OMEGA Int J of Mgmt $ct, Vol 17, No 6, pp. 543-550, 1989 Pnnted m Great Britain
0305-0483/89 $ 3 0 0 + 0 0 0 Pergamon Press pk
Sources of Personal Computing Problems T GUIMARAES St Cloud State Umversity, Minnesota, USA
V RAMANUJAM Case Western Reserve Umversity, Ohio, USA (Recetved October 1989, m revised form March 1989) The rapid growth of personal computing in organizations has given rise to numerous problems of
information management. Broadly these proHems encompass the categories of evolvabllity and integration of systems, cost of computing,data integrity and security, and MIS department--mer deportment relationships. A framework is proposed linking these problem ¢ategor~ and three organizational factors, viz., level of use of personal computing, extent of control over personal computing,and degreeof support for personalcomputing.Correlational,regression,and moderated rngresaionanalysistechniquesare employedto teat a mnnherof hypothesesrelatingthe organizational factors and problemscategories.Resultsindicatethat controlover personalcomputingis the strongest predictor of the extent of personal computing problems encountered and that, while various commonly
used support meaauresare helpfulin mitigatingMIS--u~r relationshipproblems, they do not seem to he effectivein denlingwithother typesof personalcomputingproblems.Controlis foundto exert both • directand an interactiveeffecton problemintensity.Implicationsof these findingsfor managing personal computingin organizations are noted. Key words--personal computing, mformauon management, mformatsoncenters. MIS planmng
INTRODUCTION THE EMERGENCE of end-user computing and
personal computing as viable alternatives to traditional computing represents a major and often traumatic transition wtthin the computing environments of most organizations. In many ways, end-user computing and personal computing are welcome phenomena as they afford the decentralization of the MIS funcUon consistent with the decentralization of other organizational functions and decision-making processes. In particular, personal computing makes it possible to finally realize the long pursued and often elusive goal of putting decision support at the user manager's fingertips
[10, t2, 17]. Yet, the rapidity with which the end-user computing and personal computing phenomena have been unfolding has caught many MIS managers by surprise and is giving rise to
numerous computing problems. Because they represent the ultimate in user control, personal computing environments are particularly apt to experience these problems, which range from data integrity and security ~ssues to poor or deteriorating relationships between the MIS and user departments. End-user and personal computing problems have been widely discussed in the literature [1, 7, 10, 13, 14]. They have also attracted considerable attention in the business press [2, 3, 4]. This study attempted to relate a variety of previously identified personal computing problems to a set of organizational factors that were hypothesized to have an influence on their incidence and intensity. A conceptual model of the problems and the organizational factors presumed to be related to them has been proposed by Guimaraes and Ramanujam [8]. This study was aimed at empirically testing that model which is discussed next. 543
544
Gutmaraes, Ramanujam--Sources of Personal Computing Problems
A CONCEPTUAL M O D E L O F PERSONAL C O M P U T I N G PROBLEMS
The model is shown m Fig. !. It proposes various interrelationships between categories of personal computing problems and the organization factors believed to influence them. As presented in Table 2, personal computmg problems may take many forms but usually fall into one of four broad categories.
the hardware used by other users m their department or company, creates ddticulties of integration. In many cases when users need to work cooperatively, such integration becomes essential,
2, Cost-related problems (COST)
These arise because of redundancy in the applications developed by different users and due to ineflictencies in negotiations with vendors, etc. Because of lack of technical knowl1. Evolvability and integration problems edge and mexpertence on the part of users, who (EVOL VINT) often attempt to circumvent the influence of a These problems arise because of the fact that central MIS department m hard~are selection personal computing apphcations are developed and acquisition through various capttal budgetfor the most part by individual users for thetr ing and procurement ruses and shortcuts, often own personal decision support [9]. Organiz- incorrect choices are made as new users learn by ational considerations become secondary and trial and error. As a result, various tangible and documentation is either absent or inadequate. intangible costs arise that undermine the effecThese difficulties often come to light when a tive use of the personal computing alternative. particular user is transferred or leaves the company. The applications developed by this 3. Data integrity and security problems user are either lost to the company or create aNTSEC) headaches for his/her successor. Also, the fact These problems multiply with the prohferthat the applications run on different hardware, ation of personal computers especially in enwhich may have been chosen by users on their vironments where security procedures are esther own whtms without regard to compatibility with lacking or loosely implemented and enforced.
