[Home] [Papers] [Courses] [Contact]

Jamal Munshi, Sonoma State Univesity, 1991
All rights reserved


An empirical study shows that merchandise retailers have undergone rapid automation. The rate of adoption of each technology can be measured by the diffusion coefficient. Electronic Data Interchange (EDI), Point of sale systems (POS), bar coding using the universal product code (UPC), personal computers and local area networks (PC-LAN), and magnetic strip readers are being adopted at differential rates and the differences in these rates can be interpreted in terms of MIS impact and implementation issues. Industry observers cite Quick Response Purchasing (QR) and reduction of inventory as a major reason for adoption of POS and EDI. However, no clear effect on inventory turnover or overall asset utilization is apparent. One possible interpretation of the data is that investment in information technology only serves to increase centralization and managerial concentration. Although such a behavior can have implications in terms of agency theory, indications are that concentration of purchasing power lowers cost of sales and increases the gross margin. Thus innovation adoption affects the return on investment (ROI), not through asset utilization as may have been assumed but through the profit margin component of ROI.

MIS Effectiveness

Although MIS research has contributed significantly to the understanding of organizational issues and behavioral roles of users and developers in information systems it has been less successful in developing objective measures of effectiveness of information systems. More than fifteen years ago Lucas (1975) noted: Computer information systems are being developed by many organizations in an attempt to improve organizational effectiveness and efficiency. Unfortunately, there is little research evaluating the impact of information systems .... The state of affairs regarding effectiveness research remained unchanged a decade later as evidenced by this lament by Crowston and Treacy (1986): Implicit in most of what we do in MIS is the belief that information technology has an impact on the bottom line of business. Surprisingly, we rarely know if this is true. Finally, the current condition is stated most eloquently by Miller (1989): Measuring the effectiveness of computer based information systems remains unresolved and the topic regularly appears in the top ten ...of issues requiring attention from the IS community.

It is to this issue that the current study is addressed. In order to overcome what we feel have been the obstacles to the identification of objective measures of MIS effectiveness, we present what we consider to be a new and practical approach to measures of system impact and effect. Further, we apply these methods to an empirical study of the retail industry that reveals adoption patterns in the usage of information technology. MIS research on rational effectiveness can be improved by adopting a strategy to target specific industries and functional areas rather than addressing effectiveness hypotheses of a general nature.

Rational Effectiveness Measures

The terms `rational effectiveness' is used in this paper to describe measures of effectiveness that are independent of opinions and behavioral measures of users and managers. The term is derived from organizational paradigms described by Burrell and Morgan (1979) and as applied to information systems by Hirschheim (1989). Hirschheim describes four different organizational paradigms within which an information system may be evaluated. These are Functionalism, Social Relativism, Radical Structuralism, and Neo-Humanism. These classifications (drawn from Burrell and Morgan) are mapped into a two dimensional space having the property of subjectivism-versus-objectivism along one dimension and that of order-versus-conflict along the other . The application of these social paradigms to MIS effectiveness has been proposed by Munshi (1991).

Functionalism is the dominant paradigm in the quadrant governed by objectivism and order with no conflict. The theory of rational choice in business decision making as promoted by Von Neumann and Morgenstern (1944) subsume such an organizational paradigm. Effectiveness in such a paradigm can be referred to as `rational effectiveness' and can be defined in terms of the success of the business enterprise or the `bottom line' of enterprise level performance as suggested by Crowston and Treacy. However, since there are other and perhaps more powerful intervening variables at work in determining the financial success of the organization, the contribution made by MIS may be difficult to detect if loosely defined.

Although `effectiveness' has been the subject of many papers on MIS, most have dealt with sociological and psychological issues. Many empirical studies have addressed the effectiveness issue in the past using surrogate measures such as user satisfaction (Bailey and Pearson 1983, Doll and Torkzadeh 1988, Ives, Olson, and Baroudi 1983), system acceptance, system success (Cheney, Mann, and Amoroso 1986), system usage (Baroudi and Orlikowski 1988), system utilization (Trice and Treacy 1986) system efficiency, and system quality (Srinavasan 1985). A comprehensive tabular summary of these studies is provided by Miller (1988).

These studies have used for the most part, a sociological methodology wherein the MIS staff, end users, and operational and strategic managers of the firm are polled using questionnaire instruments. These instruments are designed to gauge the subjects' perceptual views, beliefs, and attitudes on psychometric scales. Summated scores of these beliefs and attitudes are then used as measures of MIS effectiveness.

In these measures the dominant paradigms are those of subjectivism rather than objectivism since they rely on the subjective experience of the users (eg How do you feel about your new order entry system?) rather than objective measurements of business functions (eg the average number of orders mishandled per month). Even if these instruments are properly designed and validated the `soft' data they produce can only be interpreted within subjective paradigms such as Social Relativism or Neo-Humanism (Hirschheim 1989). These are, admittedly important organizational issues since it is generally agreed that one of the goals of organizational MIS is to enhance the experience of the user community.

