Technology Trajectory Mapping Using Data Envelopment Analysis

Preparing to load PDF file. please wait...

0 of 0
100%
Technology Trajectory Mapping Using Data Envelopment Analysis

Transcript Of Technology Trajectory Mapping Using Data Envelopment Analysis

Portland State University
PDXScholar

Engineering and Technology Management Faculty Publications and Presentations

Engineering and Technology Management

11-2016
Technology Trajectory Mapping Using Data Envelopment Analysis: The Ex-ante use of Disruptive Innovation Theory on Flat Panel Technologies
Dong-Joon Lim Portland State University
Timothy R. Anderson Portland State University, [email protected]

Follow this and additional works at: https://pdxscholar.library.pdx.edu/etm_fac
Part of the Operations Research, Systems Engineering and Industrial Engineering Commons
Let us know how access to this document benefits you.
Citation Details Lim, D., & Anderson, T. R. (2016). Technology trajectory mapping using data envelopment analysis: the ex ante use of disruptive innovation theory on flat panel technologies. R&D Management, 46(5), 815-830. doi:10.1111/radm.12111
This Post-Print is brought to you for free and open access. It has been accepted for inclusion in Engineering and Technology Management Faculty Publications and Presentations by an authorized administrator of PDXScholar. Please contact us if we can make this document more accessible: [email protected]

Technology Trajectory Mapping using Data Envelopment Analysis : The Ex-ante use of Disruptive Innovation Theory on Flat Panel Technologies
Dong-Joon Lim*, Timothy R. Anderson Dept. of Engineering and Technology Management, Portland State University, Portland, USA
Abstract- In this paper, we propose a technology trajectory mapping approach using Data Envelopment Analysis (DEA) that scrutinizes technology progress patterns from multidimensional perspectives. Literature reviews on technology trajectory mappings have revealed that it is imperative to identify key performance measures that can represent different value propositions and then apply them to the investigation of technology systems in order to capture indications of the future disruption. The proposed approach provides a flexibility not only to take multiple characteristics of technology systems into account but also to deal with various tradeoffs among technology attributes by imposing weight restrictions in the DEA model. The application of this approach to the flat panel technologies is provided to give a strategic insight for the players involved.
1. Introduction
Technological forecasting methods can be classified as either exploratory or normative by whether they extend present trends (exploratory) or look backward from a desired future to determine the developments needed to achieve it (normative) (Porter et al. 2011). The correct assessment of future environment and of the corresponding goals, requirements, and human desires can be better made when exploratory and normative components are joined in an iterative feedback cycle (Jantsch 1967). Here, it is crucial to have an accurate understanding of the technological inertia we have today so that exploratory methods extend the progress while normative methods determine how much the speed of such progress need to be adjusted. However, as technology systems become sophisticated, the rate of change varies more significantly, being affected by the maturity levels of component technologies (Lim et al. 2014).
1

This structural complexity makes today’s forecasting even more challenging, which leads to the question: which set of attributes have the disruptive potential to be scaled up (or down) in the future?
Technology frontier analysis has been used in several ways to consider this multidimensional and combinatorial characteristics of technology systems (Gu and Kusiak 1993; Hazelrigg 1996; Martino 1993). The simplest form is the planar frontier model (or hyper-plane method) suggested by Alexander and Nelson (Alexander and Nelson 1973). Although this approach has an advantage of a simple implementation based on multiple regression analysis, a fitted functional form of the frontier based on a linearity assumption disallows to consider dynamic tradeoffs among technology attributes. As a non-linear frontier model, Dodson proposed an ellipsoid frontier formation (Dodson 1985). This model attempts to fit the technology frontier into a priori functional form from which tradeoffs among attributes can be explained. However, ellipsoid frontier model requires that the rate of one technical capability being relinquished for the advancement of the others rely on the predefined functional form rather than the nature of data at hand. Dodson’s choice of an ellipsoid shape is analytically sound for the representation of a strictly convex surface but may not always be representative. Moreover, this model doesn’t provide a time dependent measure to estimate the future state of the technology frontier. To tackle this issue, Danner suggested the iso-time frontier using MultiDimensional Growth Models (MDGM) (Danner 2006). In this approach, the frontier surface is formed by a composite relationship between time and technological characteristics. Therefore, the frontier can be navigated to project multiple characteristics into the future (Cole 2009). Possibly the greatest limitation to the utility of MDGM is the requirement that all dimensions of technical capability integrated must be statistically independent. This presupposes that the time required to advance each attribute towards corresponding upper limit can be linearly combined to
2

