The sad story of a missed breakthrough (and
comments)
26 June 06
If Dragon
NaturallySpeaking were a person, rather than a piece of software, it
would resemble a hardworking, dedicated genius who had been severely
traumatized during its adolescence. The program’s history provides a
case study of the perils of commercializing breakthrough innovations.
Dragon Systems
existed as a company from 1982 to 2000. In 1997, after more than 18
years of work, Dragon released a pc-based voice recognition system
called “NaturallySpeaking.” The program won dozens of industry awards
and became a bestseller in a segment of the software industry that was
growing rapidly. IBM had a competing program, called “viaVoice,” and
Dragon played the role of the plucky upstart battling the giant IBM.
Dragon’s sales
tripled in one year, from 1997 to 1998, reaching $60 million.
By 1999, the
software was so popular that IDG, the publisher of the famous “Dummy
Books” series, released “Dragon NaturallySpeaking for Dummies.”
Speech Recognition for the Rest of Us, circa 1999
In June, 2000,
Janet and James Baker, the husband-and-wife owners of Dragon Systems,
sold their company to the Belgian firm Lernout & Hauspie (L&H) for $600
million in L&H stock.
Then the
situation rapidly deteriorated.
In August, 2000,
an article in The Wall Street Journal raised the specter of
financial impropriety at L&H. In November, 2000, an internal audit
revealed that the company had fabricated revenues of $277 million from
1998 to 2000.
By November,
2001, Lernout & Hauspie was bankrupt, the victim of fraudulent
activity. Both of the company’s founders spent a short time in prison,
and may still face additional charges. The stock that had been worth
about $75 a share in June of 2000 was trading at well below $1 a share
less than a year later.
Partly as a
result of this scandal, demand for pc-based voice recognition software
plummeted. According to a November 2001 article in PC Magazine,
sales of this kind of software declined by more than 50 percent from
2000 to 2001.
The technology
itself continued to have tangible value, although at much lower prices.
A company called ScanSoft bought all of L&H’s speech technology assets,
including “NaturallySpeaking,” in December 2001, for $51 million.
In 2005,
ScanSoft released the newest version of Naturally Speaking, version 8.
A colleague who works at ScanSoft (which, to further complicate the
story, changed its name to Nuance in
late 2005) recently gave me a copy of the software.
I’d used a
previous version of NaturallySpeaking in the late 1990s, when the
company was initially coming to prominence. Using the program again
prompted some painful memories of struggling with a brilliant piece of
software that was not quite ready for the mass market.
This newest
version is much better, however. Computer processing speeds have
increased and the software’s algorithms have improved, so that the
accuracy of the program is much higher than it has been in prior
versions. Experienced users can achieve an accuracy rate of 99
percent. It is quite amazing to be able to dictate a sentence into a
computer and have it appear as text in precisely the way intended.
That said, I
suspect that the program’s time for consumer success has come and gone.
This is true in large part because the mass market has moved on. The
large group of people who could most benefit from this kind of speech
recognition program, those with limited keyboard skills, now have a
variety of other ways of getting their thoughts into the computer, from
pecking out short messages on Blackberrys to communicating using the
slang of instant messaging. The alternative of putting on a headset
and talking to your laptop or desktop computer doesn’t seem very
attractive in the increasingly mobile world of 2006.
For the other
group of people who are already proficient at typing, there is little
gain in efficiency from a voice recognition program and a great deal of
startup cost. For example, I started writing this update using
NaturallySpeaking, but now find myself back in a more familiar
environment, typing on a keyboard.
As evidence for
this market decline, consider that the two dominant consumer voice
recognition programs, IBM’s viaVoice and NaturallySpeaking, once fierce
competitors, are now both owned by Nuance. Nuance itself is growing but
remains unprofitable.
Rather than
succeeding in broad-based consumer applications, voice recognition
technologies have made their mark in business-to-business services and
niche markets. Call center operators
use voice recognition technology to route customer calls.
NaturallySpeaking has specialized editions targeted at medical and legal
professionals, who still make extensive use of transcription services.
The history of a
program like NaturallySpeaking provides a case example of the “chasm”
theories developed by Geoffrey Moore. When a new high tech product or
service initially launches, the “visionaries” and “early adopters” are
eager to give it a try. But there are too few of these customers for
the product to achieve profitability. After initial launch, one of the
most important jobs for company management is to broaden the product’s
appeal, in order to reach the more conservative customers in the mass
market.
The best time
for NaturallySpeaking to have crossed that chasm was back in 2000, when
management problems and the fraud at Lernout & Hauspie crippled the
industry. By the time the industry recovered, the mass market had
moved on.
More
Information:
-
Jennifer 8.
Lee did a writeup of the situation with Dragon Systems in The New
York Times back in 2001. Here’s a
link.
-
For more on
Geoffrey Moore’s “Crossing the Chasm,” see the Wikipedia entry
here.
-
Nuance
Software’s website is
here.
-
For epinions
review of the NaturallySpeaking 8 software, go
here.
