Report No.23
*****************************************************
Japan Entrepreneur Report No. 23 September 2004
*****************************************************
- Reality jolts IT heavyweights
- Outsider breaks in by following rules
- Sigmund and the Wiz
- Bits and bytes
*****************************************************
Reality jolts IT heavyweights
What did heavyweight information technology service vendors Fujitsu, NEC,
Hitachi and NTT Data all do last fiscal year?
Lost a ton of money, even while their sales rose. What happened?
This earnings shock was looming for some time. Trying to escape their
mainframe-centered legacies and to keep up with the breathtaking pace of
technological change, the Big Boys aggressively solicited and won huge
open system-based projects. Unfortunately, they were ill-equipped to
complete them: they lacked Java expertise and the project management
skills crucial to managing hundreds of subprojects farmed out to
specialized contractors. Worse yet for these integrators, today much
project revenue goes right off the top to package vendors such as
Microsoft, Oracle and SAP.
Fujitsu, for example, posted a 10 billion yen loss in the year through
March 2004, thanks in large part to a soured banking system contract.
Meanwhile, corporate IT budget growth has slowed dramatically since 2001,
according to the Japan Users Association of Information Systems. But an
even deeper problem dogs the major players: they're starting to lose
their death grip on lucrative government contracts.
Big vendors like Fujitsu and NEC have long dominated Japan's huge
government markets. Bureaucracies' drawn out, single year budget
approval process leaves vendors little time to plan, pitch, write specs
and execute IT projects. Only the biggest players have the resources to
scramble and mobilize large teams for frenzied development during the
final three to four months of the budgetary year, when the work is
actually done. Short development fuses also mean heavier reliance on
packages over custom software.
Over the years, this recurring pattern spawned co-dependent--even
cronylike--relationships between giant IT vendors and government offices.
Some municipalities are now using multiyear budgeting practices that
unravel these cozy ties. And last year, a tenacious upstart taught the
Big Boys a tough business lesson: rules can trump relationships, even in
Japan.
*****************************************************
Outsider breaks in by following rules
"Their progress reports show what percent of the project is complete, to
two decimal places," marveled Saga City Mayor Kinoshita Toshiyuki in an
interview with a weekly news magazine. "Japanese companies never did
anything like that for us."
Kinoshita was raving about IT services vendor Samsung SDS, winner of a
large system order from Saga City. The Korean firm bested the likes of
NEC and Fujitsu, even though it didn't submit the lowest bid.
The city's bidding guidelines were clear: the selected vendor would 1)
build an open system, 2) disclose all source code and 3) complete the job
by March 2006. Yet every bidder except Samsung ignored at least one of
these three clear conditions. Some even pressured the city to change its
requirements, hoping long-standing relationships with the municipality
would carry the day. They didn't.
Samsung was the only vendor who met every proviso. And it didn't
disappoint after winning the job. City representatives were stunned by
Samsung's speed, quality and attention to detail; one client-side manager
called "unprecedented" the time and care Samsung took investigating and
documenting citizen registry and tax receipt work processes before
beginning development.
The scrappy outsider had a potent edge over its domestic rivals: a track
record building a similar system back home in Korea. That specialized
expertise, combined with a no-nonsense, rational approach to pitching and
executing work, made Samsung unbeatable.
Samsung's "outsider" experience offers clues to other IT vendors
struggling to win a bigger slice of Japan's $54 billion information
technology services market. Here are some takeaways:
Lesson 1. Specialize: Develop sector-specific expertise in health care,
finance, real estate, retail, government or another industry. Understand
that sector's unique workflow--from the employee/end-user viewpoint.
Lesson 2. Follow the rules: IT's essence is defining and encoding
business/decision rules in software and communications systems. Seek
prospective clients who take a logical, rules-based approach to engaging
vendors and follow their bidding guidelines carefully.
Lesson 3. Document workflow and project progress: Underserved Japanese
clients are astounded and delighted when IT vendors carefully create
workflow-based specifications, document progress and report
conscientiously.
