• Like
  • Save
Brand Communications Modeling: Developing and Using Econometric Models in Advertising. An Example of a Full Modeling Process.
Upcoming SlideShare
Loading in...5
×
 

Brand Communications Modeling: Developing and Using Econometric Models in Advertising. An Example of a Full Modeling Process.

on

  • 641 views

This report presents a description and a complete example of the modeling process required to build a comprehensive market response model that would account for the impact of previous marketing ...

This report presents a description and a complete example of the modeling process required to build a comprehensive market response model that would account for the impact of previous marketing actions on sales.

Statistics

Views

Total Views
641
Views on SlideShare
641
Embed Views
0

Actions

Likes
0
Downloads
22
Comments
0

0 Embeds 0

No embeds

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    Brand Communications Modeling: Developing and Using Econometric Models in Advertising. An Example of a Full Modeling Process. Brand Communications Modeling: Developing and Using Econometric Models in Advertising. An Example of a Full Modeling Process. Document Transcript

    • Copyright byEsteban Ribero 2005
    • Brand Communications Modeling:Developing and Using Econometric Models in Advertising. An Example of a Full Modeling Process By Esteban Ribero, B.A. Report Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of Master of Arts The University of Texas at Austin December, 2005
    • Brand Communications Modeling:Developing and Using Econometric Models in Advertising. An Example of a Full Modeling Process APPROVED BY SUPERVISING COMMITTEE: __________________________ John D. Leckenby __________________________ Gary B. Wilcox
    • Brand Communications Modeling: Developing and Using Econometric Models in Advertising. An Example of a Full Modeling Process Esteban Ribero, M.A. The University of Texas at Austin, 2005 SUPERVISOR: John D. Leckenby This report presents a description and a complete example of the modelingprocess required to build a comprehensive market response model that would account forthe impacts of previous marketing actions on sales in order to make better and moreinformed decisions that would help solve some advertising and marketing managementproblems. Real marketing and sales data of a big competitor in the skin-care market of aLatin American country was analyzed using multivariate regression analysis of time-series. The report presents a full description and an example of the four major stepsrequired to build a market response model: specification, estimation, verification andprediction. The model developed was used then to measure the ROI of the differentmarketing actions developed during the time period analyzed. A market sharedecomposition analysis as well as other analysis was provided in order to quantify thedirection and power of the impact of the market share drivers. The model was also usedto simulate two slightly different scenarios as an attempt to illustrate the “what-ifprocess” that can be done using a market response model suggesting different marketingand media strategies for the brand. iv
    • Table of ContentsList of tables…………………………………………………………………………….viiList of figures……………………...……………………………………………………viiiBrand Communications Modeling: Developing and Using Econometric Models inAdvertising. An Example of a Full Modeling Process……………………………………1 The Eras of Marketing Modeling………………………………………………….5The Modeling Process……………………………………………………………………..7 Specification………………………………………………………………………9 The modeler’s toolbox…………………………………………………...13 Current effects functional forms…………………………………13 Lagged advertising effects……………………………………….18 Modeling with adstock………………………………………...…23 Estimation………………………………………………………………………..24 Ordinary Least Squares…………………………………………………..25 Generalized Least Squares……………………………………………….30 Nonlinear Least Squares…………………………………………………32 Maximum Likelihood…………………………………………………....33 Verification………………………………………………………………………34 Prediction………………………………………………………………………...41 Model building Summary………………………………………………………..43An Example……………………………………………………………………………...45 Specifying the model…………………………………………………………….45 Estimating the model……………………………………………………………48 v
    • Verifying the model.……………………………………………………………52 Validating the model……………………………………………………………55Using the model………………………………………………………………………...60Summary………………………………………………………………………………..69References………………………………………………………………………………70Vita……………………………………………………………………………………..72 vi
    • List of TablesTable 1…………………………………………………………………………………...24Table 2…………………………………………………………………………………...39Table 3…………………………………………………………………………………...40Table 4…………………………………………………………………………………...49Table 5…………………………………………………………………………………...50Table 6…………………………………………………………………………………...51Table 7…………………………………………………………………………………...56Table 8…………………………………………………………………………………...58Table 9…………………………………………………………………………………...65 vii
    • List of FiguresFigure 1…………………………………………………………………………………....8Figure 2…………………………………………………………………………….….....11Figure 3…………………………………………………………………………….….....12Figure 4…………………………………………………………………………….….....12Figure 5…………………………………………………………………………….….....18Figure 6…………………………………………………………………………….….....27Figure 7…………………………………………………………………………….….....37Figure 8…………………………………………………………………………….….....54Figure 9……………………………………………………………………………...….. 57Figure 10………………………………………………………………………………....59Figure 11………………………………………………………………………………....61Figure 12………………………………………………………………………………....64Figure 13………………………………………………………………………………....68 viii
    • Brand Communications Modeling: Developing and Using Econometric Models in Advertising. An Example of a Full Modeling Process The way advertising is planned and executed is changing. The media landscapehas been changing at an impressive rate. The development of new technologies has madepossible the emergence of new and multiple media. The fragmentation of media channels,the decreasing audience’s size of traditional media and the empowerment of consumerscreate a new set of rules for marketing and advertising mangers who want to succeed inthe increasing competitive landscape. Within this framework to be accountable is no more a desire, it is a need. Thefamous statement attributed to John Wanamaker is more relevant now than ever: “I knowhalf of my advertising budget is wasted. The problem is I don’t know which half”.Finding which one is what we need now. And this is applicable not only to advertisingbut to all marketing activities. Being able to fully understand the effects of the differentmarketing policy instruments on sales should be a regular practice for marketing andadvertising mangers. Fortunately with today’s improvement in data collection and statistical analysis’techniques it is possible to address the problem in a scientific, yet subjective, manner. Aswe will se, the use of mathematical models to help marketers and advertisingprofessionals to solve management problems is not new. However, the recent use ofeconometric modeling in the advertising industry is becoming an important activity andmore and more companies are using the technique to improve their decision making 1
    • process. “Econometrics buzzes ad world as a way of measuring results” claimed a recentarticle in the Wall Street Journal (Patrick, 2005). The article mentioned the recent raiseon the number of employees working on econometric models in the advertising industry.For example, WPP’s MindShare has increase the number of people doing econometricmodeling from 20 to 150 in just 5 years. Omnicom’s OMD has its own business unit(OMD Metrics) dedicated to built econometric models for their international and localclients, and its staff members have increase from 6 to 45 in the past three years. Why is it so important to use formalized models in an industry that has beentraditionally reluctant to scientific scrutiny? Well, the game has changed: Theproliferation of options to promote the sales of a brand and the pressure for accountabilityis demanding more measurable results for the advertising industry. The pressure to comeup with ways to show which ads and media strategy boost sales of a product is thedriving force of this new interest in econometric modeling. There are many benefits of using formalized models to solve complex problemslike the ones one might encounter in marketing and advertising. John Sterman, an MITprofessor dedicated to the use of formalized model to improve our ability to comprehendand manage complex systems, discuses the advantages of using formalized models versusmental models. Following Sterman (1992), mental models have some advantages: theyare flexible, take a wide range of information into account, can be adapted to newsituations and are updated with new information. But mental models also have greatdisadvantages: they are not explicit, not easily examined by others. Their assumptions arehard to discuss, even for our own mental models. But the most important problem withmental models is that our rationality is bounded: The best-intentioned mental analysis of 2
    • a complex problem cannot hope to account accurately for the effects of all theinteractions between the variables, especially if those interactions are nonlinear. In the other hand, formal models’ assumptions can be discussed openly. Formalmodels are able to relate many factors simultaneously and can be simulated undercontrolled conditions, allowing analysts to conduct experiments which are not feasible inthe real world. This does not mean that formal models are correct. All models are wrong(Sterman, 2002): they represent the reality, they are not the reality. But formalizedmodels can help us to understand the systems we work in and for. Advertising and marketing managers can greatly be benefited by using models tosolve important problems. For example, the use of econometric models can help amanger to find the optimal or near optimal advertising budget for future periods. Theanalysis would allow him or her to find the adequate advertising budget for attaining aspecific sales goal or, if financial information is available, the model can incorporateshort-term and long-term criteria to maximize profit. (To see some examples, visit thefollowing http addresses:http://www.ciadvertising.org/sa/spring_05/adv391k/eribero/frameset.htmhttp://www.ciadvertising.org/sa/spring_05/adv391k/eribero/Solo2/frameset.htm ). Other applications of the modeling process could help managers to answer thefollowing questions: • What is the optimal mix of TV vs. Posters vs. Radio? • What happens to sales when we obtain a wider distribution? • What happens to sales when we do not advertise? • How much should we spend on advertising vs. promotion? 3
    • • What is the best pattern and level of advertising for my brand? • How effective is our pricing strategy? • Which competitors hurt my brand and how? • Which of my communications channels offers best value for money? • How does advertising work and how can we prove this to the Financial Director? • How do I spend the same budget but increase sales? • What’s the impact of economic changes on my brand? • What’s the best pattern and level of advertising for my brand? • Which copy strategy/campaign worked better? • How much sales could we make next period with X budget? Besides these direct practical applications for budgeting, forecasting andaccountability the modelling process would improve the manager’s ability to cope withhis complex environment. Leeflang, Wittink, Wedel & Naert (2000, p. 25-27) lists 8possible indirect benefits of using models in business. The benefits are described asfollows: 1. “A model would force him [a manger] to explicate how the market works. This explication alone will often lead to an improved understanding of the role of advertising and how advertising effectiveness might depend on a variety of other marketing and environmental conditions.” 2. “Models may work as problem-finding instruments. That is, problems may emerge after a model has been developed. Managers may identify problems by discovering differences between their perception of the environment and a model of that environment.” 3. “Models can be instrumental in improving the process by which decision-makers deal with existing information” 4
