Bull strategy and processes
an essay by Jean Bellec to summarize the way Bull and its parent companies worked.
During the 1980-1985 period, as the Bull and Honeywell companies were entirely separate entities, we lack the knowledge about Honeywell strategy and the formal decision process, if any, they use to make the decision to quit the computer business.
It is not in my intention here to scrutinize the
decisions taken at corporate level, dealing with financial institutions, governments
authorities, shareholders and to look at the rationale of mergers and acquisitions.
I will attempt, instead, to analyze the processes in which have shrouded the corporate culture of Bull in the last 40 years. That culture was not so different from that of other companies in the industry, while it has been obviously influenced by typical French features. Those so-called continent typical features were the needs to argue, the cartesianism, the love of elegant solutions, delaying tactics might be some of the trends that may have looked strange to an American businessman.
Old Bull strategy
One of the characteristics of Bull's, and others', organization is a powerful but transparent sales network. The higher coverage of the set of customers was the goal, at least until the downsizing of the 90s. By transparency, I do not mean necessarily that the sales organization did not play some specific tricks such as lowering the forecasts to decrease the risk of higher quota, or carrying on orders over year-end. In fact, the sales organization had had very little initiative in detecting changes in the industry trends and reflected essentially the present attitudes of customers vis-à-vis the offer from Bull and its competition.
A first global appraisal of Bull's strategy would be that it was and always has been a "follower" attitude behind the industry leader, IBM. There was also a continuous trend to do everything, and sometimes the "impossible" to keep customers faithful and captive, usually by specific technical tricks (such as different codes), sometimes by financial "concessions" very seldom by predating contracts.
the GE model
When General Electric took over, it added to this basic strategy a "scientific" analysis of the business, applying to the computer industry the models that have been successful in other branches of the manufacturing industry.
An important phase was the identification of the market
segmentation. It was important to know if independent growth of a product could
be generated without impacting others. Somewhat unfortunately, the market segmentation was
based more on the customer size than on their use of computers. This type of segmentation
has been heavily promoted by early consulting firms. The business was divided into Large
computers reserved to Fortune 500 and government agencies, Medium sized computers, and
Small computers for small private business. The success of IBM S/360 introduction seemed
to indicate that the initial 1950s segmentation between scientific and business computers
would eventually fade out. The introduction of supercomputers by Control Data and others
was seen, probably rightly, as fulfilling a niche market. So, General Electric and BGE,
and Honeywell after them, were ready to allocate one product line by market segment and
were not too worried in designing incompatible products to fill those segments.
When the "technical" market exploded in the 1970s, a new segment was declared identified, and a new minicomputer product line was created. When, BGE discovered that very small enterprises will become computerized, again a new Very Small Computers product line was added.
Inside a market segment, it was expected that several generations of products would be induced by the technology progresses. It was also expected that new applications allowed by the technology would require a growth of the customers needs offsetting the reduction of price allowed by the technology. So, the segmentation by customer size would match a segmentation by price. Overall, it was considered that the computer expenditures were to be proportional to amount of sales of the customer.
There was some consideration to segment the market by industry, but there was just a few products that were industry specific. The needs for a transaction system for banking, are not so different from the reservation system in transportation. Differences between industry existed in application software, a business where GE, Honeywell and Bull were not.
This segmentation model worked while the Grosch's law
of growing returns on CPU power was valid. When large customers started to buy a
collection of small or medium size computers for specific usages, the model validity
started to fade.
Another factor was that it discovered that the customer business stability started not to exceed the life duration of a computer system and a fortiori of a computer application. A lot of companies mergers have encountered problems due to the incompatibility of their computer systems. Such mergers usually lead to the consolidation of their EDP units and to a serious not anticiped migration problem.
Nevertheless, the segmentation model established in the mid-1960s was still accepted as the base of the strategy for twenty years. And while it was recognized as obsolete, it was not easy to replace.
To fulfill those market segments, GE defined the concept of product life cycle. In fact product life cycles came from the concept of project life cycle, that was developed by General Electric's Apollo support department in the early 1960s for NASA. It dealt with the decision process that allow a project to progress smoothly. It is based on an early identification of risks and the preplanning of contingency plans to handles glitches. It insist much on the benefits of design reviews for that purpose.
The life cycle model was extended to handle commercial products by the introduction of a "market requirements" phase and measurements ratios (essentially the time to recover investments and the return on investment). The product life was divided into phases:
Theoretically, Product Planning was the only organization in charge of phase 0, Engineering came in phase 1, Marketing in phase 3, Manufacturing in phase 4. In fact, there was some overlap because Engineering was involved in the planning evaluation, Manufacturing was involved in prototypes development. Also, the "phasing down" phase 5 was rarely addressed before the sales curve fell down significantly.
Each phase was ended by an IPR
(independent product review) and by a management formal decision to continue the product,
to reiterate a phase or to abandon the product.
IPR consisted in a several days meeting, with presentations of the project team -informal gathering of all responsible for the product - to a review board - essentially composed of individuals coming from non related parts of the company, sometimes external -, followed by an intensive Q&A session, ending by writing a review report identifying the project risks. Those risks were rated by their impact on the business consequences (unacceptable meant that the project has to be killed if risk not corrected, high meant that the business plan would be significantly impacted...) and by a presentation to the project team and to the management.
One of the characteristics of the computer projects , specially when they involve a combination of hardware and software and when they involve new marketing activities, is the long lead time they involved, typically between 2 to 3 years, in the best case. Those durations could not be managed through a yearly budgeting process. So, the processes involved in the product life cycle were both necessary and useful.
Evolution of the GE model
Inside CII, the strategy and the methodology were sensibly different. CII concluded "Conventions de Plan" that specify multi-yearly objectives and contractual milestones with the government. The milestones were materially identifiable by any civil servant such as "power on on prototypes", but they do not make always sense from abusiness point of view.
