Profit or Prodigality?
The Influence of Marketing Spending on Customer Value in the Dutch
Energy Market
Title: Profit or Prodigality?
Subtitle: The Influence of Advertising on Customer Value in the Dutch Energy Market
Author: F.C. de Jager
Department: Marketing, Faculty of Economics and Business; Qualification: Master thesis
Completion date: May 18th 2012
Address: Eperweg 29
8181 ET Heerde
Phone number: 06-51044381
E-mail: fcdejager@gmail.com
Student number: s1910205
Supervisor: prof. dr. P.C. Verhoef Second supervisor: van Nierop, J.E.M. External supervisor: S. Lhoëst-Snoeck
Management Summary
This paper set out to find out “How above-the-line and below-the-line marketing efforts of the focal company influence the total value of their customer portfolio”. Also how acquisition and retention interact; if any synergies exist between different marketing communication channels; and what the result of an increase/decrease of x%
/ €x in any of the marketing investments on the customer value, and vice versa? To
answer these questions a statistical model was estimated that linked marketing communication expenditures with the total number of acquired and retained customers, as well as the revenues of acquired and retained customers.
Of all the marketing communication channels, telemarketing, direct mail, door-to-door and online were found to deliver the best results over time. Unfortunately due to statistical problems the marketing communication channels had to be summed in above-the-line (ATL) and below-the-line (BTL) variables. ATL marketing
communication channels were found to have a negative effect on retention while they also positively influence the profits of retained customers. BTL marketing
communication channels were found to have a positive effect on acquisition and also a positive effect on the profits of acquired customers. The market simulation showed that an investment in ATL would deliver a higher return on investment than an investment in BTL.
In the definitive model that was estimated the interactions between acquisition and retention were mostly insignificant, except for the negative effect acquisition had on churn. This is contrary to what was expected. It is most likely due to the fact that current customers react on campaigns designed for new customers, this is a way for them to get cheaper energy contracts. It is also not uncommon that customers who indicate that they are contemplating switching to be offered deals meant for new customers.
maintained at current levels. Cutting marketing expenditures in ATL or BTL
marketing communication channels should be avoided as they do both have a positive return on investment. ATL marketing communication channels are found to add the most value, therefore these channels should be used more prominently in the
Foreword
Table of Content
Management Summary ...3 Foreword ...5 Table of Content ...6 1. Introduction...8 1.1 Study Context...10 1.2 Contributions...11 2. Theoretical Framework ...132.1 Value of the Customer Portfolio ...15
2.1.1 Customer Equity ...15
2.1.2 Customer Profitability Equation ...16
2.1.3 The Friction between Acquisition and Retention ...17
2.2 Communication Mix ...18
2.2.1 Above-the-line ...19
2.2.2 Below-the-line...21
2.3 General Effects of Marketing Communications ...23
5.3 Model Validation ...46
6. Market Simulations...48
7. Conclusion ...50
8. Managerial Implications ...52
9. Limitations and Future Research ...53
10. References...54
1. Introduction
One of the core tasks of a business is the maximization of value, for their share holders and stake holders. This financial view on a business can be applied to every aspect of the business. In recent years the maximization of customer value has become an important area for businesses and academics alike (Blattberg, Getz & Thomas, 2001). Maximizing the value of customers means that the business in question needs to understand what the revenues and costs of individual customers are. Maximizing customer value is thus impossible without marketing accountability, which is a 'hot' topic at the moment in marketing literature. In fact the Marketing Science Institute ranked accountability in marketing as its top priority in 2008-2010 (Lamberti & Noci, 2010).
The main objective of this paper is to determine what marketing expenditure (mix) should be used to optimize customer profitability and its underlying components.
Problem Statement:
How do the above-the-line and below-the-line marketing efforts of the focal company influence the total value of their customer portfolio?
And Research Questions:
1. How do retention and acquisition influence each other?
2. Do synergies in acquisition and retention exist between marketing communication channels?
3. What would be the result of an increase/decrease of x% / €x in any of the marketing investments on the customer value, and vice versa?
1.1 Study Context
The focal company operates in the Dutch Energy sector. The privatization of the Dutch energy market has led to major changes in how the market functions. Before the privatization each energy company had its own assigned region in which it operated and competition was non-existent. Due to the privatization of the market the incumbent energy providers were faced with a radically different and more competitive market. This has meant a radical change in how the focal company and other previously state-owned energy companies had to do business. Customers finally had a choice in their energy supplier. This is the reason why the focal companies’ current strategy is to transform the organization from being a customer’s commodity supplier to their trusted energy adviser.
