StrandWare Delivers
Expert Technical Support
on the Web with Amzi!
Company Background
StrandWare,
Inc. is a leading developer of label design, printing, and tracking software
for worldwide industrial and business applications. StrandWare's family
of products has helped bar code technology evolve from supermarket and
stand-alone applications to high tech, enterprise-wide solutions. Asset
management, time and attendance, inventory control, and a myriad of other
applications have been made more efficient through the use of proven bar
code technology. StrandWare software has enabled companies from small
manufacturers to Fortune 500 retailers to reduce labor costs by 30% and
achieve an 80% reduction in inventory returns.
Quality products and a strong commitment to customers has enabled StrandWare
to increase its installed base to over 110,000 users in 80 countries around
the globe. The commitment to customers remains Strandware's top priority,
and servicing their large customer base has required both excellent people
and products.
These fundamentals are supplemented by new technologies that keep StrandWare
ahead of its competition. This article describes one such technology,
an automated expert system based support system, available to customers
on its Web site.
Support Problem
As with most software companies, many of the technical support issues
revolve around problems that keep repeating. In particular, Strandware's
business requires knowing a lot about all the various printers they support
and how to set them up for optimal performance with their products.
They will often, for example, get customers having similar printing problems
on a particular brand of printer, and that problem can be resolved by
knowledgable twiddling of the configuration. The trick is making sure
all of their support people can find the right answers for all the possible
combinations of operating systems, printers and drivers that their customers
might have.
Support Solution
A Web-based technical support solution provides a number of benefits
for both Strandware and its customers. First, and most obviously, it can
directly solve a certain percentage of customer support issues. This means
the customer got a quick solution, and tech support time is saved for
the harder issues.
Second, it acts as a filter for human technical support. When a customer
does not get a resolution to a problem, an e-mail is sent to technical
support, containing all of the information the customer entered while
describing and trying to resolve the issue. This means the technical support
people are not wasting either theirs or the customers time gathering preliminary
information.
Third, the system gathers the data from the Web-based support sessions,
providing a valuable resource for analyzing customer problems and configurations.
This data can be used to directly drive product development and documentation
in directions that will increase future customer satisfaction.
Over the last two months they have been averaging close to 100 technical
support sessions per month on the Web site, and about 80% of those have
resulted in a resolution for the customer. The other 20% have triggered
e-mails to technical support.
Operation
When the customer starts a support session, they are asked for some initial
information, such as contact information and the product they have a question
about. The system uses that initial information to filter the rules that
might apply, and gathers additional information as needed in its search
for a resolution.
Figure 1
Figure 1 shows a question from the middle of an automated support session.
In this example the user is looking for help with a smudging problem on
a printer.
That particular session led to a relatively long answer, not shown for
space reasons, that detailed the settings that could be used to control
the quality of printing of various printers. The particular problem is
indicative of the type of expertise StrandWare must have to support it's
customers.
Figure 2
Figure 2 shows an answer from a different session that is indicative
of the sorts of support issues all software vendors in the ever-changing
Windows world have to cope with.
The Web pages for both figures 1 and 2 were dynamically generated by
the expert system in response to the data entered by the user.
Implementation
The system was implemented using Amzi!'s WebLS product. WebLS is a simple
expert system tool, designed specifically for deploying expert systems
on the Web. It carries on a dialog with the user, as do most expert system
tools, but the dialog in this case is all dynamically generated HTML.
The knowledge base is composed of questions, rules and answers. An initial
goal starts the system searching through the rules for ones that can produce
an answer. The conditional parts of those rules might refer to facts whose
values can be obtained by asking questions of the user.
Here, for example, are some questions from the knowledge base that are
presented to the user when the system is on the scent of the 16-bit ODBC
problem shown in figure 2:
question('second edition', [ prompt = $Is it Second Edition or have you installed the Y2K update?<BR>$, ask =yes_no, askAfter = ['symptom'] ]).
question('BackTrack Version', [ prompt = $What version of BackTrack are you
running?<BR>$, ask = field, askAfter = ['symptom'] ]).
Here is a rule that might drive the asking of those two questions. If
the reported symptom was as indicated, then this rule is in contention
and the questions would be asked. Otherwise, they wouldn't be asked.
if 'symptom' = 'Label Designer hangs on converting databases dialog' and ('BackTrack Version' <= '4.10')
and 'second edition' = 'yes' and (operating_system = 'Windows 98 Second Edition' or operating_system = 'Windows 98') then problem = 'Win 98se'.
And here
is the knowledge base code that generates the answer in figure 2:
answer('Windows 98se', problem, [ text = [$The version of Label Matrix you have (v. 4.60 or below) is not compatible under Windows 98 Second Edition or if you installed the Windows98 Y2K update from the Microsoft website. The reason Label Matrix is not ...
Here's a segment of the larger answer that would result from following
figure 1 to its conclusion, including HTML formatting, that helps the
user with smudging:
answer('Ink spread', problem, [ text = [$There is a way to control the ink spread on Laser and Inkjet printers by using the following commands in the products INI file. See list of products below for the appropriate INI filename. <BR><BR>BCAddToSpaces=n<BR>BCAddToBars=n<BR>Where n is the number ...
The StrandWare knowledge base currently contains over 100 of the most
common solutions.
Additional Benefits
The statistical reports from the system give a good indication of the
types of problems and equipment users have, as well as indicating how
well the system is resolving issues. For example, the current statistics
indicate that the largest area of tech support questions revolve around
printing issues, and the second largest area around label design issues.
The system also shows that technical support issues are relatively flat
across the customer base, with no particular customer having a larger
than normal number of sessions, and that Windows 98 is the most heavily
used operating system among the customers, and Zebra is the most popular
printer.
Future Directions
Based on the success of this system, StrandWare is considering two other
applications of expert system technology. One would be for sales advisors,
to help prospects pick the right products for their application. The other
would be an internal system that routes unresolved technical support issues
to the best-suited person for dealing with the issue.
You can learn more about StrandWare at www.strandware.com, and more about the WebLS
product at www.amzi.com.
This story was originally published in the Jan/Feb 2001 issue of PC
AI Magazine titled "Strandware Web-Based Expert System for Technical
Support" by Mike Strand. The magazine can be reached at PC AI, 3310
West Bell Rd., Suite 119, Phoenix AZ, USA 85023 Tel: (602) 971-1869, FAX:
(602) 971-2321, E-Mail: [email protected], Web: http://www.pcai.com
|