I
Level of Personal
Computing Usage (PCUSE)
l ..Ii
Level of Control
(CORTROL)
I |
Level of Support
(SUPPORT)
/ INCIDENCE A N D INTENSITY OF PERSONAL COMPUTING PROBLEMS
-
B v o l v a b i l i t y and i n t e g r i t y (ZVOLVZRT)
-
Costs (COST)
-
Data integrity and security
(INTSEC) MIS-Ussr
(~LTNS)
relationship
Fig. 1 A model of factors impacung personal computmg problems
Omega, VoL 17, No 6 Contamination of corporate data is said to be a widespread problem in companies where personal computer growth has gone largely unmanaged.
4. MIS-user department relationslu'p problems (RELA TNS) Relattonships between MIS and user department often undergo severe stress in companies just beginning to manage widespread personal computing. With some exceptions, MIS departments have resisted the introduction of the personal computer as they view tt as leading to the loss of their 'centrality' [11], as far as computing is concerned. On the other hand, users not infrequently take to personal computing in an effort to mitigate their dissatisfaction with traditional computing. The adoption of the personal computer by users under these circumstances exacerbates the tensions between MIS and users The incidence and intensity of personal computmg problems are thought to be affected by many organizational factors which are presented in Table 4 and fall into three groups:
(i) Level of personal computer usage (PCUSE) Intuitively, the level of personal computing problems can be expected to be related to the level of personal computer usage. The level of personal computer usage is an index of the importance of personal computing to the organization and is posited to be a key determinant of the incidence as well as intensity of personal computing problems. While, up to a point, we would expect to see an increase in the incidence of personal computing problems as personal computers proliferate in the organization, with the increasing importance of personal computing, steps to deal with personal computing problems are likely to have been taken. Thus, the directionality of this linkage will be contingent on the two factors discussed next.
(ii) Level of control (CONTROL) Some authors have argued that the chief source of personal computing problems is lack of planning and control over personal computer acquisition [13]. In organizations that profess a higher degree of attention to planning and control over personal computer acquisition and usage, we should accordingly expect to see a lower level of personal computing problems.
545
(iii) Leoel of support (SUPPORT) While planning and control are proactive or strategic measures, they alone cannot ensure freedom from personal computing problems. Accordingly, many researchers have proposed a more active role for MIS departments in helping users deal with the problems that do arise m the day to day use of personal computers. Various support mechanisms can be provided to the users by the MIS department ranging from help desks to training seminars and courses. These can be considered to be the tactical or operational approaches for dealing with personal computing problems. Strong tactical approaches are often a necessary condition for implementing any strategy.
PROPOSITIONS BASED ON THE MODEL This study aimed to test the validity of the following expected relationships among the organizational factors and the types of personal computing problems which were proposed by the model [8]:
I. PCUSE and intensity of personal computing problems As personal computer usage increases, the incidence of personal computing problems will also increase.
2. CONTROL ing problems
and intensityofpersonal comput-
Anecdotal accounts [e.g. 13] as well as intuitive reasoning suggest that greater levels of control over personal computer acquisition and usage is likely to lower the incidence and intensity of personal computing problems. It should be recognized that greater control over personal computer usage may have a salutary effect over some specific categories of personal computing problems such as evolvability and integration, costs, and integrity and security, but may have much less of a beneficial influence (or even a negative influence) on the MIS user relations problems. Indeed, any increased control by the MIS department over personal computer acquisition and usage may actually strain relationships between the two if users perceive it as unwarranted encroachment on their turf.