    • 4. “Models can help managers decide what information should be collected. Thus models may lead to improved data collection, and their use may avoid the collection and storage of large amounts of data without apparent purpose. 5. “Models can also guide research by identifying areas in which information is lacking, and by pointing out the kinds of experiments that can provide useful information.” 6. “[A] model helps the manager to detect a possible problem more quickly, by giving him an early signal that something outside the model has happened”. 7. “Models provide a framework for discussion. If a relevant performance measure (such as market share) is decreasing, the model user may be able to defend himself to point to the effects of changes in the environment that are beyond his control, such as new product introductions by the competition. Of course, a top manager may also employ a model to identify poor decisions by lower-level managers.” 8. “Finally, a model may result in a beneficial reallocation of management time, which means less time spent on programmable, structured, or routine and recurring activities, and more time on less structured ones.” The Eras of Marketing Modeling As Leckenby and Wedding said (1982), “the concept of model building inadvertising can be traced back only as far as the early 1950’s”. Even though it is arelative short history, Leeflang et al (2000) identified five eras of model building inmarketing. The first era is characterized by the emulation or transposition of OperationalResearch and Management Science into the marketing framework. The OS/MS tools thatincluded mathematical programming, computer simulations, game theory, and dynamicmodeling were initially developed to solve some of the strategic problems faced duringWorld War II. The emphasis was on quantitative method sophistication rather than on themarketing problem per se (Leckenby & Wedding, 1982). The advertising and marketing 5
    • problem was adjusted to fit the requirements of the technical methods available, ratherthan the other way around. The methods were typically not realistic, and the use of thosemethods in marketing applications was therefore very limited (Leeflang et al, 2000). The second era which ended in the late sixties early seventies was characterizedby the attempt to adapt the models to fit the marketing problems in order to overcome themisuse of the OR approach in the advertising and marketing field. The models werehowever so complex that lacked usability. The third era that started around 1970, showed and increased emphasis on modelsthat were good representations of reality and at the same time easier to use. John D.C.Little developed the concept of “Decision Calculus”. He used the term to describe modelsthat would process judgments and data in a manner which would assist the manager indecision making (Leckenby and Wedding, 1982). This emphasis in helping decisionmaking made a major change in the direction of model building in advertising. Little(1970) suggested possible answers to the question of why models were not used: goodmodels and parameterization is hard to find; managers do not understand models; andmodels are incomplete. So in order to overcome such problems a model should be:simple; robust; easy to control; adaptive; complete on important issues; and easy tocommunicate with. He also said that a model should be evolutionary (Little, 1975)meaning that a model should start with a simple structure, to which detail is added latter.The use of judgmental data as well as objective data in the model building process helpedthe raise of models implementation (Leeflang et al, 2000). Even though the third era of modeling in marketing and advertising was focusedon implementation and usability of models it was not really until the fourth era (starting 6
    • in the mid 1980) when models were actually implemented (Leeflang et al, 2000). Themain factor that helped this implementation boom was the availability of precisemarketing data coming from scanning equipment that captured in-store and household-level purchases. This era coincided with the proliferation of marketing support systems. The fifth era may be characterized by an increase in routinized modelapplications. It is predicted that in the coming decades the age of marketing decisionsupport will usher in an era of marketing decision automation (Leeflang et al, 2000;Bucklin et al, 1998). It is expected that marketing support systems take care of routinemarketing decisions like assortment decisions and shelf space allocation, customizedproduct offerings, coupon targeting, loyalty reward and frequent shopper club programs,etc. The focus of this paper is in the model building process representative of the thirdand fourth era. The Modeling ProcessThe model building process for any mathematical model, including response models, issupposed to follow a sequence of steps. The traditional view assumes the following foursteps: specification, estimation, verification and prediction (Leckenby & Wedding, 1982;Leeflang et al, 2000). Leeflang et al (2000) propose an alternative sequence more focusedon implementation (see figure 1; for a detailed explanation of the implementation viewsee Leeflang et al, 2000, chapter 5). In order to keep it as simple as possible we arefocusing on the traditional view. 7
    • Figure 1. The implementation view on model building. (From Leeflang et al,2000, p. 52) 8
    • Specification “A model is a representation of the most important elements of a perceivedrealworld system.” (Leeflang et al, 2000) In order to better understand the model building process and especially thespecification stage is important to analyze the definition provided above. The definitionindicates that models are representations, “simplified pictures” (Leeflang et al, 2000) ofreality. Those representations may be useful for decision makers trying to understand thereality they deal with. The definition above has an extremely important implication: sincea model is a representation of a perceived realworld it is something subjective. Differentmodel builder could have different perceptions and interpretations about the same reality.Modelers could also have different opinions about which are “the most importantelements” to represent. This makes the model building process not only more interestingbut very dependent on the modeler’s “theory” of the reality he tries to represent. That is why it is so important in the model building process to fully specify thevariables and the relationship between them. That is exactly what is done in thespecification stage. For example, if we consider sales as the dependent variable and advertising andthe rest of the marketing policy instruments as the independent variables, specificationwould be the process of deciding upon the functional form which will describe therelationship between advertising (and the other marketing variables) and sales (Leckenby& Wedding, 1982). In other words: “specification is the process by which the manager’stheory of how advertising works for a particular brand or company is put into testableform” (Leckenby & Wedding, 1982, p. 257). 9
    • Rephrasing Little’s suggestions for building good models (Little 1970), a modelshould be: a. simple; b. complete on important issues; c. adaptive; d. robust.Leeflang et al, (2000) pointed that it is easy to see that some of these criteria are inconflict. They state that “none of the criteria should be pushed to the limit. Instead, wecan say that the more each individual criterion is satisfied, the higher the likelihood ofmodel acceptance” (Leeflang et al, 2000, p. 53) While specifying a model one should then consider these elements. As a goal,models should be as simple as possible. That is, considering the principle of parsimony,one should choose between competing models the one that fairly represent the realitywith the simplest structure. Equally important is to consider the trade-off betweenaccuracy and usability. It is not uncommon to find two competing models that performdifferently in these two criteria. If accurate forecasting is more important thanunderstanding the effects of the independent variables then a more accurate model shouldbe chosen even though it might be more complex and then less easy to explain and use.But if it is more important to understand the market dynamics and the way the marketingvariables affect sales a simpler model should be used. Fortunately for modelers they are several functional forms to choose from whilespecifying a model. The one to be selected depends on the above criteria as well as on the 10
    • underlying theory of marketing and advertising that the manager or modeler isconsidering. We first will consider the different shapes that a response function might have.Then we will describe some of the most used response functions in advertising. The shapes of a response function could be classified as linear, concave or s-shape. Any other shape could be the result of a combination of one or more of theseshapes. Figure 2 shows a typical linear response. Figure 3 shows different concaveresponse shapes and figure 4 shows some s-shape functions. Q A Figure 2. A linear shape function 11
    • Figure 3. Some concave response functionsFigure 4. Some s-shape response functions 12
    • The modeler’s toolbox “To the craftsman with a hammer, the entire world looks like a nail, but theavailability of a screwdriver introduces a host of opportunities!” Lilien & Rangaswamy (1998) Because it is true that one should not modify to problems to fit the tools it iseasier for the modeler if he/she can choose from a series of predetermined functions thathe/she can then modify to fit the problem. The decision to pick one or the other dependson the problem at hand and the data availability. For example, a linear function (thesimplest possible response function) could fit the data pretty well if the data rangecorrespond to a linear section of a more complex response function. (Lilien &Rangaswamy, 1998) The following are some of the most used response functions in advertising. Eventhough a brief description of the functions is provided, for more details please refer toHanssens, Parsons & Schultz, 2001; Leeflang et al, 2000; or Kotler, 1971. Current effects functional forms. The simplest response functions, Current EffectsFunctions (CE), assume that the effects of the marketing variables occur in full in thesame period in which they appear. For example, advertising expenditures in April aresupposed to affect sales in April and only April. While this might not hold true for mostof the brands CE functions are useful for their simplicity and ease to explain. The Linear response model has the following form: S = a + bA + u Where: S = Sales 13
    • a = the y intercept b = slope of the function A = Advertising expenditures u = disturbance term or error term The linear response function assumes constant returns to scale. That is, salesincrease by a constant amount to equivalent constant increase in marketing effort (Figure2). The linear model would not lead to locally different conclusions than another functionif the data are available only over a limited range. While adequate for asking “what if”questions around the current operating range, the linear model would be misleading ifdata outside the range are used like it would be the case in trying to find the optimaladvertising effort. More realistic response models are said to have diminishing returns to scale.These models suppose that sales always increase with increases in advertising ormarketing effort, but each additional unit of marketing effort brings less in incrementalsales than the previous unit did (Hanssens et al, 2001). The following concave downwardresponse functions show diminishing returns to scale: The Semilogarithmic (Log) function: S = a + b ln A + u The Square-root function: S = a+b A +u The Quadratic function: S = a + b1 A − b2 A 2 + u 14
    • The quadratic function has the important property that differentiates it from theothers which is that it can represent the concept of supersaturation; phenomenon thatoccurs when too much marketing effort causes a negative response. The so called“wearout” effect is an example of a case of supersaturation in advertising. The following functions are nonlinear in the variables but linear in parameters andcan be linearizables with some algebra in order to be able to estimate them through linearregression (see the section Estimation in this paper): The Power function: a) S = aA b b) ln S = ln a + ln A The power function is very flexible since depending on the value of the parameterb it can take very different forms (see Leeflang et al, 2000 p. 75-76; Kotler, 1971 p. 33) Italso has the great characteristic that the coefficient b is actually the elasticity of thedemand to advertising (Hanssens, 2001, p. 101, Broadbent, 1997). Also, when more thanone independent variable are considered the power function, also known as themultiplicative function, accounts for possible interactions between the independentvariables. The Modified Exponential function: a) S = S (1 − e a +bA ) ⎡ S⎤ b) ln ⎢1 − ⎥ = a + bA ⎣ S⎦ Where: S = upper bound level or saturation point 15