All this will seem mere common sense to those with IT experience in India
or the U.S. But what's common practice overseas can still seem miraculous
in Japan--and that can put outsiders on the inside track.
*****************************************************
Sigmund and the Wiz
This summer we enjoyed lunchtime chats in Ebisu with software wizard Tim
Romero, longtime Tokyo resident and entrepreneur. We caught up with Tim
by e-mail this month to offer JER readers an overview of his latest
venture.
- How long have you been in Japan, and why do you favor Tokyo over, say,
Silicon Valley as the center of your work life?
I've been in Japan for 13 years. I stay in Tokyo more out of caprice of
circumstance than long-term planning. I've started and helped start up
companies both in Tokyo and in Silicon Valley, and while the atmosphere
between the two is very different, I can't say that I strongly prefer one
place to the other.
If I had to name one area in which I prefer doing business in Japan it
would be that business operates in a much simpler legal framework. This
is true both in terms of government regulations and contracts. I spend
more on legal fees in a month in America than I do in a year in Japan.
The work involved in running and developing a business, of course, is
much the same in either country-- in any developed nation, I suspect.
- Why did you choose SunBridge as an investment partner, and what non-
financial benefits does SunBridge offer startups like Wizaz?
When you first start out, a venture capitalist (VC) can do things like
provide industry access and help fill holes in your management team.
Experience marketing products in your industry and the ability to raise
significant funds in later rounds are far more important than the
relatively small amount of investment your company receives.
Of course all VCs will tell you the same thing to justify as low a
valuation as possible. In fact, I swear a lot of these guys are using the
same PowerPoint slides to make this point. I've found the only real way
to find out which VCs actually walk the walk is to talk to the owners of
their portfolio companies. And if you really want to get an understanding
of how you are likely to be treated, talk to some of the companies in
which things did not go as planned.
As you can imagine, I have quite a collection of venture horror stories.
The stories I heard from SunBridge's portfolio companies, though, were
universally positive. Even firms that went through hard times had good
things to say about SunBridge. When you add in their high exit success
rate, it was a no-brainer to go with them, even if we could have gotten a
higher valuation elsewhere.
- Did a particular problem or annoyance lead to developing Sigmund, the
engine on which Wizaz is based?
Sigmund began a few years ago when I downloaded a content-based spam
filter. It worked well in English, but not at all in Japanese, so I wrote
my own. In doing so I learned quite a bit and came up with some ideas I
wanted to try out. In a few months of experimenting, playing, and cursing
myself for not paying closer attention back in my university mathematics
classes, I had something that was not only far more accurate than my
original, but was completely language independent. From that point on it
has been a relatively simple matter of refinement and optimizing for
specific applications.
- In terms a non-techie can understand, what does the Sigmund Engine do
and what tasks does it accomplish for users?
Sigmund learns to imitate simple human judgments by watching how an
individual makes decisions. Sigmund is not "smart" in the sense of being
skilled at, say, evaluating whether poetry is good or bad. But he is
extremely good at things like determining to whom a newspaper article
might be of interest, whether an e-mail requires urgent attention and by
whom, or whether a given Web site violates corporate use policy.
When classifying information into six categories, Sigmund usually
achieves 90% accuracy after four examples, and reaches 98% accuracy after
about 40 examples. The accuracy then slowly improves to well above 99%.
Quite frankly, though, accuracy above 99% in real-world situations is too
difficult to measure since human beings aren't that accurate.
- How does Sigmund do that?
Magic! Well, not really. Until now, most work on decision systems was
rule based. Computers are great at enforcing rules, but humans are
terrible at coming up with them. The problem is that we don't think in
terms of rules. We match patterns.
Here's a simple example: None of us have any problems telling the
difference between an apple and a tomato, but it would be very hard to
write a set of formal rules that could differentiate between them. That's
because there are no "apple rules" in our brains. Even though we've
never seen this particular apple before, it is similar enough to all
others we've seen that we label it as such. No rules; patterns.