explain the technology systems’ growth rate. However, the higher the complexity of technology systems under evaluation is, the more individual growth rates are likely to be interrelated hence generated iso-time frontier without consideration on concurrent advancement would not provide an accurate picture of the feasible combinations of technical capabilities.
To overcome the disadvantages of the aforementioned methods, this study proposes an approach that can be used as a composite measure of technical capabilities as well as a tool for investigating rate of changes that enables to project the current technology frontier into the future.
2. Literature review on technology trajectory mapping
Mapping performance of technology over time can be helpful to identify potential disruptive technologies as well as to examine the maturity of incumbent technologies. Clayton Christensen and Michael Overdorf explained the theory of disruptive innovation by suggesting that “graph the trajectories of performance improvement demanded in the market versus the performance improvement supplied by the technology… Such charts are the best method I know for identifying disruptive technologies (C. M. Christensen and Overdorf 2000).”
Trajectory mapping has been employed in a wide range of applications. The most famous application of a trajectory map may be the hard disk drive case from Christensen’s original work (C. M. Christensen 1993). He used disk capacity as a performance axis and interpreted the dynamics of industry that smaller disks have replaced bigger ones improving their capacities over time. Schmidt later extended Christensen’s work by classifying the disk drive case as a lowend encroachment that eventually diffused upward to the high-end (Schmidt 2011). Martinelli conducted patent analysis in the telecommunication switching industry to find out seven generations of technological advances from the different paradigmatic trajectories (Martinelli 2012). Kassicieh and Rahal also adopted patent publication as a performance measure in search
3

of potential disruptive technologies in therapeutics (Kassicieh and Rahal 2007). Phaal et al. proposed a framework that has been tested by developing more than 25 diverse ‘emergence maps,’ analogous to trajectory map, of historical industrial evolution, building confidence that the framework might be applicable to current and future emergence (Phaal et al. 2011). Keller and Hüsig analyzed Google’s web-based office application to see if it can pose a disruptive threat to incumbent technologies, namely Microsoft’s desktop office application (Keller and Hüsig 2009). Barberá-Tomás and Consoli tried to identify potential disruptive innovation in medical industry, especially on artificial disc, by counting the number of granted patent over time (Barberá-Tomás and Consoli 2012). Husig et al. (2005) conducted one of rare ex ante analyses that mapped out trajectories of both the incumbent technology and a potential disruptive technology (Husig, Hipp, and Dowling 2005). They made a forecast based on trajectory map that Wireless Local Area Network (W-LAN) technologies would not be disruptive for incumbent mobile communications network operators in Germany. This is because the average growth rate of the bandwidth supplied by W-LAN had been overshooting the average growth rate of the bandwidth requirements of all customer groups.
There are a few studies that used composite performance measures to draw the technology trajectories. Adamson plotted R2 values from the multiple regression analysis on the trajectory map to investigate the fuel cell vehicle industry (Adamson 2005). The results showed that subcompact vehicle’s R2 values were increasing over time while compact vehicles’ were decreasing. The author interpreted that the technological advancement of subcompact vehicle was becoming similar to that of compact vehicle. This study has significant implications for identifying key drivers of technology progress using the trajectory map. Letchumanan and Kodama mapped out the correlation between Revealed Comparative Advantage (RCA), which is generally used to measure the export competitiveness of a product from a particular country in
4