1.2 Contributions
In this thesis the relations between the spending on various marketing communication channels and the returns on the value of acquired and retained customers is examined. Although multi-channel management has gotten a lot of attention in recent years (Neslin, 2006; Valos, 2008; Verhoef, Neslin & Vroomen, 2007), few models exist that try to tie together the effects of various marketing communication channels on customer profitability. The paper by Reinartz, Thomas and Kumar (2005) focuses on this subject, but they study fewer marketing communication channels and only look at profitability as a whole. Berger and Bechwati (2001) study the allocation of
promotion budget to maximize customer equity but do not look at acquisition and retention separately. In contrast to these studies this paper will examine the influences of particular above-the-line and below-the-line marketing communication channels on the costs and revenues of acquisition and retention.
Another study by the same academics focuses on building a framework that enables users to compare different marketing strategies on projected financial returns measured as customer equity (Rust, Lemon & Zeithaml, 2004). This model takes a meta-view on what drives the value of customers, this paper differs distinctly from it since it focuses on the effects of individual marketing communication channels and the interaction between acquisition and retention.
“Impact of Marketing-Induced vs. Word-of-Mouth Customer Acquisition on Customer Equity Growth” (Villanueva, Yoo & Hanssens, 2008) explores the
difference in added customer equity by company based marketing activities and word-of-mouth.
The aim of this study is to provide invaluable new knowledge on the interactions and effects of various marketing communication channels on customer value of acquisition and retention. It will also give more knowledge on the link between acquisition and retention. The managerial contribution is that it help management to allocate funds over the marketing activities regarding acquiring new customers and retaining existing customers.
2. Theoretical Framework
As noted in the introduction, analytics in marketing is gathering a lot of attention in practice and academia; this is part of the increasing focus on marketing accountability. The management of multiple marketing communication channels is of course not as hot a topic as it was, but it is still the subject of much research and debate. This paper centres on the question, ‘how do the above the line and below the line marketing efforts of the focal company influence the total value of their customer portfolio’. As discussed before this calls for the analysis of the influence of the channel mix on acquisition and retention and how these channels interact in creating value.
Figure 1: Conceptual Model
Channel Mix the Focal Company
Competition - Share of voice BTL Direct Mail Door-to-Door Telemarketing Adwords E-mail ATL Offline Online
Average Value of Customer Portfolio
# New customers
# Existing customers
Average value per new customer
The conceptual model is presented in figure 1 on the previous page. Businesses steer their actions on their performance: number of acquired customers, churn, profit, etcetera. It is reasonable to expect a feedback loop in the interactions between expenditures on the channel mix and the total value of the customer portfolio. Channel mix expenditures will affect the number of acquired and retained customers, which in turn affects the channel mix expenditures in following periods. Therefore it is expected that there are relations between marketing activities through the ATL and BTL channels (measured in expenditures) and the total value of the customer portfolio and vice versa. And since the energy market is not divided into a geographic monopoly anymore, competition is also expected to impact the total value of the focal companies’ customer portfolio.
In this chapter the value of the customer portfolio is discussed first. Secondly channel mix is presented, which is subdivided into below-the-line and above-the-line
2.1 Value of the Customer Portfolio
To practice marketing accountability a company needs measures to evaluate its performance. Since marketing focuses on the needs of the customer the measures marketing accountability needs have to be customer centric and preferably
quantifiable and objective. One important concept that is utilized by many businesses practising marketing accountability is customer equity or profitability. In this chapter, first the concept of customer equity is presented, after which customer profitability is discussed. Finally, the relation between profitability, acquisition and retention is studied.
2.1.1 Customer Equity
Customer equity management is the view that customers are a financial asset to companies and should be measured and managed as such. Customer equity
management utilizes financial valuation techniques to optimize acquisition, retention and the selling of add-on products (or up-sell) (Blattberg, Getz & Thomas, 2001). It maximizes the value of a customer to the company over the customers’ life cycle. It combines the separate focus on revenue growth and cost management by evaluating the return on investment of every marketing investment. Thus eliminating either the rigid focus on cutting costs or revenue growth through acquisition (Blattberg, Getz & Thomas, 2001).
