CUSTOMER SERVICE

Customer Service by its very nature is a difficult industry. Created to respond to and solve customer problems. Customer service professionals need to quickly determine the issue and the appropriate solution – all in a matter of seconds, minutes at the most.

A lot of the ongoing communications between clients with issues and those respective companies are unstructured information: emails, voice communications, or letters. In order to effectively “mine” the relevant information, each item must be read, usually several times: first to determine what it is about and to whom it should be routed, second by the recipient to determine what the problem is and how to solve it, and usually a third time, by an archivist who is either manually inputting the information into a database or is correlating it to a set of “known” issues or problems.

Successful companies have found that this customer information is extremely valuable – so they save it, cross-reference it, and turn it into structured information for future use.

Content Analyst enables its customers to quickly and accurately summarize, sort and search customer data with these features:

Automatic Categorization
Sentiment Analysis
Conceptual Search (Search on Steroids)

Automatic Categorization - How it Works

Content Analyst was designed from the ground up as a learning system. Much like a dispatcher needs to learn about specific product features, Content Analyst can be similarly trained about key features, issues, and options.

“Training” Content Analyst is as simple as providing a sample set of documents – exemplars – that are representative of the products, offerings, options, or other pertinent service information. In many Customer Service cases, the “training” is done with actual customer correspondence – Content Analyst will “learn” as well as identify important information while it is being trained. The only other information it needs to perform its functions are categories into which the information should be sorted.

Armed with this basic information, Content Analyst is then ready to receive documents. It will read each one, compare the concepts and word contexts against the exemplar set, and assign a category. Our categorization on average achieves 90% accuracy versus controlled human categorization which is estimated to be accurate less than 60% of the time in real-world settings.

Automatic Categorization - How it delivers Value

Automatic categorization means customer queries are instantly routed to the appropriate person or department. For the typical customer service operation, this saves at least one day, sometimes as many as 3 days, between receipt and response. It also eliminates the challenge of secondary requests because clients’ initial responses went unanswered (clients with issues are always short on patience).

In large operations, it can allow service organizations to reassign people previously tasked with the mundane job of sorting through incoming correspondence, to more rewarding customer-facing (and often revenue-producing) assignments – without the cost of backfilling those “sorter” positions.

Sentiment Analysis – How it Works

One very interesting aspect of a conceptual-based search product is that the idea of “categorization” is easily turned into value analysis. Content Analyst can take a list of terms or phrases that are identified with sentiment – concepts such as like and dislike, satisfied and unsatisfied, impressed and underwhelmed, as well as all the ways people express degrees of sentiment – and correlate them to the documents it is reading.

Content Analyst sorts the documents according to sentiment as you have described it, providing an immediate snapshot of how your clients are responding to a particular item.

Sentiment Analysis – How it delivers Value

The only way to accurately gauge sentiment is to get information directly from customers being served. Because a small percentage of answers could be biased, you need a large group of respondents. In a traditional survey mode, a company would have stores hand out cards with questions regarding service and responses measured by “1-5, where 1 is least satisfied, 2 is…” and so on. Those cards then need to be tabulated by computer, and generate data from which conclusions can be made.

But that’s only part of the solution – these forms often have an essay section where the clients can elaborate on why they rated a particular item poorly. Those sections need to be manually read and the information digested. Some surveys don’t even offer an area for customer comment – which often can be a source of very useful information – simply because they don’t have a way to compile textual answers along with numeric ratings.

Content Analyst is an ideal interpreter for such sentiment-based questions. Better still, we automatically correlate the “whys” to the “what's” for anything that has a textual response.

The result is a very accurate barometer of customer sentiment – as well as an ability to easily “mine” the concepts underlying that sentiment – far in advance of the periodic reporting most survey firms provide.

“Search on Steroids” - How it Works

Content Analyst’s powerful search software goes beyond the simple keyword and Boolean logic searches that power most search solutions. Instead our software is already trained to search out and identify concepts and relationships as it performs its indexing function. The more documents that Content Analyst reads, the more it actually learns - obscure issues, solutions, and concepts become “fleshed out” as the software reads more and more relevant documents.

Content Analyst turns these concepts and relationships into mathematical expressions – after all, language is mathematic in nature. By using the power of mathematics, Content Analyst can quickly search for related concepts, ideas, and context based on a sentence, a phrase, or even a whole paragraph.

“Search on Steroids” – How it delivers Value

The vocabulary used by the company and the one used by the client are often dissimilar: to most policy holders, it’s just a pre-tax healthcare spending account, but for the insurance company whether it’s an HSA or a PSA makes a great deal of difference. Similarly, on an HDTV is it “Input” or “Source” that the customer uses to change the video signal? Content Analyst instantly identifies the client issue or request and correlates it to the company’s products and nomenclature – without using cross-reference lists and look-up tables that go out of date as soon as they are published.

Another challenge is polysemy - these are words like “bank” that may mean the side of a river, a financial institution, or a particular pool shot - that can never be solved out of context. Another complication is unintentional errors: spelling mistakes and more often, scanning errors.

For companies implementing customer-searchable databases in “help” sections on their websites or even within their own service archives, the ability of Content Analyst to “understand” the request whether or not the terminology is correct has a profound effect. For web-based help searches, it can reduce abandon rates by a factor of 3; for internal searches, it can cut the time from request to identification from minutes into seconds.