If packaged speech recognition software is an endangered species then Lernout & Hauspie will be the first of the speech recognition vendors to be threatened with extinction. The Belgian software company has made a valiant attempt to fend off the competition by embarking upon a fast-paced acquisition strategy, buying 22 companies in two and a half years at a total cost of $340m, in a bid to shore up market share, linguistic talent and technological expertise in an ongoing battle with speech recognition supremos, IBM and Dragon Systems.
But while the company has managed to raise its profile in Europe, US success has so far remained elusive even though just under half of the company’s $100m revenues came from North America last year. IBM and Dragon Systems own the US market, says Rob Enderle, a speech recognition software analyst at Giga Information Systems.
Both IBM and Dragon Systems enjoy advantages L&H doesn’t. Development resources and marketing muscle aside, IBM can bundle speech recognition systems with its operating systems and PCs and push more flexible and cheaper deals to promote ViaVoice, its desktop natural language package. Dragon Systems can trade on its firmly established US branding, and its technologically sound products (NaturallySpeaking being its ViaVoice equivalent).
Speech recognition software has virtually become a commodity. According to market research company PC Data, the average selling price per voice-recognition title plummeted from $90 in September 1997 to $35.66 in September 1998.
But by far the biggest threat to L&H’s long term future is Microsoft. Two years ago the two peacefully co-existed under a licensing agreement where Microsoft agreed to use L&H’s text-to- speech software in future generations of its desktop operating system. Its deal with Microsoft was a coup since it would certainly provide the much vaunted market penetration needed to propel L&H’s speech recognition software into the mainstream. To cement its commitment, Microsoft took a 6% equity stake worth $45m in the company in September 97 and installed a Redmond executive on L&H’s board.
Although the equity stake still stands, the nature of the licensing agreement is now shaky. Microsoft has been quietly working on its own text-to-speech software for some five years and will embed its own speech synthesis engine into Windows 2000, the next generation of its desktop operating system. L&H is representing the news as a something it had anticipated all along. We expected Microsoft to integrate its own speech technology into future versions of the Microsoft operating system. But we can work on some 20 extra exotic languages that Microsoft won’t want to provide, says Ellen Spooren, senior VP of marketing at Lernout & Hauspie.
The ramifications are far more dramatic that L&H will admit. By the second half of 1999 when Windows 2000 is scheduled to ship, L&H will be relegated to a marginal role, providing text-to- speech software in languages Microsoft deems unimportant to the business world. (Microsoft and L&H co-developed a speech interface (SAPI) to make interoperability less of a problem.) Its ambitions of having L&H software on every desktop will be totally dashed.
Content from our partners
L&H’s only chance of future success is to dominate those markets where IBM, Microsoft and Dragon Systems are more vulnerable to competition. L&H will need to compete outside of the desktop market if it wants to stay in the game, says Jackie Fenn, a speech recognition analyst at The Gartner Group.
Vertical markets are where the company sees its future. We will play in markets where the number of users for a product are smaller and where the margins are higher, says Spooren, citing medical dictation systems and transcription software as key areas for future growth. L&H is already deemed to be the leader in medical dictation systems and is allocating more resources to both its medical and legal dictation research teams in a bid to grow that side of its business. The company also plans to play to its strength in ‘exotic’ languages (each linguist at L&H is fluent in at least five languages) so a push into the Far East is also expected.
In a bid to generate recurring revenues the company is also pursuing a strategy to position itself as a so-called ‘solutions’ provider, building itself a support and services organization. If L&H can providing training in the use of recognition or dictation systems, work with customers to provide particular customization requirement and provide support when difficulties occur, then the company can charge a percentage of license revenues for these services.
The company has set up three new divisions: Technologies and Solutions, Applications, and Services. The move came as part of a corporate re-organization at the end of January enacted to realign resources from recent acquisitions but more importantly to solve the problem of salespeople pitching different products to the same clients. Each salesperson can now sell all our products: machine translation, dictation and speech, says Spooren. Technologies will include L&H’s core technology division, language technology division and the intelligent content management group. Applications will deal with end user and retail sales and education. Services will house document creation, human and machine translation services and the new translator family of internet translators.
But a corporate restructuring and a vertical focus may not be sufficient to ensure L&H’s long term growth and prosperity. There is widespread belief that L&H placed too much on the Microsoft deal, riding Redmond’s coat tails to the desktop. This has cost the company dearly. Some analysts are already suggesting that Jo Lernout and Pol Hauspie don’t have the skill set or funding to make a profitable business. They should shape their company up as a technology sale, find the best buyer and then walk away with the money.
This article is part of ComputerWire’s M&A Impact information service. Some articles from the service are being provided to ComputerGram subscribers for a trial period only.
