[Red Herring] Generation Now Gets Up to Speed

[Adam Rifkin]

Nice quote from Rohit: 'CRACKING THE BULLWHIP. "The real-time enterprise is
becoming more real today because now we have the processing power, the
bandwidth, and enough open standards to make it happen," says Rohit Khare,
cofounder of real-time software startup KnowNow.'

Important thing to remember is that store and forward architectures are
capable of lots of important things, even though email *feels* like "real
time" sometimes.


http://www.herring.com/index.asp?layout=story&channel=10000001&doc_id=540019
854

Generation now gets up to speed
By Om Malik and Justin Hibbard 
Red Herring
July 30, 2001

This article is from the July 15, 2001, issue of Red Herring magazine.

During the first quarter of this year, while technology investors were
fretting over the Nasdaq's slump, something else was happening that hadn't
happened in five years: the output per hour of U.S. workers in the nonfarm
business sector (the traditional measure of productivity) was lower than in
the previous quarter -- 1.2 percent lower, to be exact. When word of this
development got out in May, it sent economy watchers into a tizzy,
reigniting long-standing arguments about whether investments in technology
make companies more productive and whether these incremental improvements,
in turn, produce a sustainable rate of productivity growth for the U.S.
economy.

Though the argument remains unresolved, there is growing evidence that
technology has made companies more productive than they were a decade ago.
>From 1995 to 2000, productivity at nonfarm U.S. businesses rose to an
average annual rate of nearly 3 percent, up from 1.4 percent during the
previous 20 years. During the same half-decade, U.S. corporations spent an
average of $365 billion a year on technology, about 70 percent more than in
the first half of the '90s.

Now, amid a slowing U.S. economy, corporations are eager to maintain these
productivity gains. Their demands are reviving interest among technology
entrepreneurs in a well-worn idea called real-time computing. In this model,
a company's information systems function like a 24-hour live camera on its
operations, instantly alerting managers to changes in inventory,
availability of supplies, and customer demand.

The concept dates back to at least the '80s, when it was first touted as a
way to track inventory at department stores. Advances in three categories of
technology are behind real-time computing: cheaper and more powerful
processors, networking equipment, and software. Computing power --
especially in servers -- is keeping pace with Moore's law. Faster networking
equipment has provided a tenfold increase in available bandwidth at speeds
unimaginable ten years ago. Swifter networks have in turn given rise to
software applications that can pump bigger loads of data across networks
with greater frequency. According to its proponents, real-time computing is
no longer limited to piecemeal improvements in supply-chain management and
demand forecasting. Instead, they say, the technological improvements it
promises will lead to fundamental changes that boost sustainable
productivity growth.

One area where real-time computing can improve productivity is in the supply
chain. In February, U.S. Federal Reserve Chairman Alan Greenspan told
Congress, "New technologies for supply-chain management and flexible
manufacturing imply that businesses can perceive imbalances in inventories
at a very early stage -- virtually in real time -- and can cut production
promptly in response to the developing signs of unintended inventory
building." And real-time computing has applications beyond controlling
inventory -- financial accounting, for starters, whereby a company may close
its books in a matter of hours instead of days. The cost savings and
efficiency gains of such a scenario are obvious, perhaps even too good to be
true.

The best existing example of real-time computing is the Nasdaq Stock Market.
Every change in stock price is multicast, or sent simultaneously, to
millions of users -- traders, in this case. Stock prices are then matched
with content -- mostly news like analyst upgrades or decisions of the
Federal Reserve -- from sources like Reuters or Bloomberg. The trader who
has subscribed to information relating to a specific topic -- say, IBM stock
-- can react to the information by buying or selling Big Blue's scrip.
"That, in essence, is the real-time enterprise, where events drive the
decisions of the business," says Vivek Ranadivé, the CEO of Tibco Software
(Nasdaq: TIBX), a developer of real-time computing software.