    • e = a mathematical constant equals to 2.71...16 … An attractive characteristic of the modified exponential function and some of thenext functions as well, is that it supposes an upper limit or saturation point where themarket potential reaches its maximum. One special characteristic is that it implies that themarginal sales response will be proportional to the level of untapped potential (Kotler,1971). All previous functional forms except the linear one are concave downwardfunctions (figure 3). That implies diminishing returns at all points in the response. It issometimes the desire of the modeler or manager to represent the intuitive concept of a“threshold effect” in advertising. That is, the idea that small doses of advertising does notcount for much and that there is a tipping point that must be crossed in order to expectreal effects of advertising on sales. Even though there is little evidence that such aphenomenon occurs in advertising (Kotler, 1971; Leckenby & Wedding, 1982; Hanssens2001) it is possible to represent the concept using s-shape functions (figure 4). Thesefunctions assume increasing marginal returns at first and then diminishing marginalreturns with respect to various alternative levels of advertising. The following are themost common s-shape functions: The Gompertz function a) S = Se − e e a bA b) ln(ln S − ln S ) = a + bA The Logistic function: S a) S = (1 + e − ( a + bA) ) 16
    • ⎡ S ⎤ b) ln ⎢ = a + bA ⎣S − S ⎥ ⎦ The Lower-Bound Logistic function: S + S LB e a + bA a) S = 1 + e a + bA ⎡ S −S ⎤ b) ln ⎢ ⎥ = a + bA ⎣ S − S LB ⎦ Where: S LB = Lower bound level or minimum sales when advertising is 0. As described above, these functions are just approximations of different“realities” and the modeler can modify them to incorporate other elements to betteraddress the problem at hand. For example, these functions only consider one independentvariable and do not account for special situations like seasonality or special events duringthe period analyzed. The modeler can then add different variables to these functions oruse dummy variables to represent qualitative differences or changes in the data (see someexamples at Hanssens et al, 2001, p. 97-99). Figure 5 shows some of the functionsdiscussed above. 17
    • Figure 5. Graphical representation of some CE functional forms (from Leckenby &Wedding, 1982). Lagged advertising effects. As discussed earlier, Current Effects responsefunctions assume that the effects of an advertising or marketing expenditure in period toccurs only, and completely, in period t. This assumption does not correspond withcommon understanding of advertising theory since it is assumed that a big part ofadvertising effects occur with time. So, in order to accommodate this into advertisingresponse models we need first to discuss some basic concepts about carryover effects. 18
    • Carryover effect is the term used to describe the idea that marketing andadvertising expenditures have effects on sales that carries over into future periods(Kotler, 1971). There are two major categories of carryover effects that can bedistinguished: the delayed response effect and the customer holdover effect (Leckenby &Wedding, 1982). The delayed response effect develops because delays occur between the time theadvertising dollars and programs are implemented and the time the advertising generatedpurchases occur (Leckenby & Wedding, 1982). There are four types of delayed responseeffects: Execution delay, noting delay, purchase delay and recording delay. The delayoccurs either because executing takes time, consumers do not notice the ads immediatelyor because they delay the purchase to future periods. The recording delay is a problemwith the data and may not represent a real delayed response, just a mismatch between thedata (for more detail see Kotler, 1971: Leckenby & Wedding, 1982) The customer holdover effect is clearly explained by Kotler (1971): “suppose thata marketing stimulus is paid for today, appears today, is noted today, and leads topurchase today. No delayed response is involved. The buyer finds the product agreeableand decides to remain with this brand. On this basis it can be said that marketing stimulusthis period affected sales this period and for many future periods.” (p. 124) This repurchase scenario suggests that advertising should be credited, in somepart, for holding the costumer to the brand in future time periods. Retaining new andpossibly old customers in future periods is not the only way a holdover effect can occur.A holdover effect can also occur even if the number of customer does not increase as aresult of the advertising expenditure. This can happen when the advertising or other 19
    • marketing stimulus increases the average quantity purchased per period per customer(Kotler, 1971). Regardless the type of carryover that could be present for a brand at a particulartime, it is possible to represent it with some dynamic models. To better understand someof these models we will consider the simplest linear model with lagged effects. Themodel has the following form: S t = a + bAt + bcAt −1 + bc 2 At −2 + .... Where: a = the intercept term b = regression coefficient c = carryover rate or retention rate (0 < c <1) The basic assumption behind this model is that the effect of advertising in period tdecays exponentially in subsequent periods. That is, the effect on sales in period t is theresult of the advertising in period t plus a fraction of advertising in t-1 plus a fraction ofadvertising in t-2, etc. The rate of decay, or in other words, the amount of advertisingeffect that is carried over the immediate next period is the carryover rate (c). Because estimating the parameters on this models requires us to know how manyperiods we have to look back as well as dealing with autocorrelations (see Estimation inthis paper) some modifications done by Koyck and others give us the following laggedeffect models: The Koyck Geometric Distributed Lag (GL) model: S t = a(1 − c) + bAt + cS t −1 + {ut − cut −1} Where: 20
    • u t = white noise (disturbance term) c = carryover rate or retention rate (0 < c <1) b = β (1 − c) Short-term effect of advertising b β= Long-term effect of advertising 1− c This model hypothesizes that the effect of advertising conducted in all precedingtime periods on current sales period t can be summarized in one term: lagged sales. Salesare then assumed to be a function of advertising and sales in the preceding time period.The model performs well sometimes, however where strong sales trends are noted, theeffect of previous time period sales on current sales is so strong that the effect of currentadvertising on sales can hardly be detected (Leckenby & Wedding, 1982), something nottotally in accordance with advertising theory. The Partial Adjustment (PA) model: S t = (1 − ϕ )[a + bAt ] + ϕS t −1 + wt Where: 1 − ϕ = adjustment rate w = white noise The Partial Adjustment model is similar to the Geometric Lag in its structure. Itassumes that consumers can only partially adjust to advertising stimulus in the short-termbut they will gradually adjust to the desired consumption level, which causes theadvertising effect to be distributed over time (Hanssens et al, 2001). Note: The above Partial Adjustment model should not be confused with theNerlove Partial Adjustment model (Nerlove PA). The latter may not be a carryover effect 21
    • model but it represents the concept of brand loyalty and assumes some inertia from thepast. This model could be tried after some unsuccessful attempts with the Current Effectsmodels and before the more complex models of carryover effects. The Nerlove PA functional form is: S t = a + b1 A1 + b2 S t −1 + ut Another carryover effects model similar to GL but with an autoregressivestructure is the following: The Geometric Lag Autoregressive (GLA) model: S t = a + b1 A1 − b1 ρAt −1 + (c + ρ ) S t −1 − cρS t − 2 + {u t − cut −1} Where: c = carryover rate or retention rate (0 < c <1) ρ = autocorrelation coefficient The GLA model is a nested model which means that lower-order equations arecontained within the parameters of its higher-order structure (Hanssens et al, 2001). Forexample, where ρ = 0 the GLA becomes GL; where ρ = 0 & c=0 the CE linear modeland the special case where ρ = c (≡ ϕ ) the Partial Adjustment model (Hanssens et al,2001; Leeflang et al, 2000). A modeler should first try some of the CE models, then if after estimating theparameters (see Estimation in this paper), autocorrelation appears he should try i) to addimportant explanatory (independent) variables or ii) to change model specificationsthrough transformations. If after i) and ii), autocorrelation (the fact that a variable iscorrelated with itself in previous time periods) remains it may be “true” autocorrelation.That is, a generalized carryover effect so the modeler should specify this autocorrelation 22
    • in the model (Leckenby, personal notes). The Geometric Lag Autoregressive model(GLA) is an example of that process (For others autoregressive models see Hanssens,2001, cap. 4). It is important to know that these lagged effects response models can also takedifferent functional forms in order to represent diminishing returns to scale or s-shapebehavior; pretty much like the Current Effects models discussed earlier. Modeling with adstock. The concept of carryover effect can be modeled eitherexplicitly, as we have seen in the previews models or implicitly using stock variables.The latter approach was championed by Simon Broadbent in several publications (seeBroadbent, 1979, 1984, 1997). The basic idea with the creation of stock variables is thatthey capture the present and past amount of advertising effect for any period into onesingle value for that specific period. The approach assumes the same geometrical declinein advertising effect as the models presented above. The adstock variable is then justadded to the equation like any other independent or explanatory variable. Its key advantage is the ease of communicating results to management and itssimpler estimation process since the retention rate can be estimated subjectively using theconcept of half-life (HL). Half-life is simple the time it takes for an advertising effort tohave half of its effects. Event thought this time can vary from 3 to 10 weeks it tends to bebetween 4 to 6 weeks (Broadbent, 1984). There is a carryover rate or retention rate (c) associated with every HL value.Table 1 show the retention rate for different half-lives for “first period counts full”convention or “first period counts half” (see Broadbent, 1984; Hanssens et al, 2001 for adiscussion on these conventions). To the extend that the adstock approach uses the same 23
    • model of carryover the work is not different than the one resulting from the models thatspecify the carryover effect explicitly (Hanssens et al, 2001).Table 1.Half-life and retention rate. Half Life 1 2 3 4 5 6 7 8 f=1 0.500 0.707 0.794 0.841 0.871 0.891 0.906 0.917 f = 1/2 0.334 0.640 0.761 0.821 0.858 0.882 0.899 0.912 Half Life 9 10 11 12 13 14 15 16 f=1 0.926 0.933 0.939 0.944 0.948 0.952 0.955 0.958 f = 1/2 0.922 0.930 0.936 0.942 0.948 0.950 0.953 0.956 Estimation Once the modeler has specify a model based on theoretical relations between theexplanatory and dependent variables or by examination of the available data he or shemust estimate the parameters of the function using historical or cross sectional data(Leckenby & Wedding, 1982). The essence of the process is fitting a determined equationto a set of data in order to find the best estimates of the different parameters in the model( a, b1, b2 , c , etc). There are many estimation techniques however the most “robust” andpopular is regression analysis. We will now describe the basic concepts of the simplest regression analysis:Ordinary Least Squares (OLS). We will discuss the assumptions underlying thistechnique and the problems when they are violated as well as possible remedies. 24
    • It is important to notice that the process of model building is somehow circular inthe sense that a model is specified, estimated, and verified but very often some violationsof the assumptions as well as unsatisfactory results force the modeler to choose adifferent estimation technique or to modify the model specification and start the processagain. Another annotation is that the estimation process in model building is more of aconfirmatory approach (see Hair, 1998) of multiple regression analysis. It differssomehow with an exploratory approach because a pre-established functional form basedon theoretical relations between variables is “tested” or confirmed against empirical data.However, as noted earlier, it is an iterative process where different fictional forms mightbe “confirmed” until finding satisfactory results.Ordinary Least Squares The basic idea of estimating the parameters of a response function is to find thevalues for each parameter that would minimize the sum of errors or disturbance terms inthe equation. Let us consider the simplest linear functional form: S = a + bA + u Where: S = Sales a = the intercept term b = slope of the function A = Advertising expenditures u = disturbance term or error term 25