However, these measures tell only part of the story. A more complete measure of he impact and usefulness of an MIS in a business enterprise are the `hard data' measures of business performance within the paradigm of rational objectivism (Hirschheim 1989). The hard data include accounting and financial numbers, business growth, cost reductions, and achievement of stated rational system objectives. Such evidence, together with the `soft' data on subjective experience can provide a more accurate measure of MIS impact (Crowston and Treacy 1986).

For empirical research the problem of low statistical power may be addressed by tightening the functional definitions whose success are to be directly measured. For example, it may be difficult to measure the direct effect of a new accounts receivable system on the wealth of the shareholders but the measurement of the effect on mean collection time and percentage of the Receivables actually collected within a given time frame could be attempted. Such a measure can then be used to assess the ultimate effect on firm valuation using accounting methods.

To further improve our ability to observe the impact of MIS, a specific industry segment is chosen for study. The rationale for this is three-fold. First, within a given line of business both interfirm comparisons and comparisons of firms to industry averages are possible. Second, as many information system functions are highly specialized along lines of business, inter-industry comparison of these functions would not yield clearly interpretable results. And finally, firms within an industry group are likely to be in direct competition with similar cost structures and comparable investment and financing decisions. Thus firms with different MIS strategies can be compared on the basis of commonly used measures of firm performance in terms of financial and accounting variables such as growth rates, profit margin, and inventory turn-over ratio. Bakos (1987) has also suggested industry level measures of effectiveness.

It is possible that the limited success of previous studies of rational effectiveness is attributable to a definition of MIS that is is too broadly applied to yield statistically detectable measures. In early batch processing DP systems this orientation was entirely justified. For example, there was little that distinguished one payroll system from another regardless of whether paychecks were being printed for aerospace workers or bank tellers. However, with increasing technical sophistication of information systems the traditional batch processes can lay claim only to a small fraction of the MIS budgets of today (Allen 1986). The expanded MIS functions such as MRP systems of manufacturers, ATM systems of bankers, and QR systems of retailers are highly industry specific. As a result, the MIS of a bank and that of a manufacturer can be compared only in the most superficial terms and with low statistical power. Ideally, to maximize the power of detecting the effect of MIS on the operation of the business enterprise, we would like to find a very competitive industry segment with large MIS budgets and one in which information technology is purported to play a significant role. The retail merchandising industry fulfills these requirements.

Introduction to POS and EDI

In 1989, America's merchandise retailers had sales of over $200 billion and MIS budgets that exceeded $2 billion. The use of MIS as competitive strategy has been recognized by various researchers (for example, Robertson 1986). These aspects of MIS are evident in the retail industry which has been an innovator and an aggressive adopter of information technology (NRMA 1990).

In the past decade there has been some fundamental changes in retailing principally due to the use of POS (Point of Sale) systems, EDI (Electronic Data Interchange), and related innovations in information technology. Industry observers feel that technology has changed the competitive hierarchy among the leading retailers with resultant shifts in market shares (Babcock 1986, Anonymous 1988). Table 1 shows the trend in total merchandise sales and the changes in market share among the seven largest (publicly traded) retailers.

In the 1970s, with the availability of inexpensive IC chips such as the Motorola 6809, manufacturers of cash registers began to computerize their devices. By 1978, computerization had changed the cash register beyond recognition and it became renamed as a `point-of-sale' or POS system. Since the cash register was the point at which all transactions took place, it was possible, through automation, to capture transaction data directly and later, through OCR (optical character readers) and UPC (universal product bar code) to automate the data entry itself. Retailers were receptive to the innovation because it gave them greater control of the operation at a critical point. Additionally, with what amounted to real time data capture, they were able to track sales and inventory closely (Allen 1982).

The addition of UPC bar code readers automated the entire checkout process and ushered in the next generation of POS systems. Bar codes are a series of dark and white lines that are imprinted on the product (or tag) as a unique binary identifier. The lines are read by passing them over a laser. Once the product code enters the system, the POS can look up the price from a computer database and report it to the checkout register. Beyond that, the POS can automatically update inventory, maintain a transaction log that can be used to analyze sales trends, and automatically generate the daily business report and cash balance.

At a high level of integration, the POS terminals are connected into the main MIS computing systems for direct transfer of transaction data from the POS (pull) and price data to the POS (push). The transfers can be made in a batch mode once a day or even on-line through dedicated communication channels. The degree of integration actually achieved in the industry varies over a range but only 6% of the respondents in a survey considered their system to be not integrated. This is shown in Table II (adapted from Habern 1989).

