There are many definitions of benchmarking. For example, The
Benchmarking Network, an organization of benchmarking specialists, defines benchmarking as, a performance measrement tool sed in con!nction with impro"ement initiati"es to measre comparati"e operating performance and identify best practices.# This definition sggests that benchmarking in"ol"es the following se$ence of acti"ities% &'( )easre comparati"e operating performance &*( +dentify best practices &,( +nstitte impro"ement initiati"es -teps &'( and &*( sggest that regression can be a "alable tool for benchmarking, becase regression allows s to identify how practices &the . "ariables( affect performance &the / "ariable(. +n particlar, best practices are those that impro"e performance the most. The Benchmarking Regression 0 benchmarking regression shold look something like this% 1 2 3 4 5 3 ' . ' 5 3 * . * 5 B , . , 5 6 5 noise 7here 1 is some measre of performance, and . ' , . * , . , , etc. are "arios characteristics of the firm and its processes. The coefficients 3 ' , 3 * , 3 , , etc. indicate whether the . "ariables are positi"e or negati"e determinants of performance. 8et9s examine the benchmarking regression in more detail. Unit of Observation Benchmarking stdies can be performed across firms, work nits, or e"en indi"idals. 0n indstry analyst might examine in"entory costs across a sample of parts sppliers to the ato indstry &the nit of analysis is the firm(. 0 $ality control manager at +ntel might examine chip failre rates across chip fondries &the nit of analysis is the fondry(. 0 hman resorces manager at 0T5T cable might compare cstomer call times across cstomer ser"ice representati"es &the nit of analysis is the worker(. The Benchmarking Sample The se of regression has a "ery simple implication for sample selection% /o belie"e that the nderlying model predicting for the comparison sample is the same as for the firm in $estion. By implication, yo belie"e that the 39s are the same for all the firms, inclding the one in $estion 0n easy test to see if this holds% : /o start with a sample of firms that yo are $ite certain belong in the comparison grop : -ppose there is a grop of firms yo are considering adding to the sample to impro"e the precision of yor estimates. : ;etermine if the slopes of the key predictors are the same for this new grop as they were for the initial grop. /o do thisas follows% '( <reate a dmmy "ariable for the new grop of firms *( 0dd slope dmmy# interactions to the regression ,( Test the slope dmmies &bt not the dmmy intercept( for !oint significance. Performance Measures The list of possible performance measres is "irtally nlimited. 0t the firm level, one can examine o"erall profitability &measred as =>0, =>-, etc.( >ther firm le"el performance measres inclde "arios acconting ratios &e.g., administrati"e expense ratio( and worker prodcti"ity &otpt per nit of labor(. Product level measres of performance inclde nit costs or market shares. /o shold choose performance measres that are important and within the control of managers. There is no need to always examine profits. +f yo can impro"e performance at a particlar task, profits will follow. * Predictors 1redictors inclde "ariables that are within the control of managers &e.g. prodct offerings( and those that are not &e.g. local market conditions(. +n constrcting yor list of predictors, think abot trying to identify the Key Success actors. ' The regression coefficients indicate which of the chosen "ariables really do predict sccess. !mportant Takea"ay# Regression coefficients in a benchmarking study help identify the Key actors for Success -ppose that yor regression identifies se"eral key sccess factors. 7hat do yo do with this information? 1resmably, yo are stdying the performance of a particlar firm. @ere is how to se this information to help that firm. &'( <ompare the characteristics of that firm with the key sccess factors. @ow does it score? &*( +f it is lacking one or more key sccess factors, then perhaps it cold impro"e its performance by adopting them. &,( The regression coefficients are a good measre of the potential impro"ement. ' 1lease refer to The $conomics of Strategy for a definition and examples. , The Residual The residal in a regression is the difference between the actal "ale and the predicted "ale. +n a regression of firm performance, the residal is the difference between actal and expected performance. 