Quantifying the value of customers allows marketers to calculate what a firm could and should spend on retaining and attracting customers in order to increase the value of their customer portfolio. Thus benefiting fact based marketing decisions and firm profitability as a whole. It allows for better evaluation of the effectiveness of
2.1.2 Customer Profitability Equation
The concept of customer equity is slightly modified for use in this paper. Customer equity can be dissected into a number of different components. As mentioned in the previous section it focuses mainly on acquisition and retention. The main
modification is that with customer equity the valuation of customers is done over their lifetime (that they are customer with the focal company). While this paper focuses on customer profitability.
The value of the total customer portfolio is the acquisition profitability plus the retention profitability. The acquisition profitability is the revenues generated by acquired customers in a given period minus their costs. It can also be described as the average revenue per acquired customer minus the average costs per acquired customer times the number of customers acquired. While the retention profitability, is the revenues generated by retained customers minus their costs. In other words the average revenues generated by a retained customer minus the average costs per retained customer multiplied by the number of customer retained. See equations 1 and 2. ) customers acquired of cost average -customers acquired of revenue (average * customers acquired of number portfolio customer total the of Value =
Equation 1: Acquisition profitability equation
) customers retained of cost average -customers retained of revenue (average * customers retained of number portfolio customer total the of Value =
Equation 2: Retention profitability equation
2.1.3 The Friction between Acquisition and Retention
Retention and acquisition are somewhat at odds with each other, besides the fact that firms have limited resources to divide among them (Rust, Lemon and Zeithaml, 2004). While acquisition and retention marketing both need very different approaches they are interrelated and should not be looked at in isolation.
To illustrate this point let us consider a hypothetical example. ‘Company X’ launches an aggressive acquisition campaign through multiple marketing communication channels to increase its acquisition rates. They set low introductory prices to entice as much prospects as possible to become their customer. ‘Company X’s’ aggressive low price acquisition campaign will most likely lead them to acquiring price sensitive customers with low profitability (initially). The moment a competitor drops their prices below theirs, or the firm increase their prices the newly acquired customers will most likely churn. These acquired customers flow into the customer base after their acquisition period. So the profitability of the acquired customers as well as the number that ‘survives’ the acquisition period directly influences the retention profitability in later periods. This thought experiment is backed up by the findings of Lewis (2002), who found that customers acquired with discounts have lower value than customer acquired without discounts.
2.2 Communication Mix
The range of marketing communication channels a business can utilize has rapidly increased over the past few decades. Where in the past there had been only a few so called ‘above-the-line’ channels like television, radio and print, combined with direct sales. These days thanks to modern technology, the different ways which we can use to communicate have increased and thus the communication channels marketing can utilize have increased as well. Especially the advent of the Internet has unlocked a multitude of new opportunities to reach customers, which businesses have been eager to exploit (Geyskens, Gielsen & Dekimpe, 2002). For an overview of the different channels used by the focal company see table 1. As can be seen, we classified these channels into 2 categories: Above the line and below the line.
There are numerous ways to define the difference between above-the-line and below-the-line marketing communication channels, based on reach (large vs. small), interactivity (not-interactive vs. interactive) or goals (image building vs. promotions). In this paper the channels have been classified on basis of reach. This is because channels can be used for multiple different goals; even though for ATL channels are most often used for image building below-the-line can be used for this goal also. Imagine for example a power company informing its customer of its new environmentally friendly power plant through a direct mailing.
2.2.1 Above-the-line
Above-the-line (ATL) communication channels are in essence mass media. These channels have a massive reach. Decades ago radio and later television commercials practically reached the whole population of a country. That is not the case anymore, since media consumption has fragmented in the last decades (nVision, 1979; Roy Morgan International, 2004). Television for instance is slowly losing ground to Internet, especially among younger generations (Carat, 2011). But still the reach of above-the-line is very large.
Television
These days television is the most important above-the-line marketing communication channel that large national or international businesses have. In the Netherlands television has a reach of almost 98%. While this reach is very large it is important to realise that the number of television channels has dramatically increased, so the reach of individual channels has decreased dramatically.
