Knowledge-enabled support delivery across channels and languages.

Most customer support organizations invest heavily in training and skills development to increase the effectiveness of the agents and technicians who provide after-sale customer support. Despite all good intentions, the payoff from these "soft" investments will always be bounded by three hard limits:

* Work hours. Investments in agent training can only pay off during working hours. When agents are not at work, their expertise is not readily available.

* Turnover. When agents leave and are replaced by inexperienced workers, the pay-off from training investments is obliterated.

* Language. Agents can only help customers who speak their language or languages. If you have global customers, you may be required to train agents with language skills to serve those customers, over and above the training performed in the firm's "home" language.

"Knowledge-enabled customer support" is a phrase that refers to efforts to bridge the gaps in agent expertise across working hours, languages and staff turnover. Organizations that have invested in knowledge tools for agents and technicians have hoped to improve agent effectiveness by making the problem-solving expertise of experienced staff available to others in the organization across diverse service channels and languages.

Many of the investments in knowledge tools over the last decade have had limited success in increasing agent effectiveness and reducing the cost of delivering customer support.

* Support silos. A knowledgebase created for use in a call center was too often not suitable for use through other support channels such as Web self-service access. In addition, the "knowledge silo" may also be a "language silo," meaning that a self-service or call center knowledgebase may exist in only one language, making it useless for those who do not speak that language.

* Retrieval swamps. Building the knowledgebase is an initial challenge, but helping agents find the pertinent content to solve a customer problem is another enormous challenge. Two approaches have been predominant, both with severe limitations. Agents, technicians and customers using self-service are frequently frustrated and unable to solve support problems, despite a substantial investment in the knowledgebase.

* High cost of ownership. Most first-generation knowledge tools required a great deal of special care and feeding. Special skills were required to create and maintain the system; dedicated staff was sequestered to maintain the tool. To create knowledge tools in multiple languages, a great deal of duplicated effort was required. The technical limitations of the knowledge systems resulted in high costs for the organizations using them, giving the whole knowledge tool category a bad, but well deserved, reputation.

This article will look at the causes of support silos, retrieval swamps and the high cost of ownership of knowledge tools. These issues are all crucial considerations for anyone developing an architecture for today's knowledge-enabled support organization.

Support Silo Versus Support Hub

Often, a solutions knowledgebase developed for use in one part of a support organization will have little relation to knowledge tools in use in other parts of the organization. This unfortunate situation has historical, cultural and technological roots. Fortunately, the technological barrier has recently been removed, albeit not across solutions from knowledgebase vendors.

A support silo occurs when the know-how required to solve a customer problem is available in only one place. The know-how may be restricted to one service channel; for example, the field service organization or to a language or geographic region. The will to leverage or extend the access to this know-how may exist, but the means to do so may not. The expertise may be available in the form of service bulletins, technical manuals, a knowledgebase or in the heads of service technicians, and the additional places where this expertise is needed may include the call center, the customer care Web site or the automated e-mail response system.

[ILLUSTRATION OMITTED]

In these cases, the technical requirements are often diverse. In the first place, there is a requirement for partitioning of the knowledgebase contents, since some content may not be appropriate for customer self-service access. Another requirement will be intelligence and flexibility in retrieving materials from the knowledgebase. A novice user will typically require a degree of hand-holding while navigating the knowledgebase, while an expert user, who knows just what he or she is looking for, will want direct and rapid access to it. An experienced user will know how to describe the problem, while a novice may not know how to accurately describe the problem in a way that will identify the solution. The technical requirements for a knowledge tool in customer support must accommodate these users and their needs.

The Compounding Complexity Of Multilingual Support

The challenge is compounded in global support environments. When support resources are required in multiple languages, the cost and complexity of managing these resources escalates quickly.

When a knowledge tool is introduced in the support organization, the cost of ownership of the tool escalates with the complexity of the global organization. Most of the technologies involved necessitate a manual process of localization and synchronization of the knowledgebases for each language or region, with little assurance that the solution delivered to the user in each place will be consistent.

An important consideration for the knowledge-enabled customer support organization will be the effort required to deliver and maintain a multilingual knowledge tool, and the ability of the tool to deliver consistent answers across channels and languages.

The Support Interaction Hub

Recent technical advances in the knowledge management field have made the creation and maintenance of a multilingual, multichannel knowledge resource a new reality. The knowledge-enabled global support organization can now draw on a centralized, manageable resource that delivers the know-how required to solve customer problems, regardless of the language spoken by the customer. By eliminating a great deal of duplicated effort, the cost of delivering answers to customers can be dramatically reduced.

The key technical criteria that enable the creation of a support interaction hub are:

* Partitioning of knowledgebase contents so that access can be limited to appropriate users and roles;

* A single indexing mechanism to ensure consistent retrieval results across languages, and a single storage mechanism to eliminate synchronization of knowledgebases across languages; and

* A retrieval mechanism that is effective for novice users, such as customers themselves, and experts, such as senior field technicians.