7e can se this concept to help identify firms that are not meeting expectations. Begin by rnning a regression where the predictors are A-Fs that are trly exogenos to the manager B factors otside the manager9s control. These might inclde the choice of indstry, extent of competition, and local labor market conditions. 1rodct choices may also be considered otside the manager9s control, at least in the short rn. The difference between actal and predicted performance B the residal B represents the extent to which the firm is exceeding or lagging expectations, gi"en its en"ironment.# Now add A-Fs that are within the manager9s control. The coefficients on these "ariables will reflect how they contribte to performance at the a"erage firm. By examining how yor firm performs on these A-Fs, yo can assess the extent to which management is scceeding or failing on these areas that are nder its control. The %&eficiency Score' -ppose that yor hospital has costs that exceed the mean by C'444. +n other words / i B / m 2 '444. +t is possible to decompose this into component parts. =ecall that / i 2 B. i 5 D i 0lso note that / m 2 B. m 7e can write / i B / m 2 B. i 5 D i : B. m 2 B&. i :. m ( 5 D i +n other words, the difference between yor performance and the mean can be broken down into two components% E The difference between yor firm and the a"erage firm on the A-Fs, mltiplied by the corresponding B9s, and the error term. This sggests the creation of the following table% 1redictor . i . m . i :. m 3 3&. i :. m ( A-F ' A-F * Total : : : : 3&. i :. m ( This allows yo to eyeball the specific A-Fs that contribte to the relati"e performance deficiency of yor firm. Back to the Residual The "ale 3&. i :. m ( measres the extent to which firm performance exceeds or falls short of expectations based on each A-F. The residal, or / : 3&. i :. m (, measres the extent to which firm performance differs from expectations. * @ere is how to combine the deficiency score with consideration of the residal to gain a fller nderstanding of performance. '( -tart with a regression that contains only exogenos predictors. *( <ompte the ;eficiency -core for each predictor and the residal. These tell yo how factors otside of management9s control are affecting performance, and how the firm is doing, holding those factors constant. ,( =n a regression that incldes exogenos and endogenos predictors &i.e., those the manager can control.( E( =ecompte the deficiency scores and residal. /o now ha"e estimates of how factors nder the managers control affect performance, based on their a"erage effect at all firms. * +f yo take the log of the 8@- "ariable, yo will need to make the log ad!stment described in an earlier lectre to obtain the correct residal. F The residal in any regression captres a combination of pre noise pls omitted factors. +f a firm is nderperforming expectations &e.g., the residal in a profit regression is negati"e(, then this is either becase of random chance or omitted factors. >ne candidate for that omitted factor% managerial competence. +n other words, if a firm is nderperforming expectations, we might well consider whether management is to blame. +t trns ot that this interpretation of the residal is contro"ersial in some economic circles. This interpretation re$ires two assmptions% '( There is a distribtion of managerial competenceG i.e., managers can and do make mistakes. *( The empirical model predicting performance is free bias. 7hat if the firm has high costs becase it is located in a region where the labor pool lacks proper training? 0 regression that fails to control for training wold sffer from omitted "ariable bias. /o wold not want to chalk this p to incompetence. Key Success actors vs( !ncompetence Halable benchmarking regressions shold inclde a comprehensi"e list of predictors. 0s yo add predictors to yor benchmarking stdy, the = * will increase, and the dispersion of the residals will decrease. /o will be able to explain more and more of the difference in firm performance on the basis of the A-Fs. Bt as long as there is some nmeasred managerial competence, the = * will ne"er e$al '. &)oreo"er, if some A-Fs are endogenos factors nder the manager9s control, yo can interpret those as reflecting management competence.( >nce yo ha"e a robst model, yo can explain performance "ariation de to key sccess factors &exogenos and endogenos( and residal managerial competence. )onclusion# Relative performance is determined by key success factors and managerial competence( Regression gives you the tools to measure both( I