Radio
Radio has existed as a medium since the 19th century, with the establishment of the first radio stations in the 20th century radio advertising started in the early 1920’s. It dominated as the most popular mass media channel for most of the 20th century till it started losing ground to television after the Second World War. Radio is an important channel still capable of reaching around 90% of the Dutch population (Carat, 2011).
Dekimpe and Hanssens (1995) found a substantial net long-run sales effect for image-oriented television and radio advertising. Dertouzos and Garber (2006) also found positive effects of radio advertising on army recruits applications. According to their research radio has a wear in effect and takes 2 months to reach its peak influence, after that the effect starts to wear out till it completely dissipates after about 6 months. As mentioned earlier synergy effects were found between radio and television advertising (Radio Advertising Bureau, 2011).
Print might very well be the oldest communication channel listed here; of course it has existed for far longer than that it has been used for advertisement. A study analyzing the advertising expenditures of the American army on the influx of new recruits found that of the direct effect of print (magazines and newspapers) was much larger than that of radio or television but lagged and cumulative effects of advertising campaigns on radio and television were larger (Dertouzos & Garber, 2006). Arora (1979) studied print advertising and direct mail and found that advertising elasticities are dynamic and decrease over the product life cycle.
Sponsoring
2.2.2 Below-the-line
Below-the-line advertising has a much narrower reach and is more personal,
personalized or targeted at specific individuals. The main difference being the limited reach it has compared to ATL communication instruments.
Previous research asserts that when the products/services are characterized by high involvement decision making from the side of the customers, more involved and interpersonal contact channels have higher conversion rates than less involving contact channels (Anderson & Narus, 1999). The most striking example of this is door-to-door which has a very small reach (since a sales representative needs to physically travel to each customer), but a very high conversion rate. BTL
communication instruments tend to have higher response rates while they are more expensive per customer reached.
Direct Mail
Not all BTL communication instruments are interpersonal though. Direct mail for example can be personalized thanks to modern databases but it does not allow for interpersonal contact. Regardless of that fact direct mail is still very popular with business, mainly because of the low costs associated with it. Even though one would expect direct mail to lose in popularity with the advent of more modern
communication channels like e-mail and social media direct mail expenditures still grow each year in the Netherlands (Carat, 2011). Wiesel, Pauwels and Arts (2011) found a positive direct effect for direct mail in their study and synergy effects with online channels. In the study by Arora (1979) direct mail has a positive effect that diminishes over the course of a products life cycle.
Door-to-Door
It should be noted however that Anderson and Narus (1999) found that conversion rates for more interpersonal channels is higher than for channels with less
interpersonal contact. This is backed up by the continued use of door-to-door as a channel by for example energy companies. A recent study of some interest is that of Sargeant & Hudson (2008) which explore the reasons for donors of charities that have been recruited through door-to-door to churn. Since this study deals with non-profit organizations not all findings are useful but one striking conclusion is that conversion seemed to be higher for charities that scored high on brand recognition. This might mean that businesses with a high brand recognition have a better conversion as well.
Telemarketing
Telemarketing encompasses the outbound sales by a company to existing or potential customers. This is mainly focused on add-on selling, up-selling and acquisition. The available literature on telemarketing is very sparse, the study by Coppett and Staples (1993) notes that only 10 studies had been published on telemarketing in the top 10 marketing journals in the past 10 years. According to the same paper telemarketing generates $435 billion in sales each year so we can safely assume that as a marketing channel to boost sales it is fairly successful.
AdWords
AdWords is the pay-per-click advertising scheme from Google, advertisers pay for each customer that clicks their advertisement that shows up for certain searches in Google’s search engine. Since the advertisement is directed only at consumers or businesses that are actively searching for certain information, products or services it should be more effective than regular advertising.
2.3 General Effects of Marketing Communications
As discussed in the previous chapter the channel mix influences customer value in a multitude of ways and over a period of time. Although there is some debate on the exact nature of the influences and the period in which it exerts its influence. In this chapter the general effects of marketing channels are discussed, the chapter concludes with a summary of the known synergies between marketing communication channels. There are certain effects and characteristics of effectiveness that seem generalizable over most marketing communication channels. An overview of the papers that will be discussed in this paper is presented in table 2 on page 24.
