|
|
|
Tools of the Trade |
| |
|
|
Knowledge management and
customer support desks
As
more new, high-tech products hit the marketplace, the
pressure on help desks increases, and the need for
automation grows.
Among
the most well known uses of knowledge management are
data mining and customer relationship management. If you
accept one of the definitions of knowledge management as
being 'getting the right information to the right people
at the right time' then in this example 'the right
information' is a list of products with the highest
probability of being sold and a list of the customers
who are most likely to buy them, and 'the right people'
are those who make up the sales team. The right time is
usually 'now'!
However an alternative way of populating this
definition is with 'the right people' being the
customers and 'the right information' being the answers
to the customers' queries and solutions to their
problems. In doing so, we have just defined the role of
the support desk or help desk.
Growing pains The
problem help desks face is that, as your business grows,
so does your product line and your customer base, and so
does your support headache. It has been estimated that
90% of the products that high-tech companies sell have
been on the market for less than one year. This means
that the number of different product types that are in
active use is always on the increase and that the
products never get a chance to mature. There are always
lots of issues with them. 'You can't scale a help desk
against such an aggressive product growth,' says
Verity's VP Europe, Hugo Sluimer. 'Without good ways of
knowledge expansion and knowledge management you can
forget it. You need to automate the help desk.'
Automating does several things. It makes it easier for
existing support staff to find information. It also
makes more information available to the staff. And if
you give customers access to the right information, it
can reduce the load on the support desk (see case
studies).
So
how is automation achieved? One of the simplest ways is
to publish existing information on the corporate
website. For example, putting a FAQ (Frequently Asked
Questions) page on your website should, by definition,
mean that a reasonable proportion of common questions
get answered straight away. It has the advantage of
being low cost and available 24 hours a day, seven days
a week. But it has the disadvantage of being limited,
dumb, and unable to be updated automatically.
You
could publish customer query responses as individual
documents, and allow customers to search through the
documents using a keyword-driven search engine. However
the problem here is that the search engines don't take
context into account, and you can end up being guided
down blind alleys because the search software is not
intelligent enough to know, for example, the difference
between light (as in 'lamp') or light (as in 'not
heavy') etc. So the next step in the evolution of
information retrieval is natural language processing and
contextual awareness.
Use your own words With natural language processing, the context
in which the words are used is as important as the words
themselves. In fact, some of the natural language
processing software can even work in the presence of
spelling mistakes. Equipped with dictionaries, idiom
dictionaries and thesauruses, the software is able to
form relationships within documents and generate
patterns from them that can be matched to the patterns
within customer queries.
Case Study - KMS at
Nikon
When the European technical support desk at
Nikon, the Japanese imaging company, wanted to
improve their online help, they decided they
needed to go beyond a standard FAQ and use a
knowledge-based and intelligent front-end
software. 'We needed to address the changing needs
of customers, and the increasing numbers,' said
David Ward, manager of Nikon's European technical
office of their electronic imaging division. 'We
had a good set of information, but the problem was
how the customers could access it.' This was where
Knowledge Management Software's Deskartes
application came in.
Deskartes combines a natural language user
interface with a neural network processor. This
means that users can input questions and queries
in their own language and Deskartes will find them
an answer. 'It's important to allow customers to
express the problem in their own terms,' said
Ward.
In addition to helping the customers, the
Deskartes software also helps Nikon to tune the
knowledge base. 'We get visibility of the
questions that are asked,' said Ward, 'And we can
fill in the holes.' There are a number of ways
that this can be done, including updating a
solution, adding a solution, training the
knowledge base to find the correct solution, or
even updating a product. 'Deskartes is fundamental
in getting knowledge to customers,' declared Ward.
|
Most
vendors agree that natural language processing is more
accurate at determining meaning and locating relevant
documents than keyword searching. Users at the Nikon
site (see case study: KMS
at Nikon) are advised that if they use sentences
there is a 75% probability that they will find the
answer they were looking for in the top five suggested
solutions, compared to only a 55% probability that they
will find the answer in the top 20 solutions when they
use keywords.
Verity has similar faith in the natural language
processing approach. 'The navigation precision that we
provide is proven. It's far better than the average
search engine on the Internet,' says Verity's Sluimer.
'If users have a good query (and technical people
normally express themselves with the right words) they
will immediately navigate themselves to the right part
of the manuals, FAQs or solution reports.'
In
some cases the amount of information that a customer
might need to answer a query might not be appropriate
for transmission over the internet - a lot of technical
support information comes with diagrams, such as
exploded assembly diagrams. A more appropriate
publishing medium is the CD-ROM or, better still, CD-ROM
combined with the internet. (See case study: Verity
at Cisco.)
Gauging satisfaction One of the most frustrating things, from a
user perspective, about automatic help desks is that you
sometimes can't find the information you want, either
because you can't express the problem sufficiently well,
or because this is the first time that the problem has
arisen and so a solution doesn't yet exist. What you
need in this situation is a way to escalate the problem.
Case Study - Verity at
Cisco Like other high-tech companies, Cisco
found that they had to automate their help desk to
prevent their support costs escalating. Because of
the high volume of information traffic that some
of their help queries generated, they decided to
publish their help database regularly on CD-ROMs.