    • Rephrasing, the objective in the estimation process of model building is to findthe values of a and b that would give the least value of u in the average. Because what weare trying to find is the statistical relationship between the variables there is always somerandom errors: for every value of an independent variable there might be more than onevalue of the dependent variable. These multiple values of the dependent variable forevery value of the explanatory variables are the result of random components in therelationship (Hair, 1998). The Ordinary Least Squares is the basic technique in which the parameters of alinear or linearized (see Specification section in this paper) response function areestimated by minimizing the sum of the error terms at every point of the function.Because the difference between a predicted value by the function and the observed valuecould be positive or negative, the error terms are squared so they can be added to producea measure of the fit of the model to the data in the sample. That measure is the residualsum of squares (RSS) or the sum of squared errors (SSE) (Hair, 1998). There is also ameasure of the improvement in explanation of the dependent variable attributable to theindependent variables compared to just using the media of the dependent variable. It iscalled the sum of squared regression (SSR) and it is calculated by adding the squareddifferences between the mean and the predicted value of the dependent variable for allobservations (Hair, 1998). These tow measures are crucial for assessing the model’scapacity to explain the variation of the data of the dependent variable. If the SSR isdivided by the total sum of squares (TSS), the total variance of the dependent variable,we obtain the coefficient of determination R 2 that represents the portion of the totalvariance of the dependent variable (usually sales S or market share) explained by the 26
    • model. Figure 6 shows a graphical representation of those measures. The unexplainedvariance is SSE, the explained variance is SSR and the total variance is TSS. Figure 6. Variance in regression analysis (from Leckenby & Wedding, 1982). The procedure underlying OLS has several restrictive assumptions that must becarefully considered in assessing the validity of the estimated model (see Verification inthese paper). The fundamental assumptions are the following: a.) The mean of the error terms equals 0 b.) Constant variance of the error terms c.) Independence of the error terms d.) Normality of the error terms’ distribution e.) Low multicollinearity 27
    • The basic idea behind these assumptions is that u is a random variable. This is clearlyexplained by Koutsoyiannis in his Theory of Econometrics book (1978): “(…) u can assume various values in a chance way. For each value of an independent variable the term u may assume positive, negative or zero values each with a certain probability. We said that u is introduced into the model in order to take into account the influence of various errors, such as errors of omitted variables, errors of the mathematical form of the model, errors of measurement of the dependent variable, and the effects of the erratic element which is inherent in human behavior. Now, for u to be random the omitted variables should be numerous, each one individually unimportant, and they should change in different directions so that their overall effect on the dependent variable is unpredictable in any particular period.” If we agree that what we are trying to represent in model building is therelationship between the independent and depend variables in the average, it is imperativethat the mean of the error term equals 0 (assumption a). Otherwise the parameters of thefunction are biased (Leeflang et al, 2000). Assumption b means that the dispersion of the error terms remains the same overall observations of the independent variables. It is said that the variance of the error termsaround the zero mean is homoscedastic, which means that it does not depend on thevalues of the independent variables. Conversely, the case of heteroscedasticity is whenincreasing or decreasing dispersion of the error terms is observed. The consequence ofviolating this assumption is that it is not possible to calculate an effective confidenceinterval for the parameters reducing their efficacy (Leeflang et al, 2000) and theirstatistical significance (Koutsoyiannis, 1978). Assumption c is also known as absence of autocorrelation. That means that theerror terms at any point in the function should be independent from each other. This 28
    • might be relevant only when the model is estimated using time series because theautocorrelation is actually a serial correlation (Leeflang et al, 2000) between the error atone period and the error(s) at the previous period(s). There is positive autocorrelation andnegative autocorrelation. Positive autocorrelation means that the residual in t tends tohave the same sign as the residual in t-1. Negative autocorrelation is when a positive signtends to be followed by a negative sign or vice versa (Leeflang et al, 2000). Theconsequences of violating this assumption is that even though the estimated parametersare unbiased (as when assumption b is violated) the OLS formula underestimates theirsampling variance and the model will seem to fit the data better than it actually does(Hanssens et al, 2001). The assumption of normality (assumption d) is necessary for conducting thestatistical tests of significance of the parameter estimates and for constructing confidenceintervals. If this assumption is violated the estimates are still unbiased and best, but it isnot possible to assess their statistical reliability by the classical test of significance (t, F,etc.) because this test is based on normal distributions. Multicollinearity results form the correlation between independent variables.When one independent variable “moves” at the same time as another one it is said thatthey are collinear. In marketing as in many other areas variables tend to be correlated allthe time. For example, a price reduction is announced via some TV advertising as well asradio. These variables will be correlated to each other since they vary at the same time.Managers usually do not leave all variables constant and vary only one at the same time.The degree of multicollinearity has an important impact on the parameters of theresponse function. A high level of multicollinearity limits the size of the coefficient of 29
    • determination R 2 and it makes determining the contribution of each independent variabledifficult because the effects of the independent variables are “mixed” or confounded(Hair, 1998). In consequence, the reliability of the parameter estimates is low (Leeflanget al, 2000). The assumptions discussed above limit the applicability of OLS to estimate theparameters of the function because these assumptions are often violated. There are manyreasons why the assumptions are violated but usually it is the result of misspecification ofthe response function. There are some tests and procedures to test if one or more of theassumptions are violated. Some of them would be described in the Verification section ofthis paper. Once the parameters are estimated and the underlying assumptions tested it issometimes possible to take some corrective actions if violations to the assumptions arepresent. The simplest corrective action is modifying the specification of the responsefunction and estimating it again. However, sometimes the only solution is to use adifferent estimation technique.Generalized Least Squares In the Generalized Least Squares (GLS) techniques some of the restrictiveassumptions about the disturbance term in OLS are relaxed, specifically the assumptionsof constant variance and independence of the error terms (autocorrelation). Theseestimation methods are “generalized” because they can account for especial cases ormodels. Actually, OLS is a special case of GLS where all the assumptions are met(Leeflang et al, 2000). Other special case is when the variance is heteroscedastic --forexample, when cases that are high on some attribute show more variability than cases that 30
    • are low on that attribute, and the difference can be predicted from another variable, aweight estimation procedure can compute the coefficients or parameters of a linear modelusing weighted least squares (WLS), such that the more precise observations (that is,those with less variability) are given greater weight in determining the regressioncoefficients (Leeflang, 2000). The weight estimation procedure in statistical packageslike SPSS tests a range of weight transformations and indicates which will give the bestfit to the data. Another special case, typical of time-series, is when there is strong presence ofautocorrelation of the disturbance terms but at the same time the variance ishomoscedastic. Assuming that the autocorrelation is generated by a first-orderautoregressive scheme (Markov scheme) some transformations are done to incorporate anautoregressive coefficient that would allow better parameter estimates (see Leeflang et al,2000, p. 371-376 for a detailed explanation). There are others GLS methods that accountfor especial cases of the behavior of the disturbance term. For an extensive list ofliterature on those methods see Hanssens et al, 2001, Chapter 5. One important note is that these GLS procedures for dealing with special patternsof the disturbance terms would not give better parameter estimates if the pattern is due tomisspecified models, as it is usually the case (Leeflang et al, 2000). Additionally,“robustness may generally be lost if GLS estimation method are used” (Leeflang et al,2000, p. 376). So, before using these procedures the modeler should be convinced that heor she is using the best possible model specification (Leeflang et al, 2000). 31
    • Nonlinear Least Squares There are some models that are nonlinear and nonlinearizables. Additionally,there are other models that violate the assumptions of the disturbance term in their merespecification. Those models include the Koyck General Lag (GL), Partial Adjustment(PA) and General Lag Autoregressive (GLA). Those cannot be accurately estimated bylinear regression. For solving this problem some procedures have been created to allowthe modeler to estimate those kinds of models. The general or more commoncharacteristic of this procedure is that it is iterative. In its simplest form the parameterthat is causing the model to be nonlinear is guessed by either subjective estimation or trialand error until a satisfactory result is achieved. Leeflang et al (2000) explain this gridsearch in the following terms: “For simplicity assume that for any value of y [theparameter causing the nonlinear attribute], the model is estimated by OLS, under theusual assumptions about the disturbance term. Then choose m values for y, covering aplausible wide range, and choose the value of y for which the model’s R 2 is maximized”(p. 384). This procedure is equivalent to the one using adstock models when differenthalf-life values are tested to select the one that gives the best results (Broadbent, 1984). This grid search can also be done when instead of replacing a parameter that iscausing the nonlinearity, different transformations of the predictor variables are testedsequentially until finding satisfactory results (Leeflang et al, 2000). However, grid searchprocedures are costly and inefficient, especially if a model is nonlinear in several of itsparameters (Leeflang et al, 2000). 32