The retailers wanted these systems because of the enhanced accuracy, reliability, control and flexibility they offered both to the in-store manager and to the corporate or regional office. Wholesale adoption was deterred by the high cost of these systems when they were first introduced. However, when the price of semiconductor devices began their rapid descent in the early `80s, the average cost of POS systems fell to almost a third of their original price tag of $10,000 . The falling price and increasing sophistication of the POS systems combined with the competitiveness of retailing to accelerate adoption rates. As a result, by 1990, the merchandise retailers had converted almost fully to POS.

These changing views of the industry were captured by national survey of retailers in 1981 (Allen 1981) and 1989 (Habern 1989). The surveys show that initial concerns about cost and complexity gave way to increasing optimism about the benefits of POS in terms of sales analysis and inventory control. Some of the survey findings are reproduced in Table III. Most retailers identified increased inventory turnover, better profit margins, and enhanced operating control as the primary benefits of POS.

However, as late as 1989, even when POS systems were considered mature, new barriers to diffusion of POS bar code scanning were identified in a survey by Deloitte Touche (NRMA 1990). These had to do with the logistics of maintaining industry wide bar code standards and availability of vendor markings. Retailers named `Investment costs of POS scanning equipment' and `Lack of vendor marked goods' as the major barriers to adoption.

POS Systems made it theoretically possible to track inventory and sales closely enough for retailer MIS managers to design inventory control systems, stock replenishment systems, sales reporting and analysis systems, and merchandise planning systems that would would virtually eliminate the uncertainty against which large inventories are held as insurance. The development of POS and its later enhancements have therefore facilitated the implementation of Quick Response (QR) purchasing.

In QR systems, there is virtually no inventory. The retailer is able to maintain a high in-stock rate by having accurate and current information about inventory and sales patterns. Using these data, he is able to order just those items that are being depleted and exactly at the point of exhaustion. Information was now available for replenishing inventory on a daily basis but the ordering process itself became a bottleneck to such an implementation.

Full implementation of QR therefore had to wait for EDI (Electronic Data Interchange), an innovation that could automate the re-ordering system (Hansen and Hill 1990). EDI replaces paper transaction documents such as purchase orders with direct electronic transfer of ordering information between retailer and vendor. Although originally used by McDonnell Douglas in 1974, the diffusion of this technology in the retail industry has been retarded by a lack of standards. Initial attempts by the large retailers to force their proprietary document interchange protocols on vendors have been unsuccessful. However, it is thought that the the introduction of the ANSI VICS X.12 standard in 1986 has cleared the way for large scale adoption of EDI by the retail industry (Munshi 1990).

A 1989 survey (NRMA 1990) shows that over 60% of retailers using EDI have switched over to VICS. In the same survey most retailers named `Reduce inventory levels' , Shortened re-order cycle', and `QR' as the major benefits of EDI (See Table IV). In the same survey, 56% to 65% of the respondents stated `Competitive Advantage' as the primary motivating factor for considering EDI. Some 45% of the respondents are using EDI at least on a pilot basis and are strongly committed to full conversion to EDI.

A third technology that has been adopted by retailers to control inventory is the hand held terminal. The hand-held terminal is used to facilitate the taking of physical inventory and for recording receipt of shipments. In 1988, over 68% of the mass merchandise retailers were using hand-held terminals (Habern 1989). It is expected that the three technologies together (POS, EDI, and Hand-held terminals) will assist the retailer in implementing QR. A merchandise information system with these features will help to control costs, improve asset utilization, increase margins, and thereby to increase the return on investment (ROI). This study uses empirical data to assess the rate at which these technologies are being adopted and the impact the adoption has had on the efficiency of the retail operation.

The Diffusion Model

Griliches (1957) first used the logistic model to gain insight into the process by which an innovation spreads. He looked at the acreage of hybrid corn planted in the United States and observed the pattern shown in Figure 2. The adoption curve has the following features. At inception of a new technology it undergoes what appears to be exponential growth. The exponential growth curve is defined by ∂x/∂t = k * x that is, the rate at which x can increase is proportional to x. Initially when x is small, its growth rate is also small as is evident in the lower flat portion of the curve in Figure 2. As x slowly increases, its rate of growth increases proportionately until it reaches a maximum at the inflection point in the middle of the curve.

At this point, the diffusion model ceases to resemble exponential growth. The growth rate starts to decay as the adoption asymptotically approaches a ceiling value. This portion of the curve can be described by the asymptotic equation ∂x/∂t = k(xmax - x) that is, as x approaches xmax (some maximum value), the growth rate slows until in the limit the growth vanishes and the curve becomes horizontal. The complete `S' shaped behavior is apparent in many natural phenomena including innovation diffusion.

A new technology is at first not very well known or trusted. First, a few risk takers try out the innovation. If they succeed, others are encouraged to adopt it. Thus the initial growth rate is very low. As the innovation spreads, it becomes accessible to more and more potential adopters. At the same time its credibility goes up and the cost of adoption goes down. Thus the rate of growth steadily increases in non-linear fashion as if there were an infinite number of potential adopters. However, at some point, the number of adopters approaches the same magnitude as the set of all possible adopters. Increasingly, a contact that would have generated an additional adoption is made with a previous adopter and the growth rate slows. When all potential adopters have adopted, the growth ceases.

