The personal and mobile computing industries

In order to better understand the phenomenon of over-serving gaps in the computing in-dustry this research looked for additional evidence that could shed some light on the problem. To measure the offer of performance in the personal and mobile computing in-dustries, data for 91 product lines was collected, including personal computers, smart-phones, personal digital assistants (PDAs), tablets, and operating systems from 1974 to 2017. A list from all product lines, and a detailed discussion behind all the considerations behind this database can be found in Chapter 3. Below only the aspects pertaining to product performance are discussed.

Instead of using a random sample, this dataset was built by exhaustively collecting infor-mation on as many product lines as it was possible to identify, this way the sample closely resembles the population and avoids selection bias. This dataset is comprehensive and this research is not aware of any important omissions (Reimer, 2005, 2012a, 2012b;

Dediu, 2012b), but work on expanding the dataset still continues.

This section focuses on the minimalistic definition of disruptive innovation that is most widely shared, and thus less controversial, in other words innovations that lower product performance, regardless of their market effects (Christensen, 2006; Sood and Tellis, 2011). This definition does not share all of the characteristics of Christensen’s model, but in exchange it allows for an operational definition that is easier to measure and agree upon.

During classification, this section found that in order to assess whether a product line worsened or sustained performance is not sufficient to compare two product lines, a grasp of the whole market and its trends are necessary to understand whether a product line lowers performance or not. Comparisons between different markets should only be made when two different product categories overlap, as is the case in the early stages of new market disruption.

Following Sood and Tellis (2011), this research first attempted to define disruptive tech-nologies strictly using a minimalistic definition based only on performance. However, this approach proved insufficient. Because of this, one more flexible conceptualization of dis-ruptive innovation as an innovation that shifts the basis of competition was also intro-duced to reflect the more modern understanding of the phenomenon present in disruption literature today. A shift in the basis of competition means that an innovation changes the parameters by which products are valued by customers, or new business models.

In total, this section analyzes three variables. For a detailed definition of each variable see Chapter 3.1, below only the operational definitions have been presented:

- New entrant: This variable identifies whether a company was a new entrant to a market, or instead a preexisting incumbent in the market. By definition entrant companies are in average younger than incumbent companies, however this gen-eralization can be extremely dangerous in analysis. Time of foundation is not im-portant, it is the time of entry to a market what defines an entrant company.

- Worse performance: This variable identifies whether a product line worsened per-formance at least in the short term, or instead sustained perper-formance. It is not enough that a product line lowered performance in any dimension, it must have been one of the main dimensions historically valued by customers. If a product line sustained performance in the long term, but lowered in the short term it still counts as a case of lowering performance. However, the lowering of performance must have been considerable, even if it was only temporal.

- Shift in basis of competition: This variable identifies whether a product line shift-ed the basis of competition in the market, or instead sustainshift-ed the existing basis of competition. A shift in the basis of competition means that the way competition takes place in the market has changed, this can take the form of changes to prod-ucts themselves, or new business models.

This research found that from 91 product lines analyzed, 63 product lines (69%) of them lowered performance at least temporally in comparison to the main technology trends at the time, as can be seen in Figure 16. This result validates Christensen’s identification of the counterintuitive phenomenon of innovations that lower performance instead of sus-taining it. However, it was found that this phenomenon is more prevalent than the theory suggests.

!

Figure 16. Observations of variables representing Christensen’s concepts.

On the other hand, 58 of all product lines (64%) were found to have shifted the basis of competition. The similar values might lead to the conjecture that they must overlap al-most perfectly, however only 49 product lines (54%) were found to both lower mance and shift the basis of competition. From the 63 product lines that lowered perfor-mance, 78% also shifted the basis of competition. There is a big overlap, but also a 22%

difference for two variables that are supposed to represent the same concept — that is to say, Christensen’s disruptive innovation —.

Excluded from the overlap, 14 product lines lowered performance, but sustained the basis of competition, that is 64% more in comparison to the 9 product lines that shifted the ba-sis of competition, but sustained performance. This difference can be explained by using the same reasoning put forward previously when studying the disk drive industry: innova-tions that lower performance are different from innovainnova-tions that shift the basis of compe-tition, although they are often related. If we accept that companies are able to adapt their

New entrant Incumbent Worse performance Sustains performance Shift in basis of competition Sustains basis of competition

Observations

0 10 20 30 40 50 60 70

33

58 28

63 48

43

the concept of trickling of performance offer. Companies are often adjusting their perfor-mance offer, however they do not always do so in order to shift the basis of competition

— which requires a more innovative effort —, in many cases it is just a response to the perceived customer demand of performance.