546
Gulmaraes, RamanujamwSourcesof Personal Computing Problems
3. SUPPORT and intensity of personal computing problems Along the same hnes as the arguments for CONTROL, the model suggests that the greater the level of support for personal computing usage, the lower will be the incidence and intensity of personal computing problems. Again, the effect of SUPPORT may be problem-specific. Support may lead to better MIS-user relationships and may help control some costs but may also be less beneficial m deahng with the more 'strategic' types of personal computing problems included m EVOLVINT or INTSEC. It is also reasonable to expect that the effects of PCUSE, CONTROL and SUPPORT may be interactive, in addition to each variable exerting an independent effect over personal computing problems. Only the three two-variable interacuons were considered in this study. Three way interactions would be too difficult to test effectively, and have therefore been ignored. The following interactive relationships have been tested: 4. PCUSE and control As the level of personal computing activities grows, their total costs to the orgamzat~on increase. This may attract management attention and raise the need for control. In low CONTROL environments, it is reasonable to expect a sharper increase in personal computing problems as PCUSE increases. In high CONTROL enwronments, either the level of personal computing problems will be unaffected by level of PCUSE, or increasing PCUSE may actually lead to reduced incidence of personal computing problems. Thus, the form as well as the strength of the relationship between PCUSE and level of personal computing problems is likely to be 'moderated' by CONTROL. The model suggests that there will be a significant interactive effect between PCUSE and CONTROL on personal computing problems. 5. PCUSE and SUPPORT Along the same lines as the arguments for PCUSE and CONTROL, interactive effects are posited for PCUSE and SUPPORT: There will be a significant interactive effect between PCUSE and SUPPORT on personal computing problems. Another hkely relationship which will not be tested in this study is that as the level
of computing activity in the organization increases, end-user requests are likely to force an increase in support activities. That, in turn, invites further increases in personal computer use in the organization.
CONTROL and SUPPORT Control over personal computing implies some kind of centralized or standardized policies and procedures aimed at minimizing various kinds of personal computing problems. Thus CONTROL has a strategic connotation. While SUPPORT has the same ultimate aim, it has a more operational or tactical connotation. The following relationship is expected: The benefit of CONTROL on personal computing problems will be greater for higher levels of SUPPORT than for lower levels. METHODOLOGY
A--The sample Data for empirically validating the above model and its related propositions were collected by means of questionnmres mailed to top computer executives. Four hundred companies, chosen randomly from the FORBES hst of 808 compames for 1983, were sent the three page questionnaire. One hundred and seventy three of them, representing a response rate of 43.3% of the target sample, returned completed questionnaires. All provided usable responses. This level of response rate from a population of the largest public companies m the country is higher than is usually normal for mail surveys of comparable populations, e.g. the FORTUNE 500 hsts [5]. The high response rate appears to reflect both the importance of the topic as well as the brewty and conciseness of the questionnaire. Despite the relatively h~gh response rates, tests of non-response bins were undertaken. Such tests confirmed that the sample of the study ~s free of non-response bias with respect to the key demographic variables summarized in Table I for the respondent sample. As Table 1 shows, the sample includes a diversity of firms in terms of their size (gross revenues), primary industry classifications, and MIS budgets. Since the purpose of the study is to establish broad initial generalizations about the relationship between organizational factors and personal
Omega, Vol 17, No 6 Table I Charactensucs of compames In the sample (n - 173)
547 Table 2. Malu-ttem scales for PC problems
i,
17
L Ew4mbdlry amdi~efraaoa peobkms(E¥OL FIN?') was composed of the followi~q $ item:
25 I$ 29 II 3
~I..ack of compattbdtty among toreros --Lack of compaubthty wtth corporate mainframe --Lack of compaubtltty wtth peripherals --Mtsmatch of apphcatioos and (best) eompuUng alternattves ~ L a c k of torero-mainframe mtegratton
29 21
2. Cost reded Foblems (COST) was eoAvosed of t ~ followu~ 6 item:
L Re~eMes in S ~ / i ~ s 200-499 500-1199 1200-1999 2000-4999 5000-10000 Over 101300 2. I ~ s t , ry Manufactunng Banking, thrift Institutions and msurauee compantes Elecmc and gas utthues Off refining and marketing Retad Others
5 4 3 38
3. Company MIS budget L¢ $ milkoas 2-4 9 5-9 9 10-24 9 Over 25
17 29 28 26
--Redundancy of apphcatton$ --Trial and error m software package aeqmsmon and use --Trial and error in uegoUauons wnh vendors --Poor mamtamabthty of end-user applzcatnon --Lack of vendor support for hardware --Lack of user experience with hardware/software products
~ Data mtegrrty and seewtity problems (INTSEC) was composed of the followiRI 4 Items:
All figures are percentages Non-responses are excluded
--Contammauon of corporate data on mainframe --Lack of user knov,ledge regarding integrity --Lack of user knov,ledge regardmg security --Mahctous/unauthonsed user behavior regarding micros
computing problems, the heterogeneity of the sample is believed to be a strength.