terms of world market share, and R&D intensity to examine who was making the most disruptive advancement at a national level (Letchumanan and Kodama 2000). Even though Koh and Magee didn’t utilize any function to develop composite performance measures, their research has a significance as they took different trade-offs into consideration to draw a trajectory map (Koh and Magee 2006). Their results suggested that some new information transformation embodiment such as a quantum or optical computing might continue the trends given the fact that information transformation technologies have shown a steady progress.
Table 1 summarizes 40 studies from 1997 to 2012 that have used trajectory map to identify disruptive alternatives including technology, product, and service. The majority of the studies adopted a single performance measure and simply connected time series data points, indicated as data accumulation, to draw the trajectory map.
A trajectory map should take multiple perspectives into account not to miss potential disruptive indications. This involves predicting what performance the market will demand along various dimensions and what performance levels will be able to supply (Danneels 2004). It is often recognized that new technologies would not always be superior to the prior one as well as performance disruption, i.e. intersection between trajectories, could occur from the technology that had been crossed in the past (Sood and Tellis 2005). Many ex post case studies have shown that disruptions have happened from an entirely new type of performance measure that hadn’t been considered. This implies that current performance measure may be no longer capable of capturing advancement in a new direction. Therefore, it is crucial to examine not only which performance measures are playing a major role in current progress but also which alternate technologies show disruptive potential with respect to the emerging performance measures.
5

Table 1 Summary of literatures on the technology trajectory mapping

Author (year)
Walsh (2004) Keller & Hüsig (2009)
Martinelli (2012) Phaal et al. (2011) Padgett & Mulvey (2007) X. Huang & Sošić (2010)
Kaslow (2004) Kassicieh & Rahal (2007)
Christensen (1997) Schmidt (2011) Rao et al. (2006)
Bradley (2009)
Lucas & Goh (2009) Madjdi & Hüsig (2011)
Husig et al. (2005) Walsh et al. (2005) Figueiredo (2010) Caulkins et al. (2011) Adamson (2005) Belis-Bergouignan et al. (2004)
Ho (2011)
Werfel & Jaffe (2012)
No & Park (2010)
Letchumanan & Kodama (2000)
Spanos & Voudouris (2009)
Frenken & Leydesdorff (2000)

Application area
Microsystems Office application Telecommunication S&T based industry Brokerage market General industry
Vaccine Therapeutics Disk drive Disk drive P2P and VoIP Medical operation (MRgFUS1) Photography
W-LAN W-LAN Silicon industry Forestry industry General industry Fuel cell vehicle Organic compound
General industry (Taiwan)
Smoking cessation products
Nano-biotechnology General industry (High-tech)
Manufacturing SMEs (Greek)
Civil aircraft

Watanabe et al. (2009)

Printers

Hobo et al. (2006)
Watanabe et al. (2005)
S.-H. Chen et al. (2012) Epicoco (2012) Funk (2005) Raven (2006)
Castellacci (2008)
Kash & Rycoft (2000)
Arqué-Castells (2012)

Service oriented manufacturing industry
Electrical machinery (Japan)
Smart grid Semiconductor Mobile phone Renewable energy Manufacturing and service
industries Radiation therapy General industry
(Spain)

W.-J. Kim et al. (2005)

DRAM

C.-Y. Lee et al. (2008)

Home networking (Korea)

Koh & Magee (2006)

Information technology

Barberá-Tomás & Consoli (2012)
1: MR-guided Focused Ultrasound

Artificial disc

2: Advanced Manufacturing Technology

Performance measure
Critical dimension Number of operations
Patent citation Sales
Level of service integration Capacity & Price Efficacy Patent publication Capacity Part-worth Data transfer
Noninvasiveness
Price, convenience, etc. Active Hotspot ratio Data rates Number of firms
Novelty & complexity level Market connection
Utility coefficient values Environmental performance
Technology sources and innovation drivers
Patent
Patent Correlation between Exports
and R&D intensity
AMT2
Diffusion rate (Entropy statistics)
Sales and price
Sales, income, employees, and productivity
Marginal productivity
Average age Devices per chip Mobile subscribers Energy production(TJ/yr)
Labor productivity
Capability
Patent
DRAM shipment and Memory density
Units of new household/year Megabits
Patent