Once lowering performance becomes an established trend in an industry, subsequent technologies that lower performance do not shift the basis of competition significantly.

The computing industry has had many examples of innovations that lower performance as can be seen in Table 2.

From all cases studied, 43 product lines (47%) were introduced by entrant companies, and 34 (37%) were introduced by entrant companies that also lowered performance. From a total of 63 product lines that lowered performance, entrants introduced only half (54%) of them. This finding contradicts Christensen’s theory, and is in line with the findings of other studies according to which entrants and incumbents alike introduce innovations that lower performance (Sood and Tellis, 2011; King and Baatartogtokh, 2015).

Table 2. Innovations that lowered performance in computing industry.

Innovations that lowered performance Type IBM PC compatibles that reduced price and quality Hardware Physically smaller disk drives that reduced storage capacity Hardware Budget microprocessors like Intel’s Celeron Hardware

AMD’s cheaper x86 microprocessors Hardware

ARM’s power constrained microprocessors Hardware

Non Error-Correcting Code memory Hardware

Integrated graphics that reduced graphic performance Hardware Portable computers that were less powerful than desktops Hardware Graphical user interfaces that taxed computing resources Software Higher level programming languages that were slower Software More secure operating systems that taxed performance Software The shift to mobile computing that constrained power Hardware The change from wired to wireless networking Hardware

Similarly, from all cases studied, 30 (33%) were product lines introduced by entrant com-panies that also shifted the basis of competition. From a total of 58 product lines that shifted the basis of competition, entrants introduced half (52%) of them. This finding also contradicts Christensen’s theory, and is more relevant than the previous finding pertaining to lowering performance, given that the basis of competition is a more modern concept than product performance. Despite demonstrating that innovations that lower perfor-mance and innovations that shift the basis of competition are different, this research found that in both cases incumbents are as likely as entrants in pursuing either of these kinds of innovations. In other words, incumbents are as likely as entrants to introduce innovations that have been often understood to be disruptive, regardless of whether we use a tradi-tional or modern definition.

An in-depth analysis of the causes behind this phenomenon revealed that Moore’s law often generated gaps in which customer needs for computing power were temporally over-served, and this created opportunities for innovations that lowered performance.

Since the 1960s Moore’s law has been the guiding star of the industry, and entrants and incumbents alike have regularly introduced products that lowered the performance of components such as microprocessor and memory. However, these adjustments were only temporal, given that customers still demanded more performance in the long term.

According to Gruber, “for decades, computers were starved for raw performance. CPUs were slow, RAM was scarce, disks were slow (and unreliable), graphics were slow. Print-ing was slow. NetworkPrint-ing was slow. EverythPrint-ing was slow. And the more money you spent, the more you could alleviate these problems with faster components, and more ports and peripherals” (2016). However, users were also price sensitive, and at times customer needs ‘per dollar’ grew slower than technological product performance gains, forcing companies to adjust their offering. Instead of being black and white, the demand of per-formance is better understood in relation to its intensity.

This trend has become so entrenched that companies like Intel have learned to diversify

sonal computers, while also pursuing the more profitable high-end with a different prod-uct offer. This, in turn, has enabled computer manufacturers to conveniently have a per-manent low-end offering, while also pursuing the high-end. Even companies like Apple, well known for its high profit margins, keep a mid- to low-end offer available. This con-tradicts Christensen’s hypothesis of ‘downward immobility’ in the value network, accord-ing to which “well-managed companies are generally upwardly mobile and downwardly immobile” (1997). In practice, companies in the computing industry have learned to be mobile and flexible in their value networks.

This research found that the lowering of performance was almost always temporal, and once companies found innovative ways to offer new value to customers the performance that customers could absorb also increased. This phenomenon was more remarkable than the temporal instances of worsening in performance, and customer needs have kept grow-ing exponentially to this day. Instead of over-servgrow-ing the market indefinitely, companies used the improvements in raw computing performance to unlock new uses. As can be seen in Table 3, computers became more versatile and offered new customer value thanks to the emergence of new uses like multimedia, web browsing, and nowadays mobility.

Each of these new uses were tied to key innovations that enabled them.