& MIS---user departmem relationsMp problems (RELATNS) composed of the foUowiq 3 items:
B~The measurements Incidence of PC problems. A check list of personal computing problems, compiled by Guimaraes [7] was presented to the respondents who rated the felt intensity of each problem on a five-point Likert scale. In order to obviate the possibility of obtaining inflated measure of scale reliabilities at a later stage, the order of presentation of the problems to the respondents was randomized on the questionnaire. The problems were subsequently classified into the four categories discussed earlier. Analysis of the reliability or internal consistency of these multi-item scales was undertaken to identify 'outlier' items and to further refine the multi-item scales. This procedure resulted in the four problem scales shown in Table 2. Means, standard deviations, reliabilities, and intercorrelations among the four composite scales are shown in Table 3. PCUSE. This variable was measured by a smgle-ttem estimated by IS directors indicating the proportion of total organizational computing accounted for by personal computing. CONTROL and suPPORT. These two factors were measured by the multi-item scales shown in Table 4. The problem types and organizational factors were specified a priori in this study. An alternative approach may have been to use factor analysis to derive these dimensions. However, the presence of considerable support for the dimensions of this study in the literature as well as the good internal consis-
was
--Stram m MlS--user relattonshtps due to mtcros --Lack of commumcatton between MIS and user departments --Poor tmage of MIS department All problems were rated by the respondents on a 5-point Lnkert scale ranging from I (tMs problem ts of no concern) to 5 (thns problem ns of extreme concern)
tency values of the scales lends support to our claim of their appropriateness m this exploratory study. We share the view that in the absence of good prior theory, use of techniques such as factor analysis is likely to be problematic and may lead to uninterpretable, albeit orthogonal, factors.
C--Dara analysis Correlational, regression, and moderator regression analysis techniques were used to test the six hypotheses developed earlier. In moderator regression analysis, contingency relationships are explored by introducing an interaction term of the moderator variable (X2) and the independent variable (XI) whose effects on the dependent variable (Y) it is hypothesized to moderate. Thus in the model, Y = ao + at Xl + a2Xi X=, Table 3 Means. standard devmttons, rehabdmes and ratercorrelations of the composnte dependent variables (PC problems) (n - 173)
Variable
Mean
SD
I. 2. 3 4
291 2 78 2.90 2 55
092 0.61 0 74 0 82
EVOLVINT COST INTSEC RELATNS
I
2
3
4
(0 82) 0 69 (0 65) 0 46 0 69 (0 67) 0 08 0 25 0 34 (0 57)
Rehabdnt~s (cronbach alphas) are shown wnhm parentheses as duagonal entries of the correlation matrix
548
Gummraes, Ramanujam--Sources o f Personal Computing Problems Table 4 Mulu-ttem scales for control and support
CONTROL was aUltoud of eke followittg 7 items: --Centrahzed control over selecUon of equipment by user/MIS department --Level of company-wtde IS planning --MIS department control over hardware/software uttlo.auon --User department control over hardware/software utthzatton -.-Centrahzed management over corporate data resources --AvadabdRy of eqmvalent nuunframe and personal compuuns software --Cost-benefit rauo for mainframe computing vls.i-ws personal compuUns tn key apphcauons
SUPPORT was eompoud of tlw followi~ 8 items: --Training programs --Help desk --Local area networking support --Personal computing hbrary --F.qtapment demonstrattons ~Mtero-mamframe links --Abdtty to meet user requests for mformatton --Level of staffing These variables were rated on five-point scales, usmg appropriate adjecttve Parrs For example, all CONTROL vartables were rated on a five-point scale ranging from I (very low) to 5 (very hqgh). The first sex SUPPORT variables were rated on five-point scales ranlpng from I (none or very httle use of thts mechamsm) to 5 (extenstve use of tlus mechamsm) The last two SUPPORT variables were included for internal vahdRy checking and were rated on five~pomt scales rangtng from I (very low) to 5 (very high)