Dynamic Effect
Effects of marketing communications are not static, i.e. they do not stay the same over a period. Often they change over a period of time or they change due to the changing nature of the product (life cycle) or customers using the product. Multiple studies found that advertising elasticities are dynamic and decrease over the product life cycle (Arora, 1979; Parker & Gatignon, 1996; Parsons, 1975). This is in line with the findings of Winer (1979) with regard to carryover effects, although he found that current effects increased, but that was a specific quality of the product type he analyzed (medicine). This seems logical since markets reach a level of saturation, and it is predominantly unknown new brands that benefit the most from exposure. As the brand awareness increases their sales will increase till they have reached the majority of potential customers. From that point onward the effects of advertising will decrease.
Duration
Price sensitivity
Price sensitivity is a very important factor for a company since it strives to attract customers with a low price sensitivity to increase revenues. Mela et al. (1997) found that advertising makes consumers less price sensitive and increased their loyalty, but he found promotions to increase price sensitivity, especially of non-loyal customers. The conclusion that advertising leads to lower price sensitivity was also reached by Lambin (1976).
Synergy
Type of Effect Effect Channels Industry Authors Advertising elasticities are dynamic and decrease over the product
life cycle Print, Direct mail Pharmaceuticals Arora (1979)
Advertising has persistent effect on sales in evolving markets Television, Radio, Print FMCG Dekimpe & Hanssens (1995) Advertising and sales relationships are different over all channels Television, Radio, Print Government Dertouzos & Garber (2006) Sales are only increased in 33% of the time for established brands Television - Lodish et al. (1995)
Advertising elasticities are dynamic and with an order of entry
effect Aggregated advertising FMCG Parker & Gatignon (1996)
Advertising elasticities are dynamic and decrease over the product
life cycle Aggregated advertising FMCG Parsons (1975)
While in long run brands tend to be mean reverting in sub periods brands can systematically improve performance through
promotions and advertising
Aggregated advertising,
promotions FMCG Pauwels, Hanssens (2007) Commitment and interaction are important for donor retention Door-to-door, retention, sales Non-profit Sargeant & Hudson (2008) Advertising - sales relationship changes over time Television - Winer (1979)
Sales/acquisition & retention
Advertising only has transitory effect on increasing sales Television - Winer (1980) 90% of the effects of advertising dissipate between 3 to 15 months
Aggregated advertising -
Assmus, Farley & Lehmann (1984)
90% of the effects of advertising dissipate between 3 to 15 months
Aggregated advertising
FMCG, Luxury
goods Clarke (1976) Show wear-out effect of advertising on sales Aggregated advertising - Bass & Leone (1983) Advertising has a wear-in of a week and wear-out of 3 weeks Aggregated advertising FMCG Pauwels (2004)
Duration
Most of the effects of advertising dissipate between 6 to 9 months Aggregated advertising - Leone (1995)
Synergy between television and print advertising Television, Print - Naik & Raman (2003) Combining sponsoring with television advertising leads to synergy
effects in buying intention and liking (brand level) Sponsoring, Television Apparel Olson & Thjømøe (2009) Synergy between television and radio advertising Radio, Television - RAB (2011)
Synergy
Cross channel effects for offline and online marketing efforts
AdWords, direct marketing Office furniture
Wiesel, Pauwels, Arts (2011)
Advertising leads to lower price sensitivity Aggregated advertising FMCG Lambin (1976)
Price sensitivity
Advertising reduces price sensitivity and promotions increase price sensitivity
Aggregated advertising,
Promotions FMCG
Mela, Gupta & Lehmann (1997)
Advertising does not produce barriers to entry
Aggregated advertising
FMCG, Luxury
goods Ayanian (1975) Advertising influences customer satisfaction Aggregated advertising, sales
force FMCG Baidya &Basu (2007) Advertising has positive effect on brand equity while promotions Aggregated advertising,
3. Research Methodology
The analysis in this paper will use quantitative data that is present within The focal company in their various databases. Although The focal company operates on both the business-to-business (B2B) market and the business-to-consumer (B2C) market, this paper focuses on the B2C market. This also includes a small businesses segment since they are grouped together within the business to consumer department of The focal company.
Customer and marketing channel data over the period 2005 to 2010 will be collected from these databases. Statistical analyses will provide insight in the data and eventually a descriptive model that shows the precise insight in the strength and direction of the relationships between the various communication channels and key customer metrics. In table 3 an overview of the variables is presented.