Table 3. New uses for computing devices 1974-2016.

Year New uses for computing devices 1974 Word processors

1979 Spreadsheets

1985 Multimedia (animation, graphics, and sound technologies) 1990 Photos and graphics editing

1994 3d animation

1995 Internet access, mail, and web navigation 1999 Sharing files and music

2003 Video conference calls 2005 Streaming video 2006 Social media

2016 Virtual and augmented realities

These findings contradict Christensen’s assertion that “the needs of many computer users have increased more slowly than the rate of improvement provided by computer design-ers” (1997), and instead support the results of other researchers who contend that inter-pretation. According to the interviews done by King and Baatartogtokh, customers alter they behavior in response to better performance:

“Incumbent companies, one expert noted, often use computing power to serve the needs of new, less-savvy customers — not to overshoot customer needs. Apple Inc., for example, launched its original Macintosh in 1984 with a powerful central processing unit to enable a user-friendly graphical interface…

Customer needs often respond to better performance, altering behavior and, in turn, one’s perception of needs. Today, for example, many of us feel we “need” to carry around what are in effect supercomputers that take pictures and allow us to chat with our friends. In some cases, one expert pointed out to us, user needs seem to be insatiable.”

—King and Baatartogtokh, 2015.

In order to account for these results, this research borrowed from the notion of elasticity in economics and applied it to the study of performance. This research has coined the term ‘elasticity of customer needs’ to describe this phenomenon: the responsiveness of the performance demanded by customers to a change in the performance offered by man-ufacturers, especially when new value is created. An adaptation to Christensen’s model using this concept can be seen in Figure 17.

!

Figure 17. Elasticity of customer needs.

If a small change in product performance offer is accompanied by a large change in cus-tomer demand, then cuscus-tomer needs are elastic (or responsive to product performance of-fer). Conversely, customer needs are inelastic if a large change in product performance offer is accompanied by a small amount of change in customer demand. The elasticity of customer needs can be represented by the following formula.

!

For the application of this formula is important to keep in mind that if the purpose is to compare the elasticity of very similar products, no unit conversion might be needed — this might be useful for operations management and just-in-time manufacturing —, how-ever, if dissimilar products are to be compared, then changes in performance should be expressed in the form of percentages, rather than in raw performance units. The elasticity of customer needs can take values from 0 to infinity. If elasticity is less than 1 it repre-sents inelastic customer needs, and if elasticity is more than 1 it reprerepre-sents elastic cus-tomer needs. Although unlikely to occur in the real world, there are three special cases: if elasticity is 0 is called perfect inelasticity, if elasticity is 1 is called unitary elasticity, and if elasticity tends to infinity is called perfect elasticity.

Elasticit y of cu stomer need s = %Δ Cu stomer d emand of prod u ct per for mance

%Δ Of fer of prod u ct per for mance

Notice that the formula for elasticity of customer needs compares the offer of mance with the demand of performance. Both variables are expressed in terms of perfor-mance. It would also be possible to compare the offer of performance with unit sales, in-stead of the demand of performance. In that case, inin-stead of measuring the demand of performance directly, unit sales would act as a proxy. It can also be argued that for man-agers interested in sales, customer needs are actually a proxy. This research does not rec-ommend that approach, however it acknowledges that performance demand can be rela-tively easy to measure in the computing industry, but not so much in other industries, in which case measuring unit sales might be a better alternative.

!

Figure 18. Elasticity of customer needs over time.

As can be seen in Figure 18, customer needs tend to be elastic when technologies are im-mature, in which case customers will absorb any performance gains offered to them by almost the same amount. As technologies mature, customer needs tend to become inelas-tic and more price sensitive. At this stage, it becomes possible to over-serve the market, however, contrary to traditional interpretations of ‘technology maturity’ that assume this process to be definitive (see Figure 52), it is possible for innovations to revitalize the market. Innovations can unlock new uses and create additional customer value, at which point customer needs become elastic again and respond positively to gains in perfor-mance.

!

Figure 19. Performance (offer adjusted) over time.

In the short term, as seen in Figure 18, the elasticity of customer needs changes cyclically, and can theoretically return to the same value after a cycle. However, in the long term, changes in the offer of performance in the market are cumulative, as seen in Figure 19. As a result, the cyclical nature of customer needs alters the speed at which performance gains are introduced over time, but it does not alter the overall trend of performance getting bet-ter and betbet-ter over time.