The entire process is captured by the logistic model which can be written as: ∂x/∂t = -k / (( x/xmax ) * (xmax - x) Integration yields the equation whose curve is shown in Figure 2. x = xmax / (1 + exp(-a -b * t)) where t is time, x is the variable whose diffusion is being measured in time (the number of adopters at any given time t), xmax is the maximum possible value of x ( the potential adopters ), b is the diffusion coefficient, and a is a constant of integration. The equation can be expressed in linear form by taking logarithms ln (x / (xmax - x)) = a + b * t The diffusion coefficient b is readily obtained by fitting the time series data to a linear regression model using the above logarithmic transformation. The three parameters of the diffusion model form a useful summary of the time series data being examined and they can be interpreted in meaningful terms. These interpretations were originally put forward by Griliches and later formalized by Mansfield (1961).

The term b has been the object of most diffusion research. It is an index of the diffusion speed - a single number that characterizes the rate at which the innovation spreads through the industry or firm. High values of b produce steep diffusion curves and low values are associated with gently sloping adoption curves. Thus the b values for two different innovations in the same industry can be compared to yield information about the characteristics of the technology or the industry that abetted or deterred adoption. Similarly, comparisons of the b coefficient for the same innovation in different industries or firms can yield useful information regarding policy or attitudes that affect the implementation of new technology.

The xmax term is a measure of the maximum penetration that can be achieved by a given technology within a given industry or firm. It represents the percentage of the total possible adopters that will ever adopt. It is therefore an important parameter that can be used by MIS managers in setting goals and making implementation schedules. Graphically, the xmax term determines how high the diffusion curve will be. The curve will never cross this line and will only approach it asymptotically.

The a term is the displacement or delay measure and it shifts the entire curve forward or backward. In comparing two firms with the same diffusion rate and ceiling value, the difference in the a term can be interpreted as the amount of time that passed between the initial introduction of the technology in each firm.

In short, the logistics model is very descriptive of patterns of innovation diffusion and its three parameters are a convenient way of summarizing time series data on diffusion. The parameters are easy to compute and are subject to simple and meaningful interpretation. These features of the model have made it very popular in empirical studies of innovation diffusion. Immediately following the seminal work by Griliches, Mansfield (1961) proposed a formal methodology for innovation diffusion models and used the Griliches model to study 12 different innovations in American industry. Since that time numerous empirical studies have been performed using this method.

Randles (1983) used the Griliches model to study the diffusion of computer terminals. The data were taken at a time when a large engineering department was changing over from punched cards to on-line CRT terminals. The diffusion model helped Randles identify the ceiling number of terminals that will ever be used as well as some determinants of adoption. The insight that he gained allowed him to make predictions of terminal requirements in the future.

Chatman (1985) used a modification of the model to study the diffusion of information. Norton and Bass (1987) presented an extension of the model that can be used to study the replacement of one technology by another. As the second innovation begins to rise in adoption rates, the first decays prematurely. The model was applied to the DRAM chips. As the 64K RAM chips began to spread, the earlier 16K chip adoptions began to fall even though they were still in the exponential growth phase of their diffusion cycle.

Leonard-Barton and DesChamps (1988) studied the adoption of expert systems at the firm level and correlated diffusion rates with managerial influence. Cooper and Zmud (1990) applied the diffusion model to the implementation of MRP systems in manufacturing firms and found that the various operating characteristics of the firm could be related to diffusion rates. Trajtenberg and Yitzhaki (1989) proposed an extension of the Griliches model that can be used to study a truncated process; that is, a time series that may represent only the initial segment of the total diffusion curve.

In the present study, the diffusion equation has been used to model the growth of three innovations in retailing. These are, POS bar code systems, hand held terminals, and EDI. The data represent the adoption rate of these innovations industry wide. They are summarized in terms of the Griliches model. The measured diffusion rates are then related to maturity of the technology, costs, and perceived benefits of and barriers to adoption. Data are also presented to assess whether a high rate of adoption reflects achievement of business goals that motivated the adoption.


Table V summarizes the findings of the diffusion study. The data used to generate the table covered the years 1981 through 1989 and were obtained from the NRMA and from Ernst Young who have been conducting annual surveys of retail MIS since 1981. The sample size varies from 127 to 135 and covers merchandise retailers of all sizes in the US. The variable being measured is the percentage of retailers that have adopted a given technology. A firm that has moved beyond the initial pilot stages and has become committed to a technology was considered to have `adopted' it. This definition of adoption is necessarily arbitrary however complete conversion is unnecessarily restrictive.

Figure 2 shows the raw data in graphical form. Magnetic strip readers have been included for comparison with the POS and EDI data. Magnetic strip readers are used at the POS for reading credit card information and obviating the need for a separate credit card imprint operation at check-out. These devices improve check-out efficiency. Their low cost, ease of use and installation, and obvious utility have contributed to a very high diffusion rate evidenced by the seep adoption curve.