    • More sophisticated methods have been developed where initial estimates of someparameters are reintroduced in the equation in an iterative process until the whole processconverges (Leeflang et al, 2000; Koutsoyiannis, 1978). All the techniques discussed above estimate the parameters in an attempt tominimize the squares of the differences between the estimated points and the observedones. They are all Least Squares (LS) methods. A radically different approach is theMaximum Likelihood (ML) method.Maximum Likelihood The ML method is based on distributional assumptions about the data. Basically itfinds the values of parameters that make the probability of obtaining the observed sampleoutcome as highly as possible (Hanssens et al, 2001). In other words “the maximumlikelihood principle is an estimation principle that finds an estimate for one or moreparameters such that is maximizes the likelihood of observing the data. The likelihood ofa model (L) can be interpreted as the probability of the observed data y, given the model”(Leeflang et al, 2000, p. 390). Under this assumption a certain parameter is more likelythan other. The assumptions underlying ML method are actually the ones involved inhypothesis testing in social sciences (Leeflang et al, 2000) and not surprisingly themethod is very sensible to the sample size, giving better results with large samples. The ML method can also be used to select a model between competing ones (seeSummary in this paper). For more details on ML and LS methods for estimating theparameters of a response function consult Hanssens et al, 2001; and Leeflang et al, 2000. 33
    • Verification Another important step in developing market response models is to verify that theparameters estimated in the previous step truly represent the relationship between sales(or any other dependent variable) and the marketing variables. The usual way to do this isto use statistical significant testing (Leckenby & Wedding, 1982). By verifying theparameters it is possible to determine with a certain risk level how representative they areof the true advertising-marketing/sales relationship. In market response model (ifcommercially used) the significance level often used is about 15 percent (Leckenby &Wedding, 1982). If achieving that level of significance one could say that in at least 85samples of every 100 samples of data that we use for estimating the response function theparameters would be between x and y number (the confidence interval). The first measures that should be verified are those related with the fit of themodel to the data in the sample. As discussed above, the R 2 value indicates thepercentage of the variance of the dependent variable explained by the independentvariables in model. Because this measure is affected by the number of observations perindependent variables used, the modeler should focus on the adjusted R 2 for comparingbetween competing models and to control for “overfitting” the data (Hair, 1998). It isimportant to notice that the minimum ratio of observation per independent variableshould be 5 to 1 in order to avoid making the results too specific to the sample(“overfitting”) thus lacking generalizability. Verifying the statistical significance of R 2and adjusted R 2 is critical in this step. The F ratio is the statistical significance test thatmost statistical packages use to test this. The parameters of the models should also betested in terms of their statistical significance. The t value of a coefficient or parameter is 34
    • the coefficient divided by the standard error. To determine if the parameter issignificantly different form zero (no effects or relation with the dependent variable) thecomputed t value is compared to the table value for the sample size and confidence levelselected. This test is not that important for the intercept term in a linear model since itacts only to position the model (for details see Hair, 1998, p. 184) Another measure highly related with the overall fit of the model that must be alsochecked in this step of model building is the RSS or SSE (the squared sum of the errorsor disturbance terms). Even though a high R 2 could be found for a specific model theRSS could still be very large indicating the inability of the model to accurately makepredictions. As discussed in the previous section, the assumptions underlying the differentestimation techniques are highly important for assessing the validity of the parameterestimates since violations of the assumptions give biased coefficients or, more frequently,make their statistical significance hard to estimate (Leeflang et al, 2000). If theassumptions are violated the confidence that the parameters truly represent therelationship under analysis is diminished. So another important task of the verificationstep is to verify that the assumptions used for estimating the parameters are not violated.The simplest way to do this is by a careful analysis of the residuals using scatter plots. Itis recommended to use some form of standardization as it makes the residuals directlycomparable. The most widely used is the studentized residuals, whose values correspondto t values (Hair, 1998). Figure 7 shows different plots that illustrate the pattern that thedisturbance terms could take if some of the assumptions are violated. 35
    • The null plot (Figure 7a) is the usual pattern when all the assumptions are met.“The null plot shows the residuals falling randomly, with relatively equal dispersionabout zero and no strong tendency to be either greater or less than zero. Likewise, nopattern is found for large versus small values of the independent variable.” (Hair, 1998, p.173). By analyzing these plots the modeler could find violations to the assumptions andthen find remedies for those violations. These plots are the typical pattern one should findwhen violations occur. For example, nonlinearity (b) in the relationship between thedependent and explanatory variables; heteroscedasticity of the variance (c) and (d); andautocorrelation (e). The normal histogram of the residuals (g) allows the modeler to testthe assumption of normality of variance. A pattern like (f) would result when importantevents in the data are omitted in the specification of the response function (Hair, 1998).(For example, dummy variables that account for seasonality or special promotionalevents). 36
    • Figure 7. Graphical Analysis of residuals. (From Hair, 1998). 37
    • Plotting the residuals against the independent variables is quite useful, however,the prototypical patterns depicted in figure 7 are hard to detect for small samples andsometimes large samples as well. Some statistical tests have been developed for helpingthe modeler find violation to the assumptions in a more systematic way. For example, theDurbin-Watson (D.W.) test allows the model builder to test autocorrelations of thedisturbance terms. The D.W. statistic varies between zero and four. Small values indicatepositive autocorrelation and large values negative autocorrelation (Leeflang et al, 2000).Durbin and Watson formulated lower and upper bounds ( d L , d U ) for various significancelevels and for specific sample sizes and numbers of parameters. The test is used asfollows (for details see, Leeflang et al, 2000, p. 340):For positive autocorrelation a. If D.W. < d L , there is positive autocorrelation b. If d L < D.W. < d U , The result is inconclusive c. If D.W. > dU , There is no positive autocorrelationFor negative autocorrelation d. If [4-D.W.] < d L , there is negative autocorrelation e. If d L < [4-D.W.] < d U , The result is inconclusive f. If [4-D.W.] > dU , There is no negative autocorrelation Other tests have been developed for testing violations to other assumptions. Thedescription of those tests is outside the scope of this paper, for a detailed description seeHanssens et al, 2001, chap. 5; and Leeflang et al, 2000 chap. 16. 38
    • Leeflang et al, 2000, developed a table (table 2 in these paper) that summarizesthe violations to the assumptions in model building using Least Squares as well aspossible reasons, consequences, tests for detecting them and possible remedies.Table 2.Violation of the assumptions about the disturbance term: reasons, consequences, tests andremedies. (From Leeflang et al, 2000, p. 332) 39
    • As table 2 shows when some violation of assumptions are detected by eitherplotting the residuals or applying specific test, the modeler can try to take some remedies,often modifications to the specification of the model, or the use other estimationtechnique that relax the violated assumption (see Estimation in this paper). As frequentlymentioned by Leeflang et al (2000) and Hanssens et al (2001), violation of the model areusually caused by specification errors, so the first thing a modeler should do if the resultsare not satisfactory is to try a different functional form (see Specification in this paper) orto modify the specification of the model under scrutiny. The process of model verification is clearly explained in the following table (table3) taken from Hanssens et al, (2001).Table 3.Steps in evaluating a regression model. (from Hanssens et al, 2001) 40
    • Prediction Verification is just one part of the validation of a response model. The responsefunction in order to be believed must be able to predict future sales or market share forthe brand relative to the explanatory variables (Leckenby & Wedding, 1982). Forexample, if it is true that advertising expenditures can explain sales a valid model shouldbe able to predict the amount of sales in period x given a certain level of advertisingexpenditures in period x and probably previous periods. Because waiting for future salesdata to test a model is not only risky but useless if we want to use the model to forecast ordecide on future marketing and advertising expenditure levels, a process called“postdiction” is used. Postidction refers to the idea of predicting values that are alreadyknown. For example, a model is estimated using a sample that includes all the data fromthe past two years but not from this year even thought we already know the figures. Theprocess of postdicting is the use the model to predict the sales of this year given themarketing and advertising expenditures this year too. If the accuracy of the predictions isgood the model is a valid model for future forecast and then can be used in differentmanagerial decision making tasks. The way a modeler can perform this validation process is to split the sample ofdata in two subsets: one for estimating the model and the other for validating it using theprocess described above. With large samples this can be easily done by just leaving a fairnumber of data for validation purposes. However, the modeler usually does not have a lotof data to do this, so a minimum of three data points are left for validating the model. When the model is estimated using cross sectional data, the validation sub-sampleis chosen randomly but when the model is estimated using time-series data the last three 41
    • or more periods are reserved for the validation process. The reason for doing this is thatthe modeler would like to take into account the prediction accuracy when carryovereffects are involved in the response functions (Leeflang et al, 2000) and also because themanager would be more interested in the prediction accuracy of recent events than that ofdistant ones. There are several measures of the prediction accuracy of the model (see Leeflanget al, 2000, chapter 18) but the basic principle is to compare the predicted values with theobserved ones and calculate the average error of the predictions. The two most commonmeasures are the Average Prediction of Error (APE) and the Mean Absolute Percentageof Error (MAPE). The Average Prediction of Error is calculated by averaging the differencesbetween the observed and the estimated values. The procedure allows negative andpositive errors to offset each other (Leeflang et al, 2000). In accordance with the zeromean assumption in regression analysis (see Estimation in this paper) the APE should beclose or equal to 0. However, even with an APE of 0 a model could still have largeestimation errors if they offset each other. A better estimate of the prediction accuracy is the MAPE since it is a measure thatallows the modeler to asses the error as a relative measure (percentage) of the real orobserved value. The MAPE is calculated by averaging the absolute percentage of error | y− y| ˆ( .100 ) of each pair of predicted/observed data points in the validation sub- ysample. It is important to notice that if data outside the range used to estimate the modelare used to predict the outcome of the model, misleading results can occur. This isespecially important when using “non-robust” models like the linear ones where there is 42
