Making omnichannel experience a natural conversation with Machine Learning

Omnichannel experience to natural conversation

Big Data is changing the landscape, providing an unprecedented amount of data about consumer behavior, customer needs and desires, and new economic models created by new « data driven » technologies.

A Gartner Survey in 2014 showed that 73% of respondents had invested, or planned to, in Big Data in the next 24 months, up from 64% in 2013. Improving customer experience is in the top priorities cited.

The availability of this huge amount data breaks down the established business models while at the same time creating new ones. The genesis of this movement is the interaction and symbiosis of three major developments brought by technology: mobile devices, cloud services, and Big Data management tools.

The breaking down of established business models occurs mainly in retail, although few sectors will be able to avoid it, an industry where multiple platforms and communication media for shopping come together to form a new omnichannel paradigm. Companies and businesses, regardless of industry, that successfully exploit these new communication channels, distribution options, and amounts of data from  multiple sources, will differentiate themselves and thrive in providing superior sales experience and service to their customers.

The Omnichannel paradygm

The term Omnichannel has become very common ; however, it has often been misused especially in what concerns the sales and distribution industries. Common are also the discussions on the implications and opportunities for customer experiences that offer a mix of mobile devices, web sites, and face to face interactions  using multiple communication channels, such as phone, instant messaging, email, chat, or social networks. And even though the discussions focus mainly on the sale, the implications of these new rules for customer service and support are equally significant. Used wisely, omnichannel can change unsuccessful and frustrating customer interactions  into positive experiences, highly encouraging customer loyalty.

Gathering, correlating, and analyzing customer interaction data through different channels is the key to transforming nightmare experiences in moments of pure happiness, at least in a moment of « emotional connection ».

The combination of the availability of data (Big Data) and Machine Learning (automatic data interpretation) in all forms, including predictive analytics tools and even neural networks, is the foundation that enables highly effective, informed, and profitable interactions  for both customers and businesses.

Mobile applications : the rise of social customer interactions

The trend is clear : all studies show significant changes in the relationship between companies and their customers. For example, when questioned about the preferred options of contact, two thirds of consumers still prefer to talk directly to a person, while 23% rather use the company’s web site, the main driver being the omnipresence of mobile devices and application usage.

With smartphones, always present companions, we expect immediate access to any information or any services we needed in real time. But we are not monogamous in what concerns our smartphones, as they are not the only devices available to us. This means that we routinely start interactions on a terminal (through a given channel) and resume through a different one later. At least that is what consumers expect to be able to do.

Our natural, daily conversations provide an interesting metaphor for the best possible omnichannel customer relationship. Applied to the business world, this means that making customers happy requires the reproduction of the qualities and characteristics of private conversations that we consider « successful ». The best conversations are continuous, can stop at any time and start again at the same point. They are also interactive, efficient, and mutually benefit both parties, by providing new and useful information.

In an omnichannel world, consumers change terminal according to their taste and expect to start a task, stop it, continue later and still maintain continuity throughout the process. Transforming the customer experience into a natural conversation gives strength to a brand and to the engagement with its clients, eliminating frustration related to a poor customer service and getting loyal and happy customers.

The Strategy

Redesigning  proof given and critical processes, such as those used for customer support and account management, is a major challenge ; therefore, it is important to have ambitious goals while relying on highly pragmatic plans. A goal without a plan is just a wish. We must think « big » boldly, but also have a clear strategy

The audacious part is to address the most important and most critical processes of customer support. This involves dealing with large processes and tasks, across multiple channels, of great importance for both customers and the company. The pragmatic part’s role is to quickly identify the areas where these changes would be most profitable.

It is vital to identify and prioritize the channels that can be paired, such as web and phone or web and chat, that is, channel combinations that customers use the most. This logic is particularly successful for an omnichannel customer experience. For example, banks that give support to clients about payment issues often find that the number of customers calling support within 24 hours of using the web site is 5 times greater than the average. This correlation is a good indicator of the existence of gaps in the web self-service since customers are obviously not satisfied with their online journey. The need for a second support channel increases service costs and customer dissatisfaction.