The figure also shows that POS systems have been used the longest of the technologies being studied and that their adoption by retailers is fairly complete. EDI is newer and also shows a more gradual adoption curve possibly reflecting implementation problems not associated with POS systems. In contrast, hand-held terminals, though newer have overtaken EDI and appear to be rapidly approaching full adoption by the industry. These visual patterns of adoption dynamics are captured by the diffusion model parameters shown in Table V. These parameters are obtained by using linear regression of the linearized diffusion equation obtained by logarithmic transformation. The coefficient of determination of each regression is reported in the table under `R-squared'.

A comparison of the diffusion rate coefficient b between the four technologies presented in the table yields useful information and provides insight into the diffusion process of these technologies. The extremely rapid adoption of magnetic strip readers is captured in its diffusion coefficient of 2.22, more than twice that of the other technologies. This indicates that there are few barriers to its adoption and there are important perceived benefits to the technology. For similar reasons, the hand-held terminal has a higher diffusion rate than either POS or EDI. Hand held terminals are inexpensive when compared with POS systems and they present few implementation problems if bar code markings are already in place. In that respect, they are a sort of a parasite technology that takes advantage of the technological base built up by the diffusion of POS.

POS/ bar coding itself however did not benefit from any such an advantage. Its low diffusion rate of 0.65 is evidence of several barriers to adoption and the gradual maturation of the technology itself. The systems are very expensive and, until recently, required excessive maintenance especially to the printer unit (McCready 1986). Adoption was also retarded by the time needed to standardize UPC bar codes and by the availability of vendor marked goods . Industry surveys also suggest that the reliability of the laser bar code readers has been low until recently and may have contributed to a lower adoption rate than might have been expected given the benefits.

EDI shows a higher diffusion rate than POS bar coding. This is surprising in light of the extreme difficulty of implementing EDI. Implementation requires a joint project with the vendors and, as described by Munshi (1990), the critical nature of the data to the financial success of the firm requires a `bullet-proof' installation with a low error tolerance. Data integrity and data security issues are expected to be significant barriers to the rate of adoption. Additional barriers to EDI adoption have been the lack of industry wide standards for communication protocol and legal, contractual, and psychological issues presented by business contracts that were not `on paper' and which contained no signature.

On the other hand, the technology of EDI is surprisingly simple with low hardware and software costs . Most systems involve little more than a simple ASCII file transfer over modems. Yet the benefits (see table IV) are significant to the retailer as EDI is an essential ingredient in the deployment of QR purchasing systems. It is possibly on the strength of these benefits and the facility provided by the VICS/ANSI standardization that the EDI diffusion coefficient is higher than POS bar coding.

However, the data indicates that the rate of adoption of EDI is slowing prematurely. This is reflected in the maximum expected adoption rate of 80%. The reason for such a low ceiling rate is possibly the existence of an entire class of retailers that would not use this technology. Further research is necessary to explore this question and to identify those retailers that are not candidates for EDI adoption. The ceiling values for POS and magnetic strip readers is predicted to be 100% while that of hand held terminals is slightly lower at 95%.

The magnitude of the constant of integration, a, is a measure of how recent the innovation is. For example, magnetic strip readers (a=-17.2) are a very recent innovation while POS (a=-3.3) is much older. The percent of retailers in the sample that had adopted these technologies by the end of 1988 are shown in the last column of Table V.

Having measured the diffusion rates and extent of adoption of each technology, it is now possible to compare these with the industry's operating data to assess the impact of these information technologies on business operations; that is, we seek evidence of their effectiveness. The survey data summarized in Tables III and IV suggest that the implementation of POS systems and EDI should have an impact specifically on the inventory turn over in the retail industry as a whole. The inventory turn over is a ratio that indicates the how well managers use inventory to generate sales. Numerically, it represents the sales revenue generated by $1 invested in inventory. A more general measure of the managers' efficiency in using assets to generate sales is the asset utilization ratio (TAU) that measures the amount of sales revenue generated by a dollar of total assets including inventory.

The other measure business operations is profitability and it is governed by how well the managers control costs. The traditional measure of profitability is the profit margin (PM) computed as the ratio of total profit before interest or taxes (EBIT) to sales. By excluding depreciation from this computation, profitability can be measured analogous to cash flow as the ratio of the income before interest depreciation or taxes (EBIDT) to sales. In addition to PM, retailers also pay close attention to their gross margin (GM) equal to the gross profit over sales. This measures the managers' ability to purchase goods at a low cost.

Figure 3 shows the industry average turnover ratios and asset utilization ratios for the period of the diffusion study plotted against the POS adoption rate. Visually, the graph does not suggest any trends or dramatic improvement in these measures with increasing penetration of POS systems.Figure 4 shows equally unimpressive gains in profitability measures although an upward trend in GM could be inferred.

