusing on assistance rather than challenge. Customer service can work with, not against, a caller who (in all statistical likelihood) is a genuine customer, not a con artist — innocent until proven guilty, in other words. Customer service doesn’t have to assume initially that callers might be nefarious — and identity validation can occur in parallel while the conversation is getting started. This means that the unlikely (but potentially damaging) scenario that a security threat exists doesn’t have to poison the majority of interactions with valid customers — without leaving it unaddressed. Organizations can relegate the pesky security issues to behind the scenes, where they should be kept. The authentication process is passive, churning along in the background. Security must only become visible if a problem is found. In this case, the artificial intelligence is augmenting the human employee in ways that are not visible to the customers.

As a result, valuable and expensive training time for customer service employees can be spent more on banking and less on security. While the direct result is more effective customer-service training, the indirect result is scale. When a new security threat emerges, the bank can deploy countermeasures quickly to all customer service interactions.

And more can likely come from this initial application. For example, a customer may in fact be who they say they are, but may be being coerced. Or they may be suffering from some impairment. Speech patterns that indicate these possibilities can be brought to the attention of the customer service agent for further assessment.

Because it is, by definition, an invisible process, examples like this may get far less attention than humanoid robots or chatbots. But analytics can help mitigate some of the trade-offs between the security and service that increased data collection exacerbates. These applications may have a far greater effect on customer relationships for organizations than the ostentatious examples that may be more effective at marketing than managing.

Source | http://sloanreview.mit.edu

Regards!               

Librarian

Rizvi Institute of Management