The Information Edge

Technology, both in manufacture and in product use, is among the most powerful business tools. The development of technology is often expensive and not subtle. If used wisely it gives long- standing competitive advantage providing the firm with a price premium and market leadership. Used unwisely, it can back the business into complacency and doom it to repeat past mistakes.

Technologies are the means to do things. It covers day to day procedures of managing resources as well as the sophisticated application of scientific and engineering discoveries and principles. The value of technology is not derived from its elegance or sophistication but from the benefits that are derived from its use.

We have found three key principles that must be recognized in developing creative active technology strategies:

1. All Technologies Improve - Some technologies may improve slower than others. However, all technologies whether serving old established industry or serving new "high tech" applications improve over time. A corollary to this principle is that not improving existing technology is always dangerous.

2. Established Technical Routes Show Increasing Marginal Difficulty - The easy things are done first, the remaining improvements take longer. Maintaining a continuous technology improvement based on existing methods and approaches require increasingly higher resources. Relying on existing ways and directions for developing technology will become increasingly expensive. This leads to the need for continuous improvement in the means for and direction of technology development.

3. Unique Value Comes from Unique Benefits - Long standing technology value comes from unique user benefits. Marginal product and process improvements tend only to produce marginal benefits. It is from unique capabilities that competitive advantage is derived.

Together the three principles emphasize the need for an innovative R&D process aiming for major improvements. That process must be tied to the market place where earnings and value are obtained.

Understanding the distribution of price is critical for laying out effective market strategy. Price point distributions show the percent of products sold within a category below a given price. Normally we focus on consumer products at retail for this type of analysis. However, such end-use analysis is also critical to define product position in industrial markets.

Often only imperfect data is available to estimate the price point distribution. Typical data is shown in the figure below. The data is obtain either from consumer purchase diary data, outlet audits or from analysis of advertising, promotion and distribution. It is usually based on a highly limited sample and selected outlets.

The raw data produces distributions that are crude at best. These need to be smoothed inorder to interpolate values within ranges that data exists or to extrapolate values beyond the range measured. We have found that a log normal distribution fits price point data. The rationale for using the log normal distribution comes from the psychometric observation that differences in perception tend to be proportional or logarithmic. As such, we would expect that a normally dispersed average perceived values would be log normally distributed. The figure below shows the fit of the raw data to the log normal distribution using Microsoft EXCEL. The cumulative share is converted to a normal distribution using the NORMSINV function. The log values are obtain for price using the LOG function and the fit is done with simple regression.

The figure below shows a comparison between the fitted cumulative distribution and the log normal fit. This is fairly typical.

While we use the log normal for analysis of price points, it is also an effective model for other types of value distributions.

Would it not be wonderful if we could communicate with computers by voice? It has been the promise for many years and in some regards it is here today. There are two sets of issues: (1) the capability of computers to understand speech and (2) the translation of written materials into speech. Both hold great value for the severely handicapped. However, our interests here are in this capability as productivity aids. Here too, both capabilities offer some unique opportunities and some problems.

The ability to translate text to voice has been around for some time. The capability has been incorporated into Apples' Macintosh for over ten years. Monologue for WindowsÔ by Creative Labs (Sound Blaster) had been offered as a free included software package. DecTalkÒ available from Digital Electronics is reported to be the most advanced package of this type (unfortunately I have not had a chance to work with it.). These packages basically read text out loud.

The technology has been widely used in video and computer games. It has been promoted as a means of "livening up" applications. However, the most important use to me has been for editing. It provides a means of reading back written materials. Errors that could be identified only by a second party reading of the document are quickly apparent. Grammatical errors in my work have been reduced by at least 80% using this technology.

The remaining problems with the technology are the quality of the voice and the size of the dictionary. The voice typically is mechanical and it sounds like Stephen Hawkins is reading the text. Some words are incorrectly pronounced. This can be reduced by use of a custom dictionary. It is the size and universality of the dictionary that limits the usefulness of the software. While these problems may limit its application, I strongly recommend its general use for text editing.

Voice recognition is the ability of the computer to recognize speech. There seems to have been great strides in this capability in the last few years. Like text to voice, it can serve the severely handicapped. Our interest, however, is in its use as a productivity aid, specifically in its use for dictation.

Some twenty five years ago, dictation of reports became popular with the availability of inexpensive miniature tape recording equipment. However, with the advent of personal computers and the reduction in clerical staff, dictation has declined in use. Those of us that had developed the dictation skills have been looking forward to that capability on the computer. There are two types of dictation software, discrete and continuous speech. The discrete system is based on speaking discrete words with a pause between each word. Continuous speech software is more complex, requiring significantly more computing capacity.

I've tested both types of software available from IBM. True VoiceÒ is a discrete speech system available both for OS/2 and Microsoft Windows. ViaVoiceÒ Gold is IBM's continuous speech system. In addition, Naturally SpeakingÒ is available from Dragon Software. All of these packages require that the user "trains" the system. This is a several hour process. The discrete speech recognition system, the IBM True Voice, was difficult to use since it requires a peculiar way of speaking. It is, therefore, not recommended.

There are three limitations to using continuous speech recognition software for dictation: (1) speed of use, (2) recognition accuracy, and (3) extent of the dictionary. Speed is a major problem with continuous speech. While a 90 megahertz Pentium was adequate for discrete speech, the 200 Megahertz process was barely able to handle the continuous speech system. IBM and Dragon Software both claim in the order of 95% accuracy. I found ViaVoice less than 90% accurate even after the full training session. However, 95% accuracy would itself not be adequate for dictation. This would be one wrong word in 20. This situation is even worse in regards to specialized vocabulary. Special legal and medical dictionaries are available. However, none are available in other areas such as marketing or strategic planning.

While voice to text software is a coming technology, we can not recommend voice recognition as a productivity tool just yet. It isn't ready for prime time.

We have been debating whether we should make the newsletter an annual versus a biannual publication. I appreciate any comments and contributions from you. You can get in touch with me at Custom Decision Support, Inc., Phone (610) 793-3520. Fax (610) 793-2531 or E-Mail at genelieb@lieb.com