NOW, OR LATER

To its supporters, real-time computing is a kind of corporate utopia. They
imagine that executives will become as well-informed as stock traders about
their businesses and make real-time decisions. According to this vision,
companies that use real-time computing are transformed into "real-time
enterprises," where every business process takes into account
up-to-the-minute information and its execution is immediately measured
against the overall operations of the corporation.

Such a scenario, however, is probably a decade away. But many corporations
are moving in fits and starts toward real-time computing. Rushing to supply
the necessary tools for these endeavors are scores of technology startups.
During the next four years, U.S. corporations are likely to spend $576
billion on software, according to IDC, a research firm. A sizable chunk of
that money will be spent on real-time technologies -- enough to create a
healthy market for startups and established software companies developing
tools that provide up-to-the-minute information to businesses.

The expected size of the market has enticed some of the biggest names in
venture capital. Vinod Khosla, a partner at the VC firm Kleiner Perkins
Caufield & Byers, has made real-time computing his main area of investment,
funding software companies like KnowNow, Zaplet, Oblix, and Centrata. "When
companies become real-time, they improve service performance while reducing
costs dramatically," Mr. Khosla says with enthusiasm -- rare for a VC known
for his disdain for hype -- contending that 1 percent of sales spent on
technology brings 1.5 to 2 percent savings in general and administrative
expenses. Mr. Khosla says that by the end of this decade, corporations will
spend nearly 10 percent of total sales on technology, up from 3.5 percent
currently: "You will not be competitive five years from now as a corporation
unless you have near real-time information architectures inside the
company."

Seeing the same opportunity is Promod Haque, a general partner at Norwest
Ventures, which is also targeting real-time software. His firm recently
invested $6 million in PowerMarket, a maker of real-time supply-chain
management software in Belmont, California. Mr. Haque says PowerMarket
differentiates itself by tying data from a company's manufacturing processes
into the computing systems of its customers and suppliers. The key benefit
of the shared data, he adds, is that it "helps companies make powerful,
smart purchasing decisions."

Determining the amount of venture capital flowing toward real-time computing
is next to impossible, because such a figure would have to encompass
hardware, software, and networking technologies. And ancillary markets are
affected, too. Moreover, it isn't only startups that are aiming to sell
corporations the real-time tools that they need. Large publicly held
companies like Oracle (Nasdaq: ORCL) and IBM (NYSE: IBM) are also in the
game, as is hardware maker Compaq Computer (NYSE: CPQ), which has formed the
Zero Latency enterprise group to examine the opportunity.

CRACKING THE BULLWHIP

"The real-time enterprise is becoming more real today because now we have
the processing power, the bandwidth, and enough open standards to make it
happen," says Rohit Khare, cofounder of real-time software startup KnowNow.

The combination of these technologies represents a clean break with an older
approach called batch-mode computing. Batch mode got started after World War
II, when computers were first catching on in business. In that method,
companies stockpile data for days or weeks in a central computer before
sending it -- in a batch -- to other firms. Retrieving, say, a company's
latest sales figures means waiting a week for month-old data. "In batch
mode, the answer to most questions is, I'll get back to you," says Roger
McNamee, a general partner at the VC firm Integral Capital Partners. "In
real time, you know immediately." Real-time applications are "always on," he
adds.

Batch mode works fine for tasks like accounting or analyzing customers'
histories. But it works less well for ordering supplies or managing
inventory -- jobs that, if mishandled, lead companies to miss earnings
estimates. Say a cosmetics retailer studies a batch of month-old sales data
and notices there's been a run on lipstick. Its purchasing department,
convinced it needs to stock up, sends to the lipstick manufacturer an order
that's twice the size of its usual monthly request. The manufacturer sees
the big number and, not wanting to get caught short on supply, sends even
larger orders to its suppliers of red dye and lipstick tubes. Soon several
companies have overordered and are stuck with excess inventory.