    • no limit to the response of the dependent variable for larger values of the explanatoryones (Hair, 1998). The “postdiciton” procedure described above is an adequate method for testingthe validity of a model, however, “the acid test of the model’s validity still remain withpredictive test into the future” (Leckenby & Wedding, 1982). If the model can fairly oracceptably predict sales figures which have not yet occurred, then the model is useful andcan be used to solve marketing and advertising problems. A model should always andcontinually be checked for its prediction accuracy of future events as data becomeavailable. Model building summary Developing advertising and marketing response models is a fairly structuredprocess with defined steps. However, model building is an iterative process where theresults of one of the steps could suggest revising previous ones and start the processagain. The model building process also involves subjective judgments form the part ofthe modeler as frequent tradeoffs become present and the solutions require judgment andpersonal experience. For example, a usual tradeoff that the modeler faces is when in orderto enhance the prediction accuracy of a model he must make important changes to thespecification of the model making it harder to interpret and grasp significant economicmeaning. As Hair said (1998) “Prediction is often maximized at the expense ofinterpretation” (p. 161). The important role of the model builder in developing responsefunctions is what makes it part science and part art. Summarizing the steps in model building for marketing decisions, a good modelshould first, be specified in accordance to advertising or marketing theory; second, 43
    • estimated using an adequate estimation technique; third, verified using statisticalsignificance tests and analysis of residuals to look for violations of the assumptions; andfourth, validated using postdiction and prediction accuracy tests. Sometimes a modeler has competing models that have been verified and validatedand he or she must decide on which one to choose. The principle of parsimony wouldsuggest him to always pick the simplest one. However, it is sometimes hard to find theoptimal one since there is always a tradeoff involved in selecting a model that is simplebut less accurate and one that is more precise but with increasing complexity. One shouldalways evaluate the models with the original objective of the model building process inmind. Why were we building the model in the first place? What do we want to do withthe model? What is the managerial relevance or usefulness of the model? If the answer tothose questions still does not point toward one single model, there are some additionalprocedures that can be used to solve the problem of selecting between competing models.There are informal decision rules like “choose the model with the higher adjusted R 2 ”or“choose the one that has the least residual sum of square” and formalized decision rulesinvolving hypothesis testing (Hanssens et al, 2001). The formal decision rules include theMaximum Likelihood (ML) statistic, Akaike’s Information Criteria (AIC) and Bayesianinformation criteria (for details on those tests see Hanssens et al, 2001, p. 230-239). Ideally, the model to be chosen should be the one with the higher adjusted R 2 , theleast RSS, statistically significant t values, no autocorrelation and simpler structure.Fortunately, as Hanssens et al (2001) note: “the consequences in terms of deviation fromthe optimal level of discounted profits that arise from misspecifying market response isusually not great” (p. 239). 44
    • Once a model has been specified, estimated, verified, validated and compared toother possible competing models it can be used in decision making for planning futurescenarios, running controlled simulations and deriving economic measures for betteraccountability of past actions. The latter is the essence of model building in marketingand advertising: the better we understand the past the better we will predict the future. An Example In order to illustrate the process of developing marketing and advertising responsemodels, real data from an important brand in the skin-care market in a Latin Americancountry was used to build a model. Specifying the model After an initial exploration of the data that included an analysis of the multiplecorrelations between several variables and preliminary estimations of very basic responsefunctions, the following models where specified: 1. The Linear Current Effects response model: MS = a + b1TVR + b2U + b3 RP + b4T + b5 C + b6 M Where: MS = Market Share TVR = TV GRPs U = Advertising expenditures for the Umbrella brand RP = Relative Price (brand’s price/main competitor’s price) T = Trend (linear trend over time) C = Total competitors’ advertising expenditures M = Magazine advertising expenditures 45
    • 2. The Modified Exponential Current Effects model: MS = MS (1 − e a +b1TVR +b2U +b3 RP +b4T +b5C +b6 M )Where: MS = upper bound level or saturation point e = a mathematical constant equals to 2.71...16 …3. The Gompertz Current Effects model MS = MSe − e e a b1TVR b2U b3RP b4T b5C b6M e e e e e4. The Linear Partial Adjustment (Nerlove) model: MS = a + b1TVRt + b2U t + b3 RPt + b4Tt + b5Ct + b6 M t + b7 MSt −15. The Logistic Partial Adjustment (Nerlove) model: MS MS = − ( a + b1TVRt + b2U t + b3 RPt + b4Tt + b5Ct + b6 M t + b7 MS t −1 ) (1 + e )6. The Gompertz Partial Adjustment (Nerlove) model: MS = MSe − e e a b1TVRt b2U t b3RPt b4Tt b5Ct b6M t b7 MSt −1 e e e e e e7. The Modified Exponential Partial Adjustment (Nerlove) model: MS = MS (1 − e a +b1TVRt +b2U t +b3 RPt +b4Tt +b5Ct +b6 M t +b7 MSt −1 )8. The Linear Adstock model: MS = a + b1 Adstock + b2U + b3 RP + b4T + b5C + b6 MWhere: Adstock = TV GRPs Adstock 46
    • 9. The Logistic Adstock model: MS MS = − ( a + b1 Adstock + b2U + b3 RP + b4T + b5C + b6 M ) (1 + e ) 10. The Gompertz Adstock model MS = MSe − e e a b1Adstock b2U b3RP b4T b5C b6M e e e e e 11. The Modified Exponential Adstock model: MS = MS (1 − e a +b1 Adstock +b2U +b3 RP +b4T +b5C +b6 M ) All the above models assume independent effects of the explanatory variables.For example, the Linear CE model (number 1.) assumes that the market share for thebrand is a constant, plus the effect of TV GRPs, plus the effect of the advertisingexpenditures on the umbrella or family brand, plus the effect of the price relative to themain competitor, plus a trend in time, plus (minus) the effect of the sum of allcompetitors’ advertising expenditures, plus the advertising expenditures of the brand inmagazines. The assumption about independent effects means there is no interactions betweenthe variables, for example between TV advertising and magazines advertising. This mightnot be true in reality, in consequence, some models that assumed such interactions whereestimated but failed to deliver satisfactory results and no significant interactions wereidentified. It is important to notice that a trend in the data was incorporated into the model inorder to gain more predictive power. However, as the quote says: “a trend in a model is afactor you forgot to include in the explanatory consideration set”. Considering thatusually not all the data are available, adding a trend component is a partial solution to 47
    • lack of information and helps sometimes enhancing the model’s fit and its predictionaccuracy. However, as discussed earlier, there is usually a trade-off between predictionaccuracy and explanatory power. Trend components in models should be avoided if thereis no important improvement in the capacity of the model to make fair estimations of theobserved data. Knowing when to include or exclude a trend is part of the art of modeling. Other models where also specified but where discarded early in the processbecause they failed to fairly represent the relationship between advertising and marketshare for the brand. For example, the univariate Koyck Geometric Distributed Lag(GL) model: MSt = a(1 − c) + bTVRt + cSt −1 + {ut − cut −1}and the univariate Geometric Lag Autoregressive (GLA) model: MSt = a + b1TVR1 − b1ρTVRt −1 + (c + ρ ) MSt −1 − cρMSt − 2 + {ut − cut −1}failed to deliver satisfactory results. This occurred mainly because they used only oneexplanatory variable that, alone, seems not to contribute much on explaining the marketshare variance for this particular brand. Estimating the model Once specified, the above models (number 1 to 11), where estimated usingOrdinary Least Squares. Table 4 shows the parameter estimates for the Current Effectsfunctions and their derived statistics. Table 5 and table 6 show the same information forthe Nerlov Partial Adjustment models and the Adstock models. 48
    • Table 4.Current Effects models’ statistics and parameter estimates Current Effects Models unstandarized Model t Rsq Adj. Rsq DW*** RSS coefficients a = 4.000368 4.62 * b1 = 0.001629 4.45 * b2 = 0.000112 6.08 * 1. Linear b3 = 0.036254 4.12 * 93% 91% 2.629 1.54 b4 = 0.075848 6.08 * b5 = -0.000011 -2.76 * b6 = 0.000232 1.11 a = -0.149235 -1.10 b = -0.000258 -4.51 * b2 = -0.000017 -5.90 * 2. Modified Exponential**** b3 = -0.005595 -4.08 * 92% 89% 2.655 1.76 b4 = -0.010991 -5.65 * b5 = 0.000001 1.98 * b6 = -0.000018 -0.55 a = 0.469480 1.48 b1 = -0.000587 -4.40 * b2 = -0.000038 -5.70 * 3. Gompertz***** b3 = -0.012564 -3.91 * 91% 88% 2.688 1.90 b4 = -0.023992 -5.27 * b5 = 0.000003 1.71 ** b6 = -0.000030 -0.40 Sample Size = 23 *p < .05 **p < .15 *** DW < .90 significant autor; .90 > DW <1.92 inconclusive; DW > 1.92 no significant autor. at .05 ****Upper Bound =15 *****Upper Bound =12 49
    • Table 5.Partial Adjustment models’ statistics and parameter estimates Partial Adjustment Models unstandarized Model t Rsq Adj. Rsq DW*** RSS coefficients a = 3.729706 3.88 * b1 = 0.001686 4.43 * b2 = 0.000100 4.08 * b3 = 0.031444 2.80 * 1. Linear 94% 91% 2.667 1.49 b4 = 0.068501 4.17 * b5 = -0.000011 -2.71 * b6 = 0.000243 1.14 b7 = 0.095134 0.70 a = -0.126388 -0.83 b = -0.000262 -4.37 * b2 = -0.000016 -4.11 * b3 = -0.005189 -2.92 * 2. Modified Exponential**** 92% 89% 2.654 1.74 b4 = -0.010371 -4.00 * b5 = 0.000001 1.92 * b6 = -0.000019 -0.56 b7 = -0.008031 -0.38 a = -1.020810 -3.86 * b1 = 0.000461 4.40 * b2 = 0.000028 4.09 * b3 = 0.008621 2.78 * 3. Logistic**** 93% 90% 2.668 1.52 b4 = 0.018575 4.11 * b5 = -0.000003 -2.63 * b6 = 0.000064 1.09 b7 = 0.024295 0.65 a = 0.405843 2.01 * b = -0.000353 -4.41 * b2 = -0.000021 -4.11 * b3 = -0.006763 -2.86 * 4.Gompertz**** 93% 90% 2.667 1.59 b4 = -0.014045 -4.07 * b5 = 0.000002 2.30 * b6 = -0.000038 -0.84 b7 = -0.015138 -0.53 Sample Size = 23 *p < .05 **p < .15 *** DW < .90 significant autor; .90 > DW <1.92 inconclusive; DW > 1.92 no significant autor. at .05 ****Upper Bound =15 50
    • Table 6.Adstock models’ statistics and parameter estimates Adstock Models unstandarized Model t Rsq Adj. Rsq DW*** RSS coefficients a = 4.069391 4.89 * b1 = 0.002388 4.72 * b2 = 0.000082 4.67 * 1. Linear b3 = 0.036416 4.28 * 94% 91% 2.223 1.44 b4 = 0.067203 5.67 * b5 = -0.000013 -3.39 * b6 = 0.000325 1.71 ** a = -0.933690 -4.10 * b = 0.000657 4.75 * b2 = 0.000022 4.67 * 2. Logistic**** b3 = 0.009899 4.26 * 94% 91% 2.218 1.47 b4 = 0.018084 5.58 * b5 = -0.000003 -3.31 * b6 = 0.000086 1.66 ** a = 0.347017 1.98 * b1 = -0.000501 -4.71 * b2 = -0.000017 -4.54 * 3. Gompertz***** b3 = -0.007556 -4.22 * 93% 91% 2.161 1.60 b4 = -0.013388 -5.37 * b5 = 0.000002 2.95 * b6 = -0.000056 -1.40 a = -0.162484 -1.23 b = -0.000372 -4.62 * b2 = -0.000012 -4.38 * 4. Modified Exponential**** b3 = -0.005608 -4.14 * 92% 90% 2.094 1.82 b4 = -0.009619 -5.10 * b5 = 0.000002 2.53 * b6 = -0.000034 -1.11 Sample Size = 23 Adstock Half Life = 1 period. Carry-over = 33% *p < .05 **p < .15 *** DW < .90 significant autor; .90 > DW <1.92 inconclusive; DW > 1.92 no significant autor. at .05 ****Upper Bound =15 51
    • Verifying the model As table 4, 5, and 6 show the R 2 and adjusted R 2 of all the models is 90% orabove. That means they all explain at least 90% of the variance of the market share of thebrand in the period analyzed. The adjusted R 2 is more useful for comparing the CE andAdstock models with the PA models since the Partial Adjustment models include anadditional lag parameter and the R 2 is sensible to the number of variables in the model. Other way of measuring the ability of the models to fairly represent therelationship between the explanatory variables and the dependent one is by analyzing themodel fit to the data in the sample. The Residual Sum of Squares (RSS) delivers a directmeasure of the “unfitness” of the model. The estimated models show small RSS varyingfrom 1.44 to 1.90. The Durbin-Watson statistic shows that none of the estimated models showsignificant autocorrelations. This is especially important if we desire that the estimationprocess delivers unbiased and statistically significant parameter estimates. As discussed under the verification section in the first chapter, the estimationprocess should deliver statistically significant parameter estimates so the modeler couldproject the model beyond the data sample. In other words, the parameter estimates shouldhave a value different form zero meaning that their associated variables have a real effectin the dependent variable. The estimated models vary in this criterion since not all ofthem have all statistically significant coefficients or parameter estimates. Actually, justthe Adstock Linear and the Adstock Logistic model have statistically significant b6coefficient. Interestingly, the parameters corresponding to the lagged variable ( b7 ) in theNerlove PA models are not statistically significant. This means that these PA models 52