the size and significance of the coefficient a2 gives clues as to the nature and strength of the moderator effect of X2 on the relationship between X, and Y. A positive and significant a2 would mean that the effect of Xt on Y is higher for higher values of X2. A negative and significant a2 implies that the effect of Xm on Y is higher for lower values of X2 and vice versa. Should a2 prove to be non-significant, we can conclude that the hypothesized moderating effect does not exist. An interaction, if present, can be further analyzed graphically to more precisely uncover the monotonic or non-monotonic nature of the contingency effects [15]. RESULTS AND CONCLUSIONS Table 6 shows the intercorrelations between the three organization factors studied and the Table 5 Means, standard devtauons, rehabthttes and mtercorrelattoos for the compostte independent vat'tables (n - 173) Intercorrelat]ons Variable
Mean
SD
PCUSE
CONTROL
I PCUSE 2 CONTROL (0 57) 3 SUPPORT(0 73)
9 36 2 84 2.51
6 99 0.57 047
0 02 008
0 10
Rehabthttes for CONTROL and SUPPORT scales are shown wRhm parentheses
four composite problem variables. It shows that, PCUSE has strong and significant negative correlations with INTSEC and RELTNS, a somewhat weaker but significant positive correlation with EVOLVINT, and a very weak and non-significant correlation with COST. These results imply that the level of use of personal computing is related in non-obvious ways to the intensity of the personal computing problems. The finding regarding cost was unexpected. The results regarding the measure of control over personal computing are unequivocal. CONTROL has significant negauve correlations with all four problem categories. It is shown to have the strongest impact on data integrity and security problems, and a significant but smaller impact on system evolvability and integration problems. Contrary to expectations, the level of personal computing support is found to be unrelated to the intensity of three of the four types of personal computing problems. Only MIS-user department relations problems are strongly and reversely related to the level of support. Thts suggests that the primary benefit to be expected from high levels of personal computing support is likely to be improved MIS-user department relations, and that increased support represents
Table 6 lntercorrelauons between independent and dependent variables Dependent variables PC problems Independent variables PCUSE CONTROL SUPPORT **Stgntficant at P < 0 01
EVOLVINT
COST
INTSEC
RELTNS
0.22** - 0 17"" 0 01
-0.09 - 0 36** - 0 04
-0.44** - 0 44** 0 00
- 0 33** - 0 39** - 0 32**
Omega, Vol 17, No 6
an ineffective approach for dealing with other types of personal computing problems. Another explanation for this result is that there is a substantial time lag between the introduction of support mechanisms and any noticeable effect on personal computing problems. Perhaps, only in the long run can increases in personal computing support be expected to provide benefits to the organization by reducing personal computing problems. Regression analysis was used to further assess the proposed relationships. The three organization factors (independent variables) were regressed on each problem type (dependent variables). The regression results basically reinforced the results of the correlational analysis. The interaction effects between the organizational factors were assessed by running a series of regression models for each problem and measuring the explanatory power of the models through their R 2 values. Specifically, 16 models were run for each problem type. For the sake of space, the details of the models are not presented here but can be obtained from the authors. The following is a summary of the results and conclusmons from the regression analysts. The results indicate that the expected interaction between the level of personal computing activity in the organization and the level of centralized control is significant as far as system evoivabihty and integration problems are concerned. The inverse relationship between the level of computing activity and evolvability/ integrity problems is much stronger when the level of control is low. The results are the same for the cost related problems, and data integrity and security problems. However, level of control apparently has no impact on the MIS-user department relationship problems. The results from regression analysis do not confirm the expected interaction between level of computing activtty and support. Neither do they confirm the interaction between level of control and support. A likely explanation is that s~gnificant post-implementation time is necessary before the various support mechanisms have an effect on personal computing problems.