3.1 Data Collection
The data needed for the analysis is available in the various databases of The focal company, however certain transformations had to be performed and new variables had to be calculated. This was all done in The focal company’s data system with the use of SQL. The model itself was estimated in the statistical program EViews.
The data collection has lead to two different data sets, one containing the marketing expenditures of The focal company, and one containing the aggregated data on The focal company customers. Both data sets need cleansing in order to become useful for the analysis.
To illustrate this let us consider this example: a company reserves €100.000 for direct mailings (DMs) in June and July ($200.000 in total). But due to an unfortunate product proposition the marketing campaign that the direct mailings are supporting completely fails to attract customers. The campaign manager then decides to pull the plug out of the marketing campaign at the beginning of July, saving 80% of the €100.000 DM budget in July in variable costs. The €80.000 that was not spent is booked as a negative expenditure in the next months to even out the books.
Of course outside of the technical world of accounting a negative expenditure does not exist and these data points can not be used in the model. Since it is unclear in which periods these negative expenditures were booked they have been allocated at the discretion of the researcher. Each negative expenditure has been subtracted of the highest expenditure in previous periods and the month in which the negative
expenditure was booked was transformed to zero.
Data Aggregation
The level of aggregation of the data is extremely important for detecting trends and avoiding biases. The data used in this paper is aggregated over geographical areas (no distinction between regions is made), thus it is implicitly assumed that advertising intensities are constant over space. The data is also aggregated over time, where an interval is one month; this is done because advertising decisions or budget allocations are done per month, thus making this the optimal interval for estimating a market response model (Dertouzos & Garber, 2006). The data is also aggregated over customers, thus it does not acknowledge any differences in advertisement effects that channels might have on individual customers or segments of customers.
Variable Operationalization
Radio / Television Marketing expenditure per month on radio and television Printed Media “ “ “ “ “ “ “ “ “ Printed Media
Direct Mail “ “ “ “ “ “ “ “ “ Direct Mail Door-to-Door “ “ “ “ “ “ “ “ “ Door-to-Door Outbound Sales “ “ “ “ “ “ “ “ “ Outbound Sales
Online “ “ “ “ “ “ “ “ “ on website and other online campaigns AdWords “ “ “ “ “ “ “ “ “ AdWords
Value added services “ “ “ “ “ “ “ “ “ on VAS campaigns Sponsoring “ “ “ “ “ “ “ “ “ Sponsoring
ATL Marketing expenditure per month on radio and television plus printed media plus sponsoring
Marketing communication
BTL Marketing expenditure per month on direct mail plus door-to-door plus outbound sales
Acquisition New customers compared to previous month Retention Number of customers retained
Churn Number of churned customers
Cost of acquisition Total costs that acquired customers cause (cost = cost of inbound contact costs plus costs of reminders, dunning and collections)
Cost of retention Total costs that retained customers cause (cost = cost of inbound contact costs plus costs of reminders, dunning and collections)
Average cost of new customers
Average cost of acquired customers (cost = cost of inbound contact costs plus costs of reminders, dunning and collections)
Average cost of retained customers
Average cost of retained customers (cost = cost of inbound contact costs plus costs of reminders, dunning and collections)
Average revenue of
new customers Average revenue of acquired customers Average revenue of
retained customers Average revenue of retained customers Average profit of new
customers Average revenue minus the average cost of acquired customers Average profit of
retained customers Average revenue minus the average cost of retained customers CV Acquisition New customers compared to previous month times their value
CV Retention Total customer value compared to previous month minus the acquisition times their value
Key customer metrics
CV Total Total customer value
Share of voice Share of voice Marketing expenditures (above-the-line) per month of The focal company indexed against the competitors (Essent / Eneco / Oxxio) total expenditures Price index Price index Price of regular The focal company contract compared to average price of
3.2 Research Method
As a starting point for the analysis the paper by Wiesel et al. (2011) will be used. In this paper a relatively similar analysis was performed on online and offline sales funnel progression of prospects of a Dutch business-to-business furniture manufacturer/seller. Wiesel et al. (2011) used a vector auto-regression (VAR) model as analysis tool in this paper; although other models could be an option (Baidya & Basu, 2008).
3.3 Plan of Analysis
A detailed plan of analysis is presented in the table 4.
Methodological step Relevant Literature Research question
1. Descriptives
2. Granger causality tests Granger (1969) Trusov et al. (2009)
Which variables are temporally causing which other variables?