The trends are statistically tested using linear regression against time and against lagged values. The results are shown in Table VII. No significant trend is apparent. The regression against lagged values are expected to yield slopes greater than unity for steadily increasing values. All the slopes obtained are less than one. Scatter plots of the data confirm the absence of any trend or periodic behavior.

An additional measure of managerial efficiency is defined and shown in Table VII. Called `expense efficiency', it measures the ability of managers to control administrative and marketing costs at corporate headquarters and numerically represents the amount of sales generated by a dollar of expenses. No improvement in this ratio is apparent over 19 years.

The most important result of this analysis is that the inventory turn over ratio in the industry is approximately 6 and seems to be fairly constant with time and across technologies. The data do not suggest that the rapid diffusion of information technology in the retail trade has had an observable effect on operating efficiency.

Conclusions and Summary

The study parallels similar industry level studies by Mansfield in 1961 and shows that the diffusion model can be applied to summarize and interpret time series data on the adoption of new technologY by the retail industry. The diffusion coefficient for various technologies can be compared to gain insight into the implementation issues as well the costs and benefits of adoption. Both POS bar coding systems and EDI have overcome many obstacles to adoption and achieved high diffusion rates due to the perceived benefits by retailers. Retailers hold the view that these technologies will help them control inventory and implement QR purchasing.

However, control of inventory implies an improvement in the inventory turn over ratio. An analysis of the industry's turn over ration over the past two decades does not show any trends that can be attributed to technology . No improvement in any other generally accepted measure of business operations is evident in the data except for a slight increase in gross margin.

A legitimate question is therefore. Why do retailers automate?. The answer may lie in centralization and concentration of control . For example, one large and very successful retailer has been able to achieve improvement in gross margins and this improvement corresponds with increased automation of inventory and purchasing. It is possible that the centralization of purchasing allows the retailer to buy at a lower price by increasing their bargaining power .

However, effectiveness issues can be addressed easily at the firm level. It is therefore proposed that further research in retail innovation should concentrate on obtaining firm level data on adoption. Comparison of diffusion rates with managerial and operating characteristics can then be used to infer the determinants of innovation diffusion rates and the impact of innovation penetration in terms of operating efficiency.


1981 1982 1983 1984 1985 1986 1987 1988 1989
Industry 93.37 101.1 105.2 112.3 125.0 136.7 156.7 167.8 198.2
Sears 25.19 27.36 30.02 35.88 38.83 40.72 44.28 48.44 50.25
K Mart 14.34 16.68 16.94 18.79 21.30 22.64 24.05 25.86 27.55
JC Penney 11.35 11.86 11.41 12.08 13.45 13.75 14.74 15.33 15.30
Wal-Mart 1.65 2.44 3.38 4.67 6.40 8.45 11.91 15.96 20.65
Dayton H 4.04 4.94 5.66 6.96 8.01 8.79 9.26 10.68 12.20
May 3.17 3.41 3.67 4.23 4.76 5.08 10.38 10.58 11.74
Woolworth 7.22 7.22 5.12 5.46 5.74 6.96 6.50 7.13 8.09
M Ward 3.40 4.00 4.40 4.70 5.02 4.87 6.07 4.64 4.50
Carter H 2.63 2.87 3.05 3.63 3.72 3.98 4.09 3.35 2.62
Dillard 0.47 0.59 0.71 0.85 1.28 1.60 1.85 2.21 2.56

% of Respondents
5 Fully Integrated 10
4 36
3 32
2 15
1 Not integrated 6


Barriers to Diffusion Hardware reliability problems 15%
Costs excessive for small stores 20%
POS is essentially a defensive action 11%

Aiding the Diffusion Process Reduction in inventory levels 58%
Increased inventory turn over 55%
Reduced mark downs 51%
Increased gross margins 72%
Improved customer service 38%
Reduced store operations cost 13%


Communication Protocol Percent VICS ANSI X.12 62%
Proprietary 23%

Key Advantages of EDI Rank
Reduce inventory levels Shorten re-order cycle Quick Response purchasing 1 (most important) or 2 (second)

Improved accuracy of data Reduced cost of ordering Streamlined merchandise handling 3 (least important) or - (not important) Barriers to Adoption Rank

Lack of Industry Standards Implementation Issues 1 or 2

Lack of understanding Investment costs 3


POS The percentage of retailers who have adopted POS bar code
HAND The percentage of retailers who have adopted hand held terminals
EDI The percentage of retailers who use EDI for a significant portion of their purchase orders.
MAG Magnetic strip readers at POS