Honeywell in 1970 was impressed by General Electric strategic tools and adopted GE inherited processes.
After the CII-HB merger, the GE originated methodology was maintained. It was somewhat altered by downplaying the concept of IPRs, and introducing the formality of Decision Meetings where all bureaucracies of the company were represented. Another subtle change in the process was to skip the concept of formal marketing requirements, that would have committed the sales network to their market appraisal, and to rely of centralized approval of functional specifications, the conformity with them becoming the benchmark of the product.
While the organization of the company has been built on the basis of internally developed products, the phase 0 of the development cycle included a "make or buy" analysis. The percentage of "buy" decisions went growing in the 1980s and the 1990s, but it became more and more difficult to stick with the process model. In particular, the reliance on a product line of the supplier was made implicitly, leading with decades length dependence from a data base supplier or a microprocessor company .
A few bias altered the accuracy of the decision mechanism. Inflation and the fluctuations of currencies exchange rates influenced the financial results: in the 1970s and the early 1980s, inflation reached a two-digit percentage per year and not taking it in account meant weak previsions. While Bull was a subsidiary of an American company and plans computations were in dollars, exchange rates had only an impact on manpower costs, but when Bull started to have its base accounting in French francs, all products elements were influenced by the dollar/franc - and sometimes Yen/franc - exchange rates. A more subtle bias factor has been the internal transfer prices between operations. Transfer prices have always been a secretive elements for fiscal reasons. In general, after 1970, those rules were basically symmetrical between all production operations (taking in account manufacturing costs and market prices, and assuming that each operation was taking its right share of R&D). Consequently, Bull appeared more profitable in selling American designed products, than in selling DPS-7000 or Unix boxes designed in France. When having to buy products from external manufacturers, buisness planners were sometimes surprised that gross margins were much lower than buying inside the group. A related factor was that, in accounting rules, R&D were not included in the cost of a product. Such a practice what disturbs significantly the economics, when R&D account to twice the manufacturing cost for VLSI hardware and even more for the software.
The product life cycle and the product business models have been also biased by the growing amount of derivative products. A derivative product was always less costly in terms of investments, was faster to produce, but, almost always, took its source of sales from the forecast sales of the produce it derived from. Consequently, the ROI that made the original product approved was negatively impacted, but as its model was not consolidated, no account was taken of the impact of the derivative.
Two corrections mechanisms were added: the first were product line business reviews made in front of the company top management and the second was the control of engineering expenses not by product ROI, but by yearly budgeting.
The product line business reviews were giving more light to corporate management, but the decision process that followed those review was, in general, hidden to the working level and decisions assumed to be taken by corporate officers were discovered non taken months later and ended in people beheading.
Again yearly budgeting lead to a long process where derivative products were hiden as "maintenance" , where management commitments were obtained for so-called strategic, not profitable developments paid for from subsidizing entities (government, EU) or from joint-work contracts with external companies.
Another process was entertained in the 1980s
and the early 1990s, it was the concept of a strategic plan. The
strategic plan, such as introduced by Francis Lorentz could have make the company
readdressing the market segmentation. But, at that time, the Bull management was convinced
of the eventual replacement of proprietary lines by open systems and a first segmentation
on that (somewhat artificial) basis. Then, within the proprietary business, that was then
90% of revenues, a sub-segmentation stood out : it was created by the compatibilities
within a product line and the incompatibilities with others. So, the strategic plan was
divided by product lines, although Bull's upper management was quite reluctant to endorse
the past of the company and wished to break out existing structure.
The product line strategic plan was the benchmark for the business reviews. It had a characteristic of being so sensitive that even its authors did not have the authority to read it.
Another plan, usually established for the whole company, was a multi-yearly plan (Horizon 85, Horizon 90) that was targeted more to the shareholder (i.e. essentially the French government) to justify stock increase or research subsidies. They explain the "reasons" for short term negative results and for shimmering futures. Those plans emphasized the downsizing of proprietary product line and their fast replacement by glamorous open systems. They explain why the dependency from foreign suppliers was temporary and will end soon. They often included a "wish list" established by engineering people attempting to introduce a commitment for their pet projects, specially when they know that their plan could not stand in front of a serious business review.
A lot of overtime and energy was spent by managers and bureaucrats to establish and defend those plans, and the time was scarce to address the actual customers needs and, above all, the market changes that surged with the PC revolution in that same period.
An attempt to determine a strategy adapted to the existing resources of the company, called analysis of core competence was done in the early 1990s. The idea was to identify the area where original products, recognized in advance from the competition, have been recently built, and focus the future on that direction. Such an analysis had the merit of trying to sort in the company, specially the Engineering, what can constitute an asset. The approach discovered that a company like Bull was marginally competitive in a lot of techniques, but the core competences identified such as the capability to designed SMP systems could not be mapped on specific business trend.
A new business model
In 1994, the new management initiated a big cultural change in the company. A large number of employees that had lived through the evolution and were sometimes skeptical about changes, left the company. Long development product cycles were now obsolete and a fast reaction time was accepted as the business rule. A decentralized management was instituted to cut into the bureaucracies and a monthly business report materialized by a "dashboard" was asked to managers. Bull global strategy has changed out from the development of products into a distribution and services business, leaving to others the design and the production of the bulk of the products.
Although the strategy and the related processes of Bull can, by no means, be considered as a clear success, they often have been also used in other information technology companies, and not the lowest, with a similar mix of success and failures. Much have been learned from General Electric and it is interesting to note that the same GE had evangelized IBM and some Japanese companies on the same subject.
Back to Product Strategy
Return to Corporate History
Revision : 05 juillet 2001.