    • actually reduce to the Current Effects ones since the only difference between them is theadditional lagged variable. Another very interesting result is that the coefficient for the relative price ispositive. Since the variable was defined as the ratio between the brand’s price and themain competitor’s price (brand’s price/main competitor’s price) it is surprising to realizethat, at least for the data analyzed, the highest the ratio the highest the market share allelse being equals. Since the parameter’s sign is consistent across all models it should notbe discarded. There are situations in which raising the price actually raise the demand ofthe product because it acts as a clue that signals good quality. This phenomenon has beendetected in many specialty products, including beauty products (Kotler, 1971). The brandis a competitive brand in the “wrinkle prevention” market, a highly specialized categorydriven mainly by research and product innovation. It is not unlikely that this is one ofthose special cases where the relation between price and demand is reversed. The branduse to have a lower price compared to its main competitor but it seems that the closer theprice of the brand to the price of its main competitor the higher the demand for the brand.This result should be taken with caution and would apply probably only for the datarange analyzed (min = 51.5; max = 101.7; mean = 76.13; std. deviation = 11.43). In order to check for violations of the OLS assumptions residuals’ scatter plots ofthe best four models (CE Linear model, PA Linear model, Adstock Linear model andAdstock Logistic model) where analyzed. Figure 8 shows the scatter plots of thestudentized residuals vs. the actual market share values for the four models. Nosystematic pattern is observed for any of the models analyzed showing that nofundamental assumption was violated. However, some outliers can be recognized, 53
    • especially two outliers for the adstock models. The treatment of outliers is controversial(Hair, 1998) but a careful analysis should be provided in order to asses their impact onthe overall performance of the model. We will discuss this latter.Figure 8. Scatter plots of the studentized residuals vs. the actual market share values forthe CE Linear model, PA Linear model, Adstock Linear model and Adstock Logisticmodel. Ideally the best model should have all statistically significant coefficients, noautocorrelation, the highest R 2 or adjusted R 2 and the lowest RSS. However, not alwaysall of these criteria can be found in one single model as it is the case for the Adstock 54
    • Linear model in our example. Additionally, the best model is not really identified untilthe acid test is performed. So before selecting a single model the best ones should bevalidated using the prediction/postdiction procedure. Validating the model The best competing models (CE Linear model, PA Linear model, Adstock Linearmodel and Adstock Logistic model) were selected to be validated using a subset of thesample. The sample of data was split into two subsets: one with the first 20 observationsto estimate again the parameters of the model and the other one with the last 3 to bepredicted/postdicted by the model. The Mean Absolute Percentage of Error (MAPE) wasused to compare the prediction ability of the models. Table 7 shows the results and all thestatistics for the selected models. All the models have MAPEs below 3,5% which means that they all can makeaccurate predictions of future outcomes. However, the Adstock models clearlyoutperform the CE and PA linear models. The principle of parsimony would suggestchoosing the simplest model between two competing ones. The MAPE criteria as well asall the other criteria also point the Linear Adstock model as the winner. Figure 9 showsthe modeled market share versus the actual market share for the Adstock Linear Model. 55
    • Table 7.Best models comparison Best Models unstandarized Model t Rsq Adj. Rsq DW*** RSS MAPE coefficients a = 4.000368 4.62 * b1 = 0.001629 4.45 * b2 = 0.000112 6.08 * 1. CE Linear b3 = 0.036254 4.12 * 93% 91% 2.629 1.54 2.99% b4 = 0.075848 6.08 * b5 = -0.000011 -2.76 * b6 = 0.000232 1.11 a = 3.729706 3.88 * b = 0.001686 4.43 * b2 = 0.000100 4.08 * b3 = 0.031444 2.80 * 2. PA Linear 94% 91% 2.667 1.49 3.01% b4 = 0.068501 4.17 * b5 = -0.000011 -2.71 * b6 = 0.000243 1.14 b7 = 0.095134 0.70 a = 4.069391 4.89 * b1 = 0.002388 4.72 * b2 = 0.000082 4.67 * 3. Adstock Linear b3 = 0.036416 4.28 * 94% 91% 2.223 1.44 1.65% b4 = 0.067203 5.67 * b5 = -0.000013 -3.39 * b6 = 0.000325 1.71 ** a = -0.933690 -4.10 * b = 0.000657 4.75 * b2 = 0.000022 4.67 * 4. Adstock Logistic**** b3 = 0.009899 4.26 * 94% 91% 2.218 1.47 1.71% b4 = 0.018084 5.58 * b5 = -0.000003 -3.31 * b6 = 0.000086 1.66 ** Sample Size = 23 (note: the MAPE was calculated using parameter estimates from a sample data of 20) Adstock Half Life = 1 period. Carry-over = 33% *p < .05 **p < .15 *** DW < .90 significant autor; .90 > DW <1.92 inconclusive; DW > 1.92 no significant autor. at .05. ****Upper Bound =15 56
    • Model Fit 12.00 10.00 8.00Market Share 6.00 Actual 4.00 Model R square = 94% 2.00 MAPE = 1.65% 0.00 r Ju er ec er ec er st st y y r r il il ch ay ne y ch ay ne ov e r ov e r ry ry be be be ar pr ar pr l Ju Se ug Se ug b b b ua ua ob ob M M ar ar Ju Ju an an em em em em A A em em A A M M br br ct ct Ju pt pt O O Fe Fe D D N N Figure 9. Modeled market share and observed market share for the Adstock Linear Model. It is interesting to notice that the model fits quite well the observed values except for two points corresponding to November and December of the second year. These two points are the outliers identified in figure 8. Giving the coincidence that these outliers correspond to two consecutive months, and especially November and December where the category and the brand is affected by different events for Christmas and New Year, it is not unlikely that the inaccuracy of the model is due to a special activity performed by 57
    • some of the competitors in those two months. It could also be the result of a specialactivity for the brand like an “end of the year” sales promotion. However, the graphshows that the model outperforms the actual values, which means that whatever theactivity that is creating the discrepancy is, it acts in detriment of the actual market share. A more likely event would be a sales promotion developed by one of the maincompetitors. This kind of event could be included in the model using a dummy variableaddressing the special activity for the two month. For illustration purposes only, theAdstock Linear model was modified in order to include the dummy variable. Table 8shows the comparative results of the original model and the model including the dummyvariable. Figure 10 shows the fit of the modified model.Table 8.Comparative results for the original Adstock Linear model and the Adstock Linear modelincluding a dummy variable. Adstock Linear Models unstandarized Model t Rsq Adj. Rsq DW*** RSS MAPE coefficients a = 4.069391 4.89 * b1 = 0.002388 4.72 * b2 = 0.000082 4.67 *1. Adstock Linear b3 = 0.036416 4.28 * 94% 91% 2.223 1.44 1.65% b4 = 0.067203 5.67 * b5 = -0.000013 -3.39 * b6 = 0.000325 1.71 ** a = 4.051919 8.02 * b = 0.003125 9.30 * b2 = 0.000070 6.42 * b3 = 0.035734 6.93 *2. Adstock Linear (Dummy) 98% 97% 2.400 0.49 2.44% b4 = 0.072525 10.00 * b5 = -0.000013 -5.46 * b6 = 0.000207 1.76 ** b7 = -0.821492 -5.34 *Sample Size = 23 (note: the MAPE was calculated using parameter estimates from a sample data of 20)Adstock Half Life = 1 period. Carry-over = 33%*p < .05 **p < .15*** DW < .90 significant autor; .90 > DW <1.92 inconclusive; DW > 1.92 no significant autor. at .05. 58
    • Model Fit 12.00 10.00 8.00Market Share 6.00 Actual 4.00 Model (dummy) R square = 98% 2.00 MAPE = 2.4% 0.00 r Ju er ec er ec er y y st st r r il il ay ne y ay ne ch ch ov e r ov e r ry ry be be be ar ar pr pr l Ju Se ug Se ug b b b ua ua ob ob M M ar ar Ju Ju an an em em em em A A em em A A br br M M ct ct Ju pt pt O O Fe Fe D D N N Figure 10. Modeled market share and observed market share for the Adstock Linear model that includes a dummy variable. As table 8 shows and comparing figure 10 with figure 9, including a dummy variable that represents the atypical activity during November and December for the second year enhances the model’s performance. However, the MAPE is bigger by 0.79%; but considering that the MAPE was calculated using only the last three data points it is not a significant difference. As can be seen from this exercise, modifying the specification of the model by including dummy variables allows the modeler to represent especial events that enhance the performance of the model. It also forces the model builder and manager to search for 59
    • possible explanations of the atypical events. Because there is no information available tofully understand what the dummy variable is really representing in our exercise, theoriginal model will be used. Using the model Once the model was specified, estimated, verified and finally validated it can beused to solve marketing and advertising management problems. For example, measuringthe impact of the past marketing actions in sales or market share is critical in currentmarketing practices. The pressure to deliver accountable results is partially responsiblefor the raise of econometric modeling in advertising. As it was discussed at the beginningof this paper, understanding the past in order to make better decisions in the future is thecore idea of econometric modeling in marketing and advertising. An easy way of observing the effects of the different marketing activities on thebrand’s market share is by performing a market share decomposition analysis. Once theeffect of each marketing activity is isolated it can be plotted in conjunction with theothers effects so as the sum of all the effects plus a base level result in the actual marketshare. The different effects can then be compared by analyzing the resulting graph.Figure 11 shows the decomposed market share for the brand. 60
    • Market Share Decomposition 12.00 10.00 8.00Market Share 6.00 4.00 Total Competitors Adspend Magazine Adspend TV GRPs Adbank 2.00 Umbrella Adspend Change in relative price Base 0.00 r r ec er ec er t y st y r r ch il ay e y ch il ay e ov e r ov e r ry ry Se ugs be be be be ar pr ar pr n l n Ju ug b b ua ua ob b M M ar Ju ar Ju an an em em em em em A em A o A A br M br M ct ct Ju Ju pt pt O O Fe Fe Se D D N N Figure 11. Market Share Decomposition It is important to notice that the effect of the relative price was modified in order to show the effects of variations from the minimum value (50% June of the second year). As figure 11 shows changes in price relative to the main competitor have an important effect on market share. When the ratio was close to 100 (August, September and October of the first year) the additional contribution in market share was the highest. As discussed earlier this model suggests to raise the price to match the main competitor’s price since the effect is positive, possibly because it signals product quality. 61
    • It is interesting to notice that the GRPs’ adstock has the second highest effect. Byanalyzing the graph it is easy to recognize its additional contribution to market share. Thehalf-life used is 1 month suggesting that 33% of the effects in one period are carried overthe next ones. More interesting is to realize that the advertising expenditures for the umbrellabrand have an important direct effect in the brand’s market share. It is in part responsiblefor the good performance of the brand at the beginning and end of the period analyzed.Since the cost of the umbrella advertising is shared among all the other brands of thecompany it is a fascinating cross effect. It also reveals the importance of the brand’sname and brand’s associations in the performance of the product. The advertising expenditures in magazines also show an immediate (no carryovereffect) but small effect in market share. It is important to say that we are only analyzingthe short-term effects of advertising but the reader should keep in mind that advertisingalso have an important role in brand building that could be accurately identified with theanalysis of a longer time period (minimum 3 years). It would be inaccurate to say that theeffects of advertising for the umbrella brand as well as advertising for the brand limits tothe additional contribution of market share over the base. Advertising actually affects thebase of the brand. This effect can be modeled with a more sophisticated technique inwhich the base is not assumed to be constant but floats over a dynamic tension (fordetails see Broadbent, 1997). It is also possible to model the variations in the base withkey performance indicators like brand awareness and brand likeability as independentvariables. 62