3. Unit root and co-integration tests Augmented Dickey-Fuller test Zivot-Andrews test
Enders (2004)
Zivot and Andrews (1992) Johansen et al. (2000)
Are variables stationary or evolving? Are unit root results robust to unknown breaks?
Are evolving variables in long-run equilibrium?
4. Model of dynamic interactions Vector Autoregressive model
Dekimpe and Hanssens (1999) Bronnenberg et al. (2000) Pauwels et al. (2007)
How do performance and marketing interact in the long run and short run, accounting for the unit root and co-integration results? 5. Policy simulation analysis
Unrestricted impulse response
Pesaran and Shin (1998) Pauwels et al. (2002) Pauwels (2004)
What is the net dynamic impact of a marketing change on performance? What is the direct dynamic impact of a marketing change, controlling for its indirect effects?
6. Validation analysis VAR lag specification Regression analysis
Ventzislav and Lutz (2001) Wiesel et al. (2011)
Are the results robust to the lag selection criterion?
Do the key results replicate in regression analysis?
3.5 Descriptives
The charts in this sub-chapter show the development of acquisition, churn and the "profit" of acquired and retained customers from January 2005 up to December 2010.
As figure 2 shows there seems to be a slight upwards trend in the total number of customers acquired per month. While the churn (figure 3) seems to be relatively stable albeit higher than the acquisition, which is logical since The focal company is slowly losing market share over the past few years.
Churn 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 Acquisition 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 1 3 5 7 9 11 13 15 1719 21 23 25 27 2931 33 35 37 39 41 43 45 47 49 51 5355 57 59 61 63 65 6769 71
Figure 2: Acquisition in total customers
The profit of acquisition however is shows a clear upward trend (figure 4), this is contrary to the profit of retained customers which seems to be stable but has a slight drop off in the last half year of 2010 (figure 5).
Profit_Acquisition 0 10000 20000 30000 40000 50000 60000 70000 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 Profit_Retention 5200000 5400000 5600000 5800000 6000000 6200000 6400000 6600000 6800000 7000000 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71
The graphs of the Above-the-line and Below-the-line expenditures can be found in the appendices 1 and 2.
Figure 4: Profit of acquisition
3.6 Granger Causality
While regular regressions simply model correlations between multiple variables time series can be used to show ‘temporal causality’. Variable X is said to Granger-cause variable Y if past values of variable X contain information that helps predict variable Y beyond the information that is contained in the values of Y alone. More formally it is said that variable X Granger-causes Y if the mean squared forecast error (MSFE) of a model containing past values of X and Y is smaller than the MSFE of a model containing only past values of Y (Leeflang et al. 2000; Hanssens et al. 2003).
A Granger causality test is used to test the null hypothesis “variable X does not Granger-cause variable Y”. A significant outcome rejects the null hypothesis, in this paper the statistical software package EViews is used to test for Granger causality. Outcomes that have a p-value below 0.05 are statistically significant on a 95% confidence interval; outcomes that have a p-value below 0.10 are statistically significant on a 90% confidence interval.
Granger-Causes Overview
The most interesting Granger-causes are discussed below; the number behind the Granger-cause indicates where it can be found in appendix 3, which shows the complete set of Granger-causalities that were tested. In appendix 3 the correlation (Corr.) indicate the direction and strength of the relation between the variables; green coloured lags indicate that the lag is significant on a 95% confidence interval, while yellow coloured lags indicate the lag is significant on a 90% confidence interval. For an overview of the most striking Granger-cases see table 5. Which can be read in the same manner as appendix 3.