Plotting method
Growth curve Data accumulation Data accumulation Data accumulation Data accumulation Data accumulation Data accumulation
Patent mapping Data accumulation Data accumulation Data accumulation
Data accumulation
Data accumulation Data accumulation Data accumulation Data accumulation Data accumulation
Skiba curve Data accumulation Data accumulation
Data accumulation
Reduced form model
Data accumulation
Data accumulation
Data accumulation
Data accumulation
Technology price function
Data accumulation
Data accumulation
Data accumulation Data accumulation Data accumulation Data accumulation
Data accumulation
Growth curve
Poisson model
Data accumulation
Data accumulation Data accumulation
Data accumulation

6

3. Methodology
To supply insight into the approach we are proposing, this section introduces Technology Forecasting using Data Envelopment Analysis (TFDEA.) The DEA model, which underlies TFDEA, is unique in that it allows each Decision Making Unit (DMU) to freely choose its own weighting scheme, and as such, the efficiency measure will show it in the best possible light (Charnes, Cooper, and Rhodes 1978; Fried, Lovell, and Schmidt 2008). This flexible weighting characteristic has shown practical advantages in a wide range of applications especially when the assessment involves complex tradeoffs that are difficult to model as a universal set of weights (Lim, Anderson, and Kim 2012). When the application area calls for limits on relative weights, upper or lower bounds of weights can also be implemented by imposing weight restrictions (Dyson and Thanassoulis 1988; R G Thompson et al. 1986; Russell G Thompson et al. 1990; Wong and Beasley 1990).
Based on the strengths of DEA, TFDEA has been used in a number of forecasting applications since the first introduction in PICMET ’01 (Anderson, Hollingsworth, and Inman 2001; Cole 2009; Lim, Anderson, and Shott 2014; Tudorie 2012). Figure 1 shows the TFDEA rate of change (RoC) calculation process with AR-I (Assurance Region type 1) weight restrictions implementation in a multiplier model (R G Thompson et al. 1986). Specifically, the variable serves as the objective function and represents the weighted sum of inputs using the most favorable set of weights, , , for technology at time period . Since each reference set only includes technologies that had been released up to , indicates how superior (or efficient) the technology is at the time of release. The effective year, , is determined by calculation of (1) to specify a weighted average of the old technologies that technology k is being
7

compared against. Note that the benchmarking parameter, , , is obtained from the envelopment model and calculation of (1) can be simplified as (2) in the case of VRS.
∑∑ ∙ , , , ∀ 1

∙ ,, ∀

2

The RoC, may then be calculated taking all DMUs that were efficient at the time of

release,

1 , but were superseded by technology at time ,

1 . For a more

comprehensive treatment of TFDEA, the interested reader is referred to earlier studies (Inman

2004; Lim, Anderson, and Inman 2014).

Figure 1 TFDEA RoC calculation process with AR-I implementation 8

4. Trajectory mapping on flat panel industry
To illustrate the use of the methodology presented in this paper, we provide an example of trajectory mappings that is applied to the flat panel industry to examine technology progresses from various perspectives. 4.1. Dataset
Lim, Runde, and Anderson investigated the technology advancement of Liquid Crystal Display (LCD) to forecast future state of the arts (SOAs) specifications (Lim, Runde, and Anderson 2013). This study examined 389 LCD panels with five characteristics that were determined from a group of LCD technologists. As a follow up study, the dataset has been updated to include 442 LCD panels and 29 Organic Light Emitting Diode (OLED) panels that have been introduced from 1998 to 2012 (see Table 2 for the summary of data). Variables included for this study are as follows:
 Company / Name (text): manufacturer and name of panel  Backlight (text): illuminating source  Year (year):year of release  Screen Size (inches): diagonal length  Bezel Size (millimeters): length from the outside shell to the beginning of the active
area  Weight (kilograms)  Resolution (pixels): horizontal times vertical resolution  Contrast Ratio (ratio): the ratio of luminance of brightness 0 to 100% energized
pixel(s)
9
TimeTechnologiesTechnologyTrajectory MapIndustry