Sramana Mitra: Let’s have some examples of those problems.
Sid Banerjee: The companies will often localize website content to certain market or product needs. That localization can often lead to inefficiency. When they find that people have a particular problem using websites or completing transactions, there will be feedback that comes in a variety of forms – both online and traditional support forms – that can quickly help to isolate the root cause of the problem, so they can fix the actual customer experience on the website.
We have had other companies in the telecom space. There are things that are known to create adverse experiences with customers. For instance, there could be network outages or defective phone units. But a bigger question is, “How important is a negative experience to the long-term customer opinion about the brand?” We used Clarabridge to look at all the negative experiences a customer might have with a particular vendor in the telecom space, and isolated those which tended to spike, but [we also looked at] resolution causes versus those that have a likelihood to have a long-term negative impact on customer loyalty. You can see that by the words used and the curve that comes after a negative experience. Does it immediately turn down, or does it stay high for a period? If you think about customer emotion, something that causes a customer to have a negative opinion for a long period generally has a much more negative impact on loyalty and customer retention. By using Clarabridge, we can piece out the sentiment and the attitude that a customer has about various things. We can see the impact is not just that a bad experience has caused a bad opinion, but that opinion seems to be creating a long-term damage to that brand or product.
We use this information to help clients identify how to prioritize the problems that have a long-term negative impact on customer loyalty and satisfaction versus those which are – for better or worse – the cost of doing business, where customers still get upset when they happen, but these incidents are not going to negatively impact customers’ opinion and loyalty as much.
As an example, we find that network outages generally do not impact customer loyalty in the long run,if they are brief and quickly resolved. Billing and pricing changes, which have an appearance of being disadvantageous to the customer, will create an equally negative response as a network outage, but they have a much higher likelihood of turn because of the long-term impact of that change. We can look at those kinds of changes as well as many other kinds and try to determine whether it is something to worry about or something that if you fix, it will go away.
SM: There is obviously one aspect of this problem that is horizontal. It is managing and processing large volumes of data, involving infrastructure. Then there is the core business logic and heuristics. Can you talk a bit about where your real sweet spot and core competency is?
SB: The way to think about the Clarabridge stack is that it has four primary components from an offering perspective. When you are implementing a Clarabridge software solution, the first part of the solution is a set of predefined, configurable connectors to a wide variety of data sources that can come either come from the social sphere, market research, or other online or offline forms.
Then there are customer interaction platforms like CRM systems, chat platforms, workforce management platforms, etc. We connect to all the data that contains insights that we want to mine and make sense of. The second part of the stack is essentially algorithms that take unstructured data, process it, and transform it into quantitative insights. The main algorithms that play here are text and entity extraction algorithms, classification algorithms that apply industry and horizontal ontologies to the data. We look at raw data coming in to a call center for a hotel, and we apply lodging and hospitality classification models, call center models, social models, etc. We also do sentiment scoring. A big part of making sense of customer feedback is sensing what people are saying and with what sort of emotion or sentiment are they expressing a concept. We have a whole set of algorithms, including some that will identify how words are used, how they are modified, and the intensity at which a word is used. Each word that conveys qualitative amplitude is given a number and connected to a concept.
We can also bring in structured data and do other things to basically bring in a complete data set as we need to, regardless of whether the data originates intact or in structured form. If you have CRM loyalty scores, predictive analytics scores, typographic or demographic data, we can bring all that in as well. The last two phases are analytics and distribution. In analytics we apply a range of statistics, derivative and trending measurements, correlation and driver analytics, spike analytics, etc. Those analytics power three uses: dashboarding use cases – typically for business analysts – alerting, distribution, and collaboration processes that go through a process called Clarabridge Collaborate. The way to think of that is if I am a hotel manager or store manager, I can log on to Clarabridge Collaborate. Or I get an email from the system that tells me “You have an issue that is relevant to you because it is your hotel or your customer and you have the ability to resolve it.”
The third interface is the Clarabridge Engage Interface. That is basically a product that sits on top of those algorithms and data migration capabilities that will recommend specific customers that a customer management professional should communicate with. That product can be used by people in social care, customer care, retail or hotels, where increasingly we had clients saying, “We want to empower the customer-facing representative with tools and techniques to be able to communicate back to customers so that they provide feedback about an experience or product that they have control over.”
So it is a full stack. There is data migration, transformation with a particular focus on sentiment, and text analytics processing to make sense of the data so that it can be insightful. Finally, there is data distribution.