Customer Centricity

Too often, companies  start by defining internal processes and then tools to optimize the work of their agents or internal administration without really thinking about the experience they provide to their customers.

They should  do precisely the opposite : analyze the customer experience from the outside, as seen by the client. An effective and structured technique for such analysis (from externally to internal) implies the definition of client personality types, such as user archetypes, representative of demographic groups and client segments  with different needs and different omnichannel customer journeys  within the customer service support.

Each profile corresponds to an archetype, or an idealized representation, of a key customer group. Following their respective paths, through the different channels, for different issues, throughout the support organization can reveal bottlenecks and blocking points, but also opportunities for optimization and automation.

Big Data : from “Mountains of Data” to useful (and used) Data 

E-commerce and advertising were the pioneers of understanding the value of gathered data and its analysis, but each company has mountains of data that can be used to improve the customer experience. However, it is not enough  to aggregate that amount of information, structured and unstructured, into a melting pot.

Machine Learning and predictive analytics require a thorough and comprehensive customer analysis. That means that historical data transactions from other sources, such as CRM, ERP, marketing tools, or satisfaction surveys can be gathered into a single, common data structure.

Start with the indicators and transaction history that can best highlight the progress of the client in the labyrinth of his journey. Look for indicators that show obvious gaps, inefficient segments, frustrated customers, ruptures and breaks in the omnichannel journey. Identify the tasks that have led to multiple attempts, measure their success rate and transfers (from one channel to another, from one service to another …).

Also identify the data that may reveal problems, such as the transition from web to phone, that are good indicators of customer frustration and the need to escalate problems that must be solved at a higher level.

Data mining with Machine Learning

Let us remember that the goal is not only to provide a consistent, multichannel experience, but also to be proactive instead of reactive, to anticipate customer needs and prevent problems, not simply solve the issues as they arise. Using a sports metaphor, run to where the ball is going, not to where it is. For example, one of the greatest customer frustrations is having to ask the same question or explain the same situation to different people or different online forms.

A classical approach is to manage the information transfers from place to place. The proactive approach is to anticipate problems and to ensure that customers do not even have to ask. Predictive analytics can eliminate customer frustration not only by contextualizing customer queries and handling, but also use statistical and forecasting models to be able to anticipate requests.

One of the most common topics of customer interactions, regardless of the type of service provided, concerns billing issues. When the number of interactions for a specific type of client, for a given subject, is abnormally high, it’s probably a sign of potential problems. Combining detailed interaction history (Web and telephone, for example) in a predictive analytics approach can make sense of what appears in the first approach as a background noise. Many customer interactions, spread (to not say lost) over time and channel, become intelligible and workable in an omnichannel conversation.

In what concerns financial institutions, Machine Learning allows them to understand and anticipate customer behavior. For example, Machine Learning can identify customers with problematic accounts due to the repetition with which they check the account online or via mobile applications to see whether the current payments have been accepted.

Redesigning the customer experience using data and Machine Learning allows financial institutions to be proactive. They can send an SMS or generate automatic calls to these customers when their account balance is about to reach problematic levels according to the analysis of their behavior and usual monthly expenses.

Learning and Optimization

A predictive multichannel system is based on two fundamental elements: optimization and self-learning. The best predictive systems are improving over time, learning from past events, adapting to changing conditions, andoptimizing based on key performance indicators.

These elements are particularly important in managing customer support, where the customer multiplies the use of different channels and the quality and quantity of data and business priorities are very dynamic.

Ultimately, providing a first class customer support in the multichannel era involves analyzing large volumes of data to understand customers and their needs and integrate all the available elements in a fluid and successful customer journey. The motto must be: learn and anticipate.