549
of interest to professional supporting and managing personal computing in large organizations: (1) As the level of personal computing activity increases (and it will) t h e types and intensity of problems will follow. To ultimately avoid end-user frustration and high costs to the organization, the MIS department, in cooperation with internal auditing and end-user departments, must take the iniUative to choose appropriate support and control mechanisms. (2) Control measures, as defined in this study, have been found to have a significant impact on the four major types of personal computing problems, w~thout exception. Organizations should use a cooperative approach between the various parties involved to reduce the hkelihood of 'turf wars'. The results indicate that in companies with a h~gh level of personal computing activity, the implementation of control measures have a more noticeable effect on the intensity of personal computing problems than m companies with lower personal computing levels. (3) In the short run, tactically oriented support measures (such as training programs, hbraries, new equipment, etc.) are quite ineffective in deahng with many, if not most, types of personal computing problems. This finding reinforces Keen and Woodman's contention [12] that the effecuve management of personal computing requires a long term strategic orientatton. Therefore, organizations must carefully plan the types of support mechanisms they will offer to their personal computing community and avoid the ubiquitous 'quick fix'.
ACKNOWLEDGEMENTS SUMMARY AND RECOMMENDATIONS Through the process of testing a previously proposed model, this study has found evidence
This paper was a collaboraUveeffort ~th equal contnbutson from each author. The authors thank the numerous MIS execuuveswho sharedtheirknowledge,expenenceand perspecuvesvath them.
550
Gu:maraes, Ramanujam--Source$ of Personal Computing Problems REFERENCES
! Benson DH (1983) A field study ofend-user computing: findings and issues MIS Q 7(4), 35-46 2 Business Week (1984) How personal computers can trip us execuUves September 24, p 94 3. Engstrom T (1984) New kind of manager coordinates helter-skelter computer growth. Wall Street J August 24, p. 19 4 Freeman RM (1984) A shp of the chip on computer spread sheets can cost milhons Wall Street J August 20, p 10 5 Gaedeke RM and Tootehan DH (1976) The Fortune 500 hst" an endangered species for academic research. J Bus Res 4(3), 283-288 6 Green PE (1978) Analyzmg Multwartate Data The Dryden Press, Hmsdale, IL 7 Gmmaraes T (1984) The evolution of the reformation center Datamat:on, July 15, 127 8. Gmmaraes T and Ramanujam V (1986) Personal computing trends and problems" an empirical study M1S Q 10(2), 179-187 9 Gmmaraes T and Gupta Y (1987) Personal computing and support services Omega 15(6), 467-475 l0 Hackathorn RD and Keen PGW (1981) Orgamzatlonal strategies for personal computing in decision support systems MIS Q 5(3), 21-27
1!. Hlckson DJ, Hmmgs CR, Lee CA, Schneck RE and Pennlngs JM (1971) Strategic contangencaes theory of intraorganizauonal power Admm Sc:. Q 16(2), 216-229 12 Keen PGW and Scott Morton MS (1978) Dec~swn Support Systems. An Organ:zattonal Perspectwe Addison-Wesley, Reading, MA. 13 Keen PGW and Woodman LA (1984) What to do with all those micros. Harv Bus. Rev 62(5), SeptemberOctober, 142-150 14 Rockart JF and Flannery LS (1983) The management of end-user computing. Commun. ACM 26(10), 776-784. 15 Schoonhoven CB (1981) Problems with contingency theory testing assumpuons hidden vathin the language of contingency theory Admm Scz Q 26(3), 349-377. 16 Sharma S, Durand RM and Gur-Ane O (1981) Identtficatton and analysis of moderator variables J. Marketmg Res 18(3), 291-300 17 Sprague RH Jr (1980) A framework for the development of decision support systems MIS Q 4(4), 1-26 ADDRESS FOR CORRESPONDENCE:Professor V Ramanujam,
The Weatherhead School of Management, Enterprtse Hall, Dwts:on of Management Pohcy, Case Western Reserve Umverslty, Cleveland, OH 44106, USA