This phenomenon is called the "bullwhip effect," a term coined by Hau Lee, a
professor at Stanford Graduate School of Business. "Sometimes the bullwhip
effect happens because people are panicking," Mr. Lee says. When suppliers
can't see their customers' most recent data, they assume the worst and
overcompensate. But if a manufacturer can monitor a retailer's sales as
they're happening, it may churn out only as much merchandise as the retailer
requires. And if the manufacturer's suppliers see retail sales, too, they
can ship just enough materials. Outside of the financial markets, few
companies have built systems capable of this.

Instead, some companies have what are called just-in-time inventory systems.
Companies have been attempting to build these inventory systems for at least
ten years. The results are mixed. They require a lot more than technology
(see "Debunking the myth of zero inventory"). And companies must be willing
to share sensitive information -- the kind that determines the prices at
which suppliers will sell goods to them. But the fear of coöperating too
much with competitors and the expense of network security have kept
companies from sharing the information necessary to benefit from real-time
computing's full potential. Until these problems are overcome, no company
will make the leap from just using real-time computing to becoming a
so-called real-time enterprise.

Consider Cisco Systems's (Nasdaq: CSCO) much-publicized woes earlier this
year, when its valuation plummeted from $500 billion to about $160 billion
in a matter of months. The company bungled its own sales and inventory
forecasts, and ordered far too many parts. Because of misjudged demand, its
net inventory rose by $577 million from the first quarter to the second. The
result: Cisco will write off about $2.5 billion's worth of inventory.

A close cousin of the just-in-time inventory system is what's known as
build-to-order. In this model, popularized by Dell Computer (Nasdaq: DELL),
companies don't build a product until a customer has paid for it. Upon
payment, the company orders the required parts -- say, a Pentium III
processor and 10 GB of memory -- from suppliers. When the parts arrive, the
company assembles and ships the product. In this way, it avoids carrying any
excess inventory. Mr. Ranadivé at Tibco describes this as "cash-to-cash"
business, in which there is little or no delay between the time a company
spends cash to produce goods and the time it receives cash from its
customers.

STOCK ANALYSTS

Software makers like Manugistics (Nasdaq: MANU) and its arch rival, I2
Technologies (Nasdaq: ITWO), are best known for their supply-chain
optimization engines, programs that analyze historical patterns of supply
and demand and recommend how much a company needs to stock up. The
applications have been widely available for five years, and there is little
room for new brands. But startups like Rapt, Saltare, SeeCommerce, and
Tilion are building applications that supplement -- and, in some cases, may
improve upon -- optimization engines.

The newcomers' applications perform analytics. That is, they pull
up-to-the-minute data from several sources, including supply-chain and sales
systems. Because they usually crunch less data than optimizers do, they can
pull in the latest numbers hourly or daily and process them quickly, letting
companies respond to last-minute changes in supply and demand. "If
optimizers present a what-if scenario, analytics applications present what
is," says Paul Albright, CEO of SeeCommerce.

Raymond Lane, a general partner at Kleiner Perkins, and for many years the
chief operating officer of Oracle, believes analytics applications will
change the role of buyers in corporations. "In the not-too-distant future,
we'll see companies that don't have order-management departments to do
transactions," he says. "They'll just be for analyzing how the company is
doing." Mr. Lane recently funded a startup called Valdero, which makes
software for analyzing supply chains in real time.

Of course, the data in an analytics program is only as timely as the source
from which it's pulled, which is often an application from I2 or
Manugistics. Terry Austin, an executive vice president at Manugistics,
cautions against companies that use the term "real time" too liberally.
"Real time is one of those terms that gets thrown around a lot," he says.
"You have to question what they mean by real time, because they're sourcing
the same data as everyone else. If it's an inventory system that gets
updated from around the world, you have to get the data from somewhere, and
that has latency built into it."

Attacking this latency are other companies developing technologies that
generate real-time data at the source. Savi Technology, for instance,
attaches sensors to containers and trucks to track the shipment of goods.
Radio transmitters beam the data to a computer loaded with Savi's software,
which manages inventory by tracking shipping along with manufacturing and
purchasing.