The Retrieval Swamp 

Retrieving solutions from a knowledgebase has relied on two main technical approaches until recently. Unfortunately, both approaches have considerably limited the usefulness of a knowledgebase in numerous situations. Users too often feel themselves lost in a swamp as they try to locate pertinent material in the knowledgebase to solve a customer's problem. We'll have a look at why these approaches have resulted in frustrated users and have curtailed the hoped-for financial returns from these systems.

The first approach uses some form of keyword matching. A user types in a problem description or a sequence of keywords. For example, if a customer was having a problem with an LCD projector, she might describe a problem as "a yellow spot in the center of the screen." In this phrase, the words "yellow," "spot," "center" and "screen" would be salient. The knowledgebase would produce solutions that match these keywords and perhaps rank them by some criteria such as frequency of use or user feedback on usefulness of the solution.

The first problem with this approach is that the user may not choose the right keywords. The same problem might be described by another user as a "brownish area in the middle of the screen." This phrase would produce an entirely different set of keyword matches. Today's approach strives to eliminate the dependence on word choices to deliver more consistent, accurate results regardless of how the problem is described by the user. This is an extremely important capability in self-service applications where the user, a customer, has likely not been trained on which words to use to describe the problem.

The second limitation of the keyword matching approach is that the user is typically presented far too many or far too few results. In the example above, "yellow" and "spot" may produce 80 matches, while "yellow," "spot," "center" and "screen" may produce zero matches. In the first case, too many responses, the user will have to comb through the list to sort out irrelevant material. This is what we all experience when we sort through lists of FAQs to find a solution to a problem or an answer to a question.

A related issue is the "under-specified problem." The user may describe the problem above as "poor image quality." This problem could have myriad causes. The only way to solve this problem is to ask a few intelligent questions. If your self-service Web site delivers nothing but keyword matches to the words "poor," "image" and "quality," you will be delivering a frustrating experience to the user and will probably end up with a call to the contact center to solve the problem. At that point, we can only hope the call center agent has better tools to help pinpoint the source of the problem and identify an appropriate solution.

The other primary knowledgebase retrieval strategy can deliver a less frustrating experience for the customer, but at much greater cost to the support organization. These are the common decision-tree-based knowledge systems. The user navigates through a sequence of questions until the system identifies a solution to the problem. The discriminating questions are developed essentially by hand. While the user experience is more positive because the system is more effective at pinpointing an appropriate solution, the long development time and costly demand on resources to maintain these systems have led many large support organizations to abandon the decision-tree approach.

A new technology, dynamic induction, has been successfully applied in some advanced knowledge applications in customer support organizations. This approach enables the system to ask questions to pinpoint the root cause of a problem, but does not rely on hand-crafted decision trees to generate the questions. Key advantages of this approach are that updates to the knowledgebase can occur more quickly because the bottleneck of special skill to edit decision trees is not required, and the users themselves can navigate the knowledgebase without following a pre-ordained sequence of questions. The ability to skip questions and resume along a more productive line will enable users to use intuition, hunches and their own good judgment in seeking a solution from the knowledgebase.

Lowering The Cost Of Ownership Of Knowledge Solutions

Decision-tree-based knowledge systems have been plagued by high development costs and high resource demands for maintenance of the knowledgebase. On the other hand, the newer keyword search approaches are limited in their effectiveness at problem-solving, which means the savings expected from their use has been limited.

Both of these first-generation approaches required a great deal of effort to deliver and maintain multilingual knowledge resources, and many early approaches were suitable for only one or another service delivery channel. For these reasons, the issue of manageability is the central criterion to examine when comparing approaches to knowledge-empowering the support organization, particularly when support through multiple service channels and languages is a requirement.

Efficiency And Effectiveness Result In Cost Savings

At the end of the day, delivering the right information at the right time in the right place is what will maintain satisfied customers. Quickly delivering an accurate solution to a customer's problem, regardless of the language or the interaction channel used, is a central goal of any support organization. Doing so cost-effectively is the goal of every enterprise. The support interaction hub is a concept that should be closely examined by organizations seeking to knowledge-enable support operations in a manageable, cost-effective way.

For information and subscriptions, visit www.TMCnet.com or call 203-852-6800.

BY Dominique Chatelin, Kaidara Software

Dominique Chatelin is president and CEO of Kaidara Software (www.kaidara.com). Kaidara is a provider of advanced support solutions that accelerate the delivery of consistent and accurate multichannel, multilingual customer support services. Chatelin brings more than 20 years of expertise in both management and sales at leading organizations, including his previous positions as president and COO of iMediation and VP of worldwide e-commerce sales of Netscape Communications, and has also held senior management positions in Europe with Sybase and Computervision.