Computers, however, don't perceive the world as we do. They don't have
five senses. They understand ones and zeros and not much else. The basic
concept, which I admit was a bit of a leap of faith, was to build
patterns from a stream of data and then compare that to patterns
generated from other streams.
The Sigmund engine uses statistical techniques to compare the patterns,
so strictly speaking it is not a neural network or artificial
intelligence. The results are impressive, though, and our approach
results in much lower resource requirements. Sigmund can process about
two megabytes of data per second on a standard desktop PC.
- Does double byte character handling still pose challenges in
applications like this?
Not at all. It's all ones and zeros. If you know the target language, a
bit of preprocessing can be useful, but the Sigmund engine neither knows
nor cares what language it is analyzing.
- Working remotely with programmers in Eastern Europe and elsewhere,
you've achieved something close to the "virtual office." How does the
experience match up with your expectations?
I've had a lot of experience with software outsourcing, but this is the
first time I've tried to manage a startup this way. Quite frankly, I
don't recommend it. Things have worked out well, and most of our
development will continue to be done in Eastern Europe.
The early days of a software company are chaotic. Requirements change.
Specifications change. There is a sense of teamwork, everyone contributes,
and some of the most important information exchange is informal. This
social dynamic simply does not occur remotely. The overhead and formality
required to integrate a small, remote team into this kind of system more
than offsets the cost savings. Fortunately, things worked out well. But
it took a lot of blood, sweat and profanity to get us to this point.
- You are now a serial entrepreneur--Wizaz is your second startup. Tell
us about the post-sale experience with your first company.
Actually, this is my third startup but I don't talk much about the first
one. The sale of my last company went pretty well. There were rough spots,
but at the end of the day, our technology was fully integrated into their
product line and they are now successfully developing and selling the
software even though most of the original team has moved on.
I have to admit, however, the post-sale experience was hard on me. One
week I was working with a resourceful, fanatically dedicated, 16-person
team, and the next week I am on the board of directors of a public
Japanese corporation where my job is primarily political. I'm not really
a political animal by nature. On top of that, we're talking Japanese
politics being played out in Japanese--it wasn't something I was
particularly effective at. I departed on good terms soon after the
integration of the companies was complete.
- What does Wizaz mean?
Wizaz is an alternate spelling of an ancient Indo-European word that
means "knowledge of hidden things."
Tim Romero
http://www.wizaz.co.jp
*****************************************************
Bits and bytes
Two Japan-based SunBridge investees went public in 2004: Do-it-yourself
on-line research service MACROMILL (www.macromill.com) and micromotor
manufacturer Shicoh Giken (www.shicoh.com). Our informal survey found
these companies ranked fifth and 14th worldwide in post-IPO market
capitalization. World-class IT ventures are indeed coming out of Japanc
"How to succeed as an entrepreneur in Japan"--sound intriguing? Hamaguchi
Naota will reveal all Tuesday October 5 at the Entrepreneur Association
of Tokyo meeting. See <www.ea-tokyo.com> for location and details.
Next month I'm all about "human resource management": hiring, training
and helping people be happy at work. Stay tuned...
Tim Clark
Senior Fellow
SunBridge Corp.
Voice (U.S.) 503.235.4419
Fax (U.S.) 503.235.4429
clark@sunbridge.com
*****************************************************
Address change and unsubscribe instructions
To change your subscription address, visit:
<http://www.japanentrepreneur.com/change.html>
To unsubscribe from Japan Entrepreneur Report, visit:
<http://www.japanentrepreneur.com/unsubscribe.html>
*****************************************************
Copyright 2002-2005 Tim Clark
Reproduction in whole or in part without express written permission is
prohibited, but feel free to pass along or quote with URL
(www.japanentrepreneur.com).
<< Archive | Page Top | Home
Copyright © 2002-2005 Tim Clark All rights reserved