Software that quickly transports data from its source to an application is
also essential in real-time computing. Companies like Tibco and SeeBeyond
Technology (Nasdaq: SBYN) (formerly Software Technologies) have provided
such software for more than a decade. TXU (NYSE: TXU), a power utility in
Texas, is using SeeBeyond's software to send updates over the airwaves from
home electricity meters to its headquarters. With real-time information
about power consumption, the utility knows precisely how much electricity to
buy on spot markets. Power utilities in California have faced severe
shortages in the past year partly because of inaccurate forecasts of demand.

Making real-time computing equipment work with a company's older information
systems is daunting. One approach is to divide a corporate computing system
into four layers. At the bottom are existing applications like enterprise
resource planning (ERP) and customer relationship management (CRM) systems,
as well as existing databases and custom-built applications. These are
connected to one another and to the other layers using a software adapter,
typically called an information bus. Different types of process-specific
software reside in the next layer. These are event-driven programs, meaning
they execute business processes based on up-to-date information from the
computing system. At the very top is a layer of software that sends alerts
by email, or multicast, to employees, vendors, and customers regarding, say,
their orders or their accounts. The four layers are further tied together by
management software that routes information to employees and to the
process-specific programs, at the same time keeping information
synchronized.

Addressing various needs of such a real-time system are at least two dozen
startups, including Bang Networks, FineGround Networks, SeeCommerce,
Vianeta, OpenDesign, KnowNow, and Oblix. Picking those that will be
successful is difficult; success may depend on their ability to make their
software work with software from established companies like IBM and Oracle,
which have very large customer bases that have already invested millions in
the use of their software. And not only will they have to coöperate with
these giants -- they could wind up competing against them.

ONLY TIME WILL TELL

Doubts linger about the potential effect of real-time computing systems on
productivity. For one, there is the possibility that the technological
challenges of integrating disparate and aging computing systems with a
real-time computing system are simply insurmountable. Then, there is the
notion that real-time computing is merely an evolutionary technological
development akin to the so-called advances made in ERP systems during the
'90s. Corporate efforts to use real-time databases and online transactional
processing methods to improve ERP systems consumed hundreds of millions of
dollars, often for naught. The paybacks "are smaller and less dramatic than
we've been led to believe," says Stephen Roach, chief economist at Morgan
Stanley. "They're concentrated largely in back-office functions, allowing
companies to process more volume per unit of work time than ever before. But
that has a relatively low payoff."

Nevertheless, there is a strong possibility that real-time computing offers
the chance of technological advancement, a restructuring of business
processes, and fundamental economic change. If real-time computing goes
mainstream, entire industries may consist of highly efficient real-time
enterprises. Such structural changes could boost economic productivity. In
this scenario, the question would be when this boom in productivity will
occur. "Structural changes in the economy never result in immediate changes
in productivity," cautions Mr. Roach. But, as the U.S. economy's performance
from 1995 to 2000 shows, these structural changes spurred by technology can
lead to increased productivity growth.

Real-time computing will likely boost productivity at individual companies,
but viewing these improvements as directly leading to a substantial, overall
boost in economic productivity calls for caution. Such a finding assumes
that companies begin sharing "instant snapshots" of their respective
operations and creating more efficient infrastructures within their
industries. A more reasonable approach would say: not so fast. Competitive
pressures continue to prevent companies from sharing information, and the
technological hurdles in the way of a real-time utopia are high. Though
real-time computing is a hotbed of entrepreneurial capitalism today, its
macroeconomic fruits are ten years away, if it ever bears fruit at all.

Additional reporting by Michael Fitzgerald. Write to om.malik@redherring.com
and justin.hibbard@redherring.com.



----
Adam@KnowNow.Com

FoRK is going down soon for movement from LA to Santa Clara.  Probably
Wednesday.  Be back up a few days after that.  Probably Friday.  Maybe
Monday.  Aw, who knows.  I'm looking forward to using the time off to find
out again what sunlight looks like...