Research/Development magazine, July 1976

Having told how to manage projects to best advantage (yours) Putt now discloses the secret of selecting the best project (for you) and how to foresee the best time to get out of it

The Successful Technocrat 4: The S-curve law

Archibald Putt

The Laws of Crises and the Law of Failure, described in earlier installments, provide a formal structure on which the technologist can base his management of high-technology projects. But what about the selection of projects? Is there any methodology that can be followed to select projects likely to succeed?

There is little advice that I or anyone else can offer in this regard. The success or failure of a project is largely determined by technological factors that are not understood until after the project is well underway. The probability of success for advanced development projects, at their inception, is only about 30 per cent, and there is no general method for predicting which ones are most likely to succeed.

It has been rumored that a Harvard graduate student of management became intrigued with this problem and investigated a technique in which random numbers were assigned to research projects. The success or failure of these projects was then predicted using a formula based on the numbers carried by horses in the win, place and show columns at Belmont. News of the remarkable success of his method, as compared to the more heuristic methods now used by the research directors themselves, was said to have been suppressed by a clandestine International Association of Research Directors. The members of this august group rightfully feared that their own hard-won positions as technology soothsayers would be paced in jeopardy if the success of the method became generally known.

It occurred to me that if one good approach may have been suppressed then there could be other concepts worth considering that had also been suppressed. With this in mind, I proceeded to interview research directors and other high-level managers of technology. These interviews were, however, unproductive. There appears to be a paucity of ideas on how research or development projects should be selected.

The most universal suggestion I got from successful managers was, "Just ask me, and I'll tell you." However, when I queried any one manager about the quality of advice I should expect from one of his colleagues, the response was usually noncommittal: all too frequently, it was accompanied by a bit of a sneer.

Fortunately, I did not terminate my quest here, but turned my attention to more promising areas. The stock market was particularly interesting because the problem of selecting a good stock is quite similar to the problem of selecting a good research project. Selection of stocks on "fundamentals" is generally not successful unless the fundamentals include "inside information." In technology, however, all the expletives uttered in God's name do little to make available to man any "inside information" that He may possess.

Because most stock market services don't have inside information either, and would hardly share it if they did, they base their recommendations on "technical factors," relying primarily on curves of past performance projected into the future. Even this approach has quite limited success for stocks because of the lack of universal laws governing their performance.

This I realized was the key to success in technology, for there is a universal law governing all progress in technology, the S-Curve Law:

All progress in technology follows an S-Curve.

Tile S-Curve is the solid line in Fig. 1. While it does not provide any insight into the value of projects before they are initiated, it can be used to determine when projects will become successful.

In the early part of the curve, a lot of time and effort is expended for very little progress. At this stage there appears to be little chance of reaching the successful level indicated by the dashed line in the figure. However, in projects that are destined to be successful, the pace of progress gradually quickens, as indicated at point A on the curve. By point B, the rate of increase of progress is quite noticeable, and yet the rate will continue to accelerate further, until a yet higher rate of progress is achieved well before the level of success is reached at point C.

Following the initial success of the project, progress usually continues at a rapid rate, leading to an over-optimistic corporate straight-line projection, as shown in the figure. Then a gradual leveling-out occurs. This causes great trauma in the marketing and financial sectors of the company, which results in increased pressure on the technical groups, for it is their job to solve the problems and get the technical progress back onto the optimistic straight-line projections.

Clearly there is no reason to seek responsibility for a project in the early stages when progress is slow and success uncertain. It is best to be given responsibility after point B (when progress is picking up) but before C (when the project has already been proclaimed a success).

There is great temptation to remain with a project long after the point of success has been reached. This is technically referred to as "basking in the glory." It is, however, an ill-advised luxury. For once the rate of progress begins to level out at point D, it may be too late to avoid the recriminations associated with failing to meet the straight-line projections.

Even worse than staying on a project too long is the mistake of stepping in to carry the ball once point C has been reached. By this point, further progress is already assumed, but it is more likely to be attributed to the technology than to the project manager. Furthermore, incompetent management cannot be masked at this stage by the myriad of uncertainties and technical problems that can cover a manager's incompetence in earlier stages of a program.

This qualitative discussion of the S-Curve is sufficient for most purposes and provides a good initial basis for project selection by ambitious technocrats. Substantial refinements are, however, available through mathematical analyses. Of particular interest is the application of the S-Curve Law to progress in the development and manufacture of semiconductor components.

No field of modern technology has moved more rapidly and provided more opportunity and risk than this one. The first replacement of electronic vacuum tubes (once called radio tubes) by transistors occurred in the 1950s. Since then a two-fold improvement in the cost-performance of semiconductor devices has occurred every two to three years.

For many years, these improvements have been closely linked with size reduction, which in turn has been made possible by reducing the amount of dust and other contaminants during fabrication. This has been accomplished through the use of carefully designed "clean rooms." A detailed analysis of the yield of devices in manufacturing showed that it was related to the density of dust particles, according to the well known S-Curve. In this case percentage yield was plotted on the vertical (progress) axis and the reduction of dust density was plotted along the horizontal (time) axis as shown in Fig. 2.

Several years ago, an engineer for a major manufacturer of semiconductor devices became aware of this recently derived relationship at a most opportune time. He was serving on a five-man task force established by the president to find out why there was still no yield of a newly designed semiconductor device. The task force learned that recent steps taken to reduce dust in the "clean room" had improved yield from something less than 1 per cent to about 5 per cent. This, however, was still far below the 40 to 50 per cent yield required to achieve profitability in the program. The project leaders and the task force members were, therefore, not optimistic about the chances for success.

The mathematical analysis of yield versus dust particle density, however, revealed that the improvement in cleanliness required to get from a 1 to a 5 per cent yield was actually greater than that needed to get from a 5 to a 40 per cent yield. Since this recent analysis was known only to the engineer, only he had reason to be optimistic.

He wrote a damaging report on the incompetence of the program's manager and submitted it with a long list of recommended actions. Most of these actions, he failed to note, were already being implemented or were in the planning stage by the present manager.

When top management reviewed the various reports on the project, they were understandably distressed at the very low semiconductor yield. The negative comments about the present project leader suggested a solution on which they were eager to act.

Proof of the wisdom of this decision came in less than six months, when the engineer from the task force, now the new manager of the project, was able to report a 43 per cent yield, with further improvements expected. In making the recommended change in project leaders, management had truly done its work effectively.

Unfortunately, the engineer became so enamored by his own success that he continued to manage the project beyond the point where rapid progress could be expected. His reputation tarnished, and his opportunities in the company vanished. Eventually he was replaced by one of his younger and presumably more innovative subordinates.

Next: Laws Governing Values