Each month 200,000 CD-ROMs containing a searchable
database of technical information as a
self-support application are sent out, both
internally to Cisco and externally to partners and
customers. Even updating them monthly is still not
frequent enough or good enough for the rate at
which products are changing.
So Cisco collaborated with Verity to
develop CD-Web publisher to enable them to
disseminate large amounts of information directly
to their users through CDs. The system combines
the inexpensive packaging and presentation
advantage of CD-ROMs with access to abundant
on-line information of the web. The Verity
solution provides the best of both worlds by
allowing the customer access to most of the
information they need direct from the CD-ROM.
Over the past few years, the load on the
Cisco help desk has grown by a factor of five but
they still have the same staff. Most of the
problems are solved by the customer base itself
and the number of live incoming calls to the Cisco
call centres has reduced dramatically, leading to
considerable cost savings. In addition there have
been $50 million in printing costs saved each
year. Product information, marketing materials,
technical specifications and manuals need only be
published and maintained once for the web, and
shared between CDs and the web site. If the user
is on-line to the internet then any information
that is more up-to-date can be provided
transparently when available. 'The user always
sees the most recent and most correct
information,' says Verity's Sluimer.
|
There are several ways that this can be done. For
example, each time customers get a response they can be
asked how well their question was answered. 'If the
question wasn't answered immediately we also want to
provide a way of escalating the problem,' says
BrightWare's Brian Tuller. 'We want to improve the
knowledge base and the operation of the software'.
Some
software systems let you log the fact that you didn't
find a solution and as soon as a solution is found to
match your problem, the help desk will email it to
you. A
similar approach can be taken when a customer problem is
reported that might affect other customers. The software
can be proactive and automatically email all other
customers who could have the problem.
Not perfect Automated help desks are generally successful
in what they do - they provide a front line of support
and they help keep support costs down, especially in
high-growth, high-tech companies. Automation works well
with technical customers who like to solve their own
problems. They are used to doing so and are normally
very successful. Although automated help desks are not
perfect yet, they are improving through the use of
natural language processing and self-learning
techniques.
Links http://www.kmsplc.com/ - Knowledge
Management Software http://www.verity.com/ - Verity http://www.brightware.com/ -
Brightware http://www.netg.co.uk/ - NETg
Case Study : Brightware at
Skydesk Skydesk provides an on-line storage
service - @backup - that automatically backs up
your files when you are on the internet. 'The most
important thing about a subscription-based model
like Skydesk is the first experience you have,'
says BrightWare's SVP of marketing and business
development, Brian Tuller. 'Because once you are
set up you are unlikely to change things. The
first few interactions are critical for
establishing customer satisfaction.'
Skynet uses the BrightWare 2000 eCustomer
Assistance software. 'When you ask it a question
it will answer based on what it knows about you.
So if you are a Gold-level customer you might
receive a different answer than if you were a
Silver-level customer.' Other clues used by the
software could be the wording of the questions,
the web page the questions are coming from, etc.
BrightWare 2000 can respond in a number of
ways - it can provide the answer in real-time,
pushing a webpage that already exists somewhere on
the site to you. It can route you to a chat
solution, to an email solution, or to a voice
solution. It can also generate a dynamic answer to
a question. So, for example, if you have a
question about a particular order or technical
configuration, it can create a web page for you
with various specific information embedded in it.
It can also automate the handling of email
requests. These typically get routed to an
automatic answering service because, 90% of the
time, the questions come from the web and are from
people who can't find an answer to their problem.
But most of the time the answer is there and they
simply haven't managed to find it. Typical
requests that get handled automatically are
customers looking for a particular backup disc,
account cancellation, credit refund, account
signup, error messages, wanting to know how much
space is left, how to start the service, how to
install the service, etc.
|
Profile - NETg and Just-In-Time Learning
NETg believes it will take the
evolution of the automated help desk one step further
when it launches its new just-in-time help desk software
later this year. The product is aimed at corporate users
who might normally call their support desk for help on
how to perform a particular task. Whereas the response
from a standard automated help desk to a query such as
'How do I print my Word document as a PDF file?' or 'How
do I cast in Java?' might produce a web page for the
user, the new NETg training help desk would search
through NETg's 50,000 learning objects to find the ones
that most suit the question being asked. Learning
objects are small learning modules of between three and
eight minutes in length that focus on a particular
topic. They can be strung together to build a complete
on-line training course or, as in this case, taken one
at a time. When the user inputs a query, the software
will send back a list of those objects that have a high
percentage match. 'When the user gets to the one which
they think best matches what they want,' said NETg's
Director of Strategic Deployment Jon Butriss (pictured),
'They click on it, it launches the learning object and
20 seconds later they're right in the middle of the
training course.'
NETg's learning objects have two features of that
set them aside from the standard web page response.
First, they can be customised using tools supplied by
NETg . So, if a particular company has its PCs set up in
a special way, perhaps with a particular background, the
learning objects can be likewise customised. And the
customisation is simple to do because of the generic way
that NETg's learning objects are created. Secondly, the
objects use simulation techniques to enhance the
learning. Users are not simply told the sequence of
steps to perform, they are led through them and given
the chance to perform them to ensure that they are
capable of performing them in the real situation. 'It's
not the final step,' said Buttriss, 'But we believe it's
the next step.'
|