Technology Xmax a b R-squared Adoption

POS 100% -3.3 0.65 0.987 93%
HAND 95% -6.65 0.88 0.986 80%
EDI 80% -6.35 0.72 0.978 45%
MAG 100% -17.2 2.22 0.900 100%

Managerial Characteristic Percent of Retailers

Uses Steering Group 60%
Has Long Range MIS Plan 34%

Managerial Variable Value

MIS Budget as Percent of Sales 1.12%
$Millions of Sales per MIS Employee $8.15



1971 TO 1989 (n=313)

Variable Mean Std. Dev. b-Time F b-lag F

Inventory turn over 5.79 1.655 0.022 0.65 0.9114 1209
Asset utilization 3.15 0.824 0.011 0.97 0.8567 1139
Gross Margin % 31.1 4.753 0.003 1.44 0.9519 2766
Profit Margin % 7.28 2.495 0.002 1.12 0.8621 863
ROI % 22.2 8.216 0.009 1.05 0.8015 558
Expense Efficiency 4.37 0.896 -0.022 1.31 0.9696 3714

Std. Dev = standard deviation, b-Time = the regression weight for time, b-lag = the regression weight for one-year lagged values, F = F-statistic


Allen, Randy L., POS Trends in the 80s, NRMA Report , National Retail Merchants Association, (1982)

Allen, Randy L., and Daniel W. McCarthy, Point of Sale: Current Trends and Beyond, NRMA Report, National Retail Merchants Association, (1986)

Anonymous1, Barcode/EDI Survey Results, Touche-Ross and the National Retail Merchants Association, (1989)

Anonymous2, Barcode/EDI Survey Results, Touche-Ross and the National Retail Merchants Association, (1990)

Anonymous3, EDI and QR: A Lot More Than Alphabet Soup, Chain Store Age Executive, January (1988), pp. 89-90

Anonymous4, Levilink Makes QR a Reality, Chain Store Age Executive, January (1988), pp. 92

Anonymous5, Today's Needs Versus Tomorrow's Possibilities: POS Systems are Being Used For Much More Than Collecting Sales Data, Chain Store Age Executive, January (1988), pp. 94-99

Anonymous6, Retail MIS Field Reflects Technology Trends, Chain Store Age Executive, January (1988), p.100-102

Anonymous7, Retailers Learn To Handle Data Capture: Industry Getting a Better Grip On Exploiting Available Technology, Chain Store Age Executive, January (1988), pp. 106

Anonymous8, JC Penney Ready to Ride the Third Wave, Chain Store Age Executive, January (1990), pp. 72-74

Anonymous9, Talbots Links Banking Credit Authorization: Retailer Develops Cost Effective Ways to Handle Transactions, Chain Store Age Executive, January (1990), pp. 81-85

Anonymous0, MIS Budgets: EDI Up Bar Coding Down: Retailers Experimenting With New Applications, Says Survey, Chain Store Age Executive, January (1990), pp. 96-98

Anthony, R. N., A Framework For the Analysis of Planning and Control Systems, Harvard Graduate School of Business Administration, (1965)

Babcock, Charles, Information Systems Sharpen Retailers' Competitive Edge, Computerworld, Vol. 20 January 20 (1986) pp. 14

Bailey, James E., and Sammy W. Pearson, Development of a Tool for Measuring and Analyzing Computer User Satisfaction, Management Science, Vol. 29 (1983), pp. 530-544

Bakos, Yanis J.,, Dependent Variables for the Study of Firm and Industry Level Impacts of Information Technology, Proceedings of the Eighth International Conference on Information Systems, (1987) pp. 10-23

Barnett, George A., and Gary Siegel, The Diffusion of Computer Assisted Legal Research Systems, Journal of the American Society for Information Science, Vol. 39 (1988), pp. 224-234

Baroudi, Jack J., Margarethe H. Olson, and Blake Ives, An Empirical Study of the Impact of User Involvement on System Usage and Information Satisfaction, Communications of the ACM, Vol. 29 (1986) pp. 232-238

Baroudi, Jack J., and Wanda J. Orlikowski, A Short Form Measure of User Information Satisfaction: A Psychometric Evaluation and Notes on Use, Journal of Management Information Systems, Vol. 4 (1988) pp. 44-50

Brown, Lawrence, Innovation Diffusion: A New Perspective, Methuen Press, New York, (1981)

Burrell, G. and G. Morgan, Sociological Paradigms and Organizational Analysis, Heinemann, London, (1979)

Chatman, Elfreda A., Diffusion Theory: A Review and Test of a Conceptual Model in Information Diffusion, Journal of the American Society for Information Science, Vol. 37 (1986), pp. 377-386

Cheney, Paul H., Robert I. Mann, and Donald L. Amoroso, Organizational Factors Affecting the Success of End-User Computing, Journal of Management Information Systems, Vol. 2 (1986) pp. 65-80

Cook, Kathleen, Retail MIS: Expenses and Trends, Ernst Young Sixth Annual Report on Retail MIS, Ernst Young, New York, (1987)

Cooper, Randolph H. and Robert H, Zmud, Information Technology Implementation Research: A Technological Diffusion Approach, Management Science, Vol. 36 No. 2 (1990) pp. 123-139