    • The grey area in the graph represents the market share that is lost or not gainedbecause of the effects of all competitors’ advertising. It is actually taken away from thebase and it is represented here as potential market share that would be gained if there wasno competitors’ advertising. These effects are probably the primary factor responsible forthe poor performance from March to December of the second year. This analysis suggestsraising advertising under such periods in order to compensate the negative effects ofcompetitors’ advertising. It is also interesting to notice that because the adstock takestime to built, once it fall to zero (February of the second year) the brand is heavilyaffected by competitors advertising (see figure 11 from February to June) and recoversonly after several periods under heavy advertising investment. It is also important to recall that the model incorporates a trend component that isnot evident in figure 11. It was added to the base and is the responsible for its tendency togrow over time. Once this visual analysis is performed it is useful to quantify the contribution ofeach of the market share drivers in order to compare them and derive more importanteconomic measures. Figure 12 shows the relative contribution to market share ofadvertising and price. It is interesting to notice that advertising in TV and magazines forthe brand and advertising for the umbrella brand, all together, have a very similarcontribution to market share than changes in the relative price. Also important to notice isthat the base for the brand is responsible for almost 80% of the market share for thebrand. The base as it was considered in the model includes the contribution of thedistribution strategy, packaging, sales promotion and all the other marketing variables notexplicitly represented in the model. The basis also includes a trend component. Under 63
    • certain circumstances the base can be understood as an important component of the brandequity. Market Share Contribution Change in relative price 12% Umbrella adspend 2% Magazines adspend 2% GRPs 6% Base 78% Figure 12. Contribution of advertising and price to the brand’s market share. Another useful tool derived from the model is a comparative table with theprincipal influences, their impact on the brand’s market share, their power and theirelasticities. That information is reported in table 9. 64
    • Table 9.Significant Influences on the brand’s market share. Significant Influences Factor Impact Power Elasticity GRPs positive short-term 0.24 points / 100 GRPs 0.06 Magazines adspend positive direct 0.03 points / 100,000$ 0.02 Umbrella adspend positive direct 0.008 points / 100,000$ 0.02 Total competitors adspend negative direct -0.01 points / 1,000,000$ -0.09 Relative price positive direct 0.036 points / 1unit increase 0.36 Trend positive direct 0.07 points / period - It is interesting to notice that even though the advertising expenditures inmagazines and the total advertising expenditures for the umbrella brand have the sameelasticities but their power is different. Because the advertising expenditures for theumbrella brand is a much bigger number than the advertising expenditures in magazinesfor the brand, a 1% change is a lot of money and as the table shows, 0.008 points ofmarket share are added per every $100,000 in advertising for the umbrella brand. Thetrend component even though is small it contributes to the brand’s market share. Furtherresearch should be done in order to identify what the trend really represents and to beable to fully comprehend the significant influences on the brand’s market share. The advertising and price elasticities represent the change, as a percentage, inmarket share per 1% change in price or advertising expenditures. The percentages for theadvertising variables in figure 12 are actually the advertising elasticities multiplied by100. It can be interpreted as what would happen if there is no advertising for the brand orthe umbrella brand. It other words, one could said that a change of 100% in advertising(dropping all advertising) would result in lost of 10% of the brand’s market share. 65
    • One final analysis that can be performed once the model was developed is thereturn of investment for the different advertising activities during the time periodanalyzed. Table 10 shows the result of such analysis. Table 10. ROI analysis. ROI Analysis Medium/Brand Investment Return Gross ROI TV adspend 107,562.81 56,251.2 52% Magazines adspend 9,914.72 19,279.3 194% Umbrella adspend 37,397.28 22,219.5 59% The return was calculated multiplying the number of market share points thatresulted for the advertising activities during the whole period and the total industry salesfor the same time period. Interestingly, even thought the effects of the advertisingexpenditures in magazines are small compared to TV the money investment is paid backwith an additional 94% percent of return. It is not the same case for TV expenditures. Anarrow interpretation of this result would be that in the immediate and short-termadvertising on TV does not pay back and only 52% is returned by the end of the timeperiod analyzed. However, as it was discussed earlier advertising not only have short-term effects in market share and sales but it also have more pervasive effects in the brandequity. It is said that advertising’s primary role is brand building so the analysis of itsROI should be taken with cautious since the model does not account for those effects ofadvertising. With a larger time period the model could be modified to include a brandingor long-term effect of advertising as part of its main components. 66
    • The ROI for advertising for the umbrella brand is only 59%, however, this isactually a big figure since it is a cross effect and the cost of those expenditures areactually shared with the umbrella brand and the rest of the company’s brands under thesame name. So, it means that 59% of what was invested in advertising for the umbrellabrand was returned via the sales that the brand under analysis generated. One of the advantages of building a model is that it can be used to test differentstrategies or “policies” and evaluate them before there are even implemented. Anothervery practical use of models in advertising is in determining the advertising allocation.For some examples of advertising models built and used to help determine the advertisingor communications budget go online at:http://www.ciadvertising.org/sa/spring_05/adv391k/eribero/frameset.htm orhttp://www.ciadvertising.org/sa/spring_05/adv391k/eribero/Solo2/frameset.htm As an illustration of using the model to run some scenarios we use the same datafor the period analyzed and just vary the advertising expenditures in magazines and theGRP allocation for the same period. Two scenarios are compared to the original modeledmarket share. Figure 13 shows the result of the two simulations. 67
    • Scenario Planning 12.00 10.00 8.00Market Share 6.00 4.00 Modeled market share Scenario 1 2.00 Scenario 2 0.00 r Ju er ec er ec er y y r r st st il il ne y ne ry ch ay ry ch ay er ov er be be be ar pr ar pr l Ju Se u g ug b b b ob ua ob ua M M ar ar Ju Ju an an em em em em A A em em A A br M br M ct ct Ju ov pt pt O O Fe Fe Se D D N N Figure 13. Comparison of the simulation of two different scenarios for the brand using the model developed. In scenario 1 we increased by 23% the number of GRPs and by 38% the advertingexpenditures in magazines. The extra GRPs and money where distributed between themonth of March and October of the second year in order to protect the brand againstcompetitors’ advertising. The red line in figure 13 shows that by applying the describedstrategy the market share for the brand moves more smoothly preventing it to drop duringthat time period as it did in reality. In scenario 2 we followed the suggestion from themodeling process to increase price in order to match the main competitor’s price.Scenario 2 also takes into account the modification done for scenario 1. By analyzing Figure 13 it is evident that by raising the price the brand performsconsiderably better. Actually the total difference between the modeled market share and 68
    • the one simulated under scenario 2 is 21.4 points during the entire time period. And thebrand is not only getting more points of market share it is receiving more money perevery unit sold! Recall however that the suggestion to raise the price should be taken withextreme caution. Here we use the extreme scenario in which we match the maincompetitor’s price just to illustrate the point. However, in a real situation the managersshould first, confirm that by raising the price the demand for the product would actuallyraise as well; and second, the price should be increased gradually and not in just onemoment. Summary Econometric modeling in advertising is an exiting activity. It gives the modelersand brand managers as well as advertising managers the opportunity to evaluate theadvertising performance and impact on sales and solve recurrent managerial problems ina systematic way. As discussed earlier, modeling is part science part art. The modelerexperience and subjective view of how advertising works for a specific brand is testedusing advanced statistical methods. The model is specified in an attempt to represent themost important elements of a reality, but even for the simplest problem this task requiresjudgment as well as deep knowledge of advertising theory. Once the model is specifiedand estimated, it is verified, tested and revised. Then it can be used to derive importanteconomic measures, or to set the communications’ budget or to test strategies underdifferent scenarios. Understanding the past is sometimes the only way to improve future’sperformance. It is important to say however, that modeling is just a tool that wouldhopefully help mangers to make better decisions by delivering valuable information butsuccess will always rely on good judgment and probably some intuition too. 69
    • References Broadbent, S. (1979) One-Way TV Advertisements Work. Journal of MarketResearch Society, 21, 139-165. Broadbent, S. (1984) Modeling With Adstock. Journal of Market ResearchSociety, 26, 295-312. Broadbent, S. (1997) Accountable Advertising. A Handbook for Managers andAnalysts. UK: NTC Publications. Henley-on-Thames. Bucklin, R.E., Lehmann, D.R. & Little, J.D.C. (1998) From Decision Support toDecision Automation: A 2020 Vision, Marketing Letters, 9, 234-246. Hair, J.F., Tatham R.L., Anderson, R.E., Black W. (1998) Multivariate DataAnalysis (5th edition). Prentice-Hall. Hanssens, D.M., Parsons, L.J. & Schultz, R.L. (2001) Market Response Models.Econometric and Time Series Analysis. (Second edition). Boston. Kluwer AcademicPublishers. Kotler, P. (1971) Marketing Decision Making: A Model Building Approach. NewYork. Holt, Rinehart and Winston. Koutsoyiannis, A. (1978) Theory of Econometrics (2nd edition). Textbookbinding. Rowman & Littlefield Publishers Inc. Leckenby, J.D. & Wedding, N. (1982) Advertising Management: Criteria,Analysis, and Decision Making. New York: John Wiley & Sons, Inc. Leeflang, P.S.H, Wittink, D.R., Wedel, M. & Naert, P.A. (2000) Building Modelsfor Marketing Decisions. Boston. Kluwer Academic Publishers. 70
    • Lilien, G.L. & Rangaswamy, A. (2002) Marketing Engineering (2nd Edition).Reading, MA. Addison-Wesley. Little, J.D.C. (1970) Models and Managers: The Concept of a Decision Calculus.Management Science, 16, 466-485. Patrick, A.O. (2005) Econometrics Buzzes Ad World As a Way of MeasuringResults. The Wall Street Journal (October 16th). 71
    • Vita Esteban Ribero was born in Bogotá, Colombia, on April 7, 1978 the son of RafaelRibero and Ana Isabel Parra. After attending high school at the Colegio Helvetia, in 1997he entered the Universidad de los Andes in Bogotá, Colombia and attended the PotificiaUniversidad Javeriana for courses in marketing and advertising. He received the degreeof Bachelor of Arts in Psychology from the Universidad de los Andes in 2002. Heworked as a strategic planner for three years at the TBWAColombia advertising agencybefore entering The Graduate School at The University of Texas at Austin in August,2004..Permanent Address: 2501 Lake Austin Blvd K 208 Austin, TX, 78703.This report was typed by the author. 72