Crowston, Kevin and Michael E. Treacy, Assessing the Impact of Information Technology on Enterprise Level Performance, Proceedings of the Seventh International Conference on Information Systems, (1986) pp. 299-310

De Lio, Richard M., Retail MIS: Expenses and Trends, Ernst Young Seventh Annual Report on Retail MIS, Ernst Young, New York, (1988)

Doll, William J., and Gholamreza Torkzadeh, The Measure of End-User Computing Satisfaction, MIS Quarterly, Vol. 12 (1988) pp. 259-273

Goslar, Martin D., Marketing and the Adoption of Microcomputers: An Application of Diffusion Theory, Journal of the Academy of Marketing Science, Vol. 15 No. 2 (1987) pp. 42-48

Griliches, Zvi, Hybrid Corn: An Exploration in the Economics of Technological Change, Econometrica, Vol. 25 (1957), pp. 501-522

Habern, Glen L., Retail MIS: Expenses and Trends, Ernst Young Eighth Annual Report on Retail MIS, Ernst Young, (1989)

Hansen, James V., and Ned C. Hill, Control and Audit of Electronic Data Interchange, MIS Quarterly, December (1989), pp. 403-412

Hirschheim, Rudy, and Heinz K. Klein, Four Paradigms of Information Systems Development, Communications of the ACM, Vol. 32, (1989), pp. 1199-1216

Ives, Blake, Margarethe H. Olson, and Jack J. Baroudi, The Measurement of User Information Satisfaction, Communications of the ACM, Vol. 26 (1983) pp. 785-793

King, William R., and Jaime I. Rodriguez, Evaluating Management Information Systems, MIS Quarterly, September (1978) pp. 43-51

Leonard-Barton, Dorothy, Experts as Negative Opinion Leaders in the Diffusion of a Technological Innovation, Journal of Consumer Research, Vol. 11 March (1985) pp. 914-926

Leonard-Barton, Dorothy, Isabelle DesChamps, Managerial Influence in the Implementation of New Technology, Management Science, Vol. 34 (1988) pp. 1252-1265

Lucas, Henry C. Jr., Performance and Use of an Information System, Management Science, Vol. 21, (1975), pp. 908-918

McCready, A.C., Retail MIS: Expenses and Trends, Ernst Young Fifth Annual Report on Retail MIS, Ernst Young, New York, (1986)

McFarlan, F.W. and J.L. McKenney, The Information Archipelago - Gaps and Bridges, Harvard Business Review, Vol. 60 No. 5 (1982) pp. 109-119

Miller, Jonathan, Information Systems Effectiveness: The Fit Between Business Needs and System Capabilities, Proceedings of the Tenth International Conference on Information Systems, (1989) pp. 273-288

Mansfield, Edwin, Technical Change and the Rate of Imitation, Econometrica, Vol. 29 (1961), pp. 741-766

Munshi, J., MIS: Cases In Action, McGraw Hill Book Company, NY (1990)

Munshi, J., A Framework for MIS Effectiveness, SW DSI Meeting, Dallas, 1991

Norton John A., and Frank M. Bass, A Diffusion Theory Model of Adoption and Substitution Successive Generations of High Technology Products, Management Science, Vol. 33 (1987) pp. 1069-1086

Olshavsky, Richard W., Time and Rate of Adoption of Innovations, Journal of Consumer Research, Vol. 6 March (1980) pp. 425-428

Raho, Louis E., James A. Belohlav, and Kirk D. Fiedler, Assimilating New Technology Into the Organization: An Assessment of McFarlan and McKenney's Model, MIS Quarterly, March (1987) pp. 47-56

Robertson, Thomas S. and Hubert Gatignon, Competitive Effects of Technology Diffusion, Journal of Marketing, Vol. 50 July (1986) pp. 1-12

Rogers, E.M., The Diffusion of Innovations, Third Edition, The Free Press, New York, (1984)

Randles, F., On the Diffusion of Engineering Terminals in an Established Engineering Environment, Management Science, Vol.29 (1983), pp. 465-475

Sharda, Ramesh, Steve H. Barr, and James C. McDonnell, Decision Support System Effectiveness, Management Science, Vol. 34, (1988) pp. 139-159

Srinavasan, Ananth, Alternative Measures of System Effectiveness, MIS Quarterly, September (1985) pp. 243-253

Trajtenberg, Manuel, and Shlomo Yitzhjaki, The Diffusion of Innovations: A Methodological Reappraisal, Journal of Business and Economic Statistics, Vol. 7 (1989), pp. 35-47

Trice, Andrew W., and Michael E. Treacy, Utilization as a Dependent Variable in MIS Research, Proceedings of the Seventh International Conference on Information Systems, (1986) pp. 227-239

von Neumann J., and O. Morgenstern, Theory of Games and Economic Behavior, Princeton University Press, Princeton, NJ, (1944)