Probability per Lag
Null Hypothesis: Corr. 1 2 3 4 5 6 7 8 9 10 11 12 13
ACQUISITION does not Granger cause
RETENTION -0,61 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,01
D2D does not Granger cause ACQUISITION 0,73 0,04 0,05 0,28 0,11 0,27 0,47 0,51 0,59 0,57 0,69 0,61 0,28 0,38
Direct mail does not Granger cause ACQUISITION 0,39 0,43 0,09 0,04 0,01 0,01 0,01 0,02 0,02 0,03 0,00 0,00 0,00 0,01 Direct mail does not Granger cause the value of
acquired customers 0,31 0,03 0,00 0,02 0,02 0,04 0,06 0,04 0,06 0,10 0,08 0,12 0,06 0,05
Direct mail does not Granger cause Revenue of
acquired customers 0,39 0,52 0,10 0,06 0,02 0,02 0,02 0,03 0,03 0,07 0,00 0,00 0,01 0,03
Price index The focal company does not Granger
cause average cost of retention -0,27 0,53 0,30 0,52 0,64 0,71 0,67 0,41 0,03 0,04 0,02 0,00 0,01 0,00
Telemarketing does not Granger cause
ACQUISITION 0,64 0,03 0,11 0,19 0,33 0,37 0,29 0,32 0,08 0,07 0,09 0,01 0,00 0,01
WEBSITE/ONLINE does not Granger cause
ACQUISITION 0,49 0,01 0,01 0,04 0,07 0,23 0,01 0,02 0,05 0,09 0,05 0,01 0,03 0,05
WEBSITE/ONLINE does not Granger cause the
value of acquired customers 0,48 0,00 0,01 0,05 0,11 0,17 0,13 0,26 0,39 0,52 0,35 0,16 0,14 0,05
WEBSITE/ONLINE does not Granger cause
Revenue of acquired customers 0,48 0,00 0,00 0,01 0,02 0,06 0,00 0,00 0,01 0,03 0,02 0,01 0,02 0,03
WEBSITE/ONLINE does not Granger cause
Revenue retention -0,37 0,01 0,05 0,11 0,13 0,14 0,39 0,73 0,66 0,75 0,79 0,87 0,93 0,96
Table 5: Most Striking Granger-causes
Note: (Corr.) indicate the direction and strength of the relation between the variables; green coloured lags indicate that the lag is significant on a 95% confidence interval, while yellow coloured lags indicate the lag is significant on a 90% confidence interval.
Of all the marketing communication channels, telemarketing, direct mail, door-to-door and online seem to deliver the best results judging by the Granger causes. Especially online and direct mailings seem to have exert influence over longer
periods, which could be explained by the fact that the door-to-door and telemarketing are more like sales pitches than marketing. Another striking conclusion that can be drawn from the results in table 5 is that online and direct mailing seem to influence the value of new customers, thus these channels seem to attract more valuable customers. Lastly the price index of The focal company is negatively related to the average cost of retained customers, thus higher prices seem to attract customers that cause less costs or expel customers that cause more costs.
Pairwise Granger Causality Tests
Probability
Null Hypothesis: Corr. Lags: 1 Lags: 2 Lags: 3 Lags: 4 Lags: 5 Lags: 6 Lags: 7 Lags: 8 Lags: 9 Lags: 10 Lags: 11 Lags: 12 Lags: 13
ATL does not Granger Cause
ACQUISITION -0,11 0,75 0,46 0,65 0,76 0,89 0,87 0,9 0,96 0,61 0,52 0,57 0,45 0,53
ATL does not Granger Cause BTL 0,02 0,94 0,05 0,04 0,14 0,18 0,14 0,13 0,07 0,06 0,16 0,24 0,24 0,06
ATL does not Granger Cause CHURN -0,23 0,2 0,28 0,44 0,69 0,88 0,95 0,95 0,87 0,79 0,68 0,67 0,83 0,72 ATL does not Granger Cause
PROFIT_ACQUISITION -0,09 0,32 0,45 0,7 0,89 0,96 0,81 0,49 0,59 0,42 0,39 0,58 0,58 0,61
ATL does not Granger Cause
PROFIT_RETENTION 0,04 0,04 0,11 0,18 0,27 0,45 0,59 0,4 0,53 0,64 0,65 0,74 0,82 1
BTL does not Granger Cause
ACQUISITION 0,83 0 0,01 0,03 0,04 0,02 0,01 0,03 0,04 0,07 0,14 0,15 0,1 0,26
BTL does not Granger Cause ATL 0,02 0,15 0,19 0,43 0,59 0,67 0,63 0,59 0,62 0,42 0,63 0,68 0,72 0,7
BTL does not Granger Cause CHURN 0,26 0,23 0,59 0,02 0,02 0,01 0,02 0,04 0,03 0,04 0,05 0,03 0,05 0,08 BTL does not Granger Cause
PROFIT_ACQUISITION 0,8 0 0,01 0,04 0,09 0,16 0,27 0,38 0,48 0,59 0,59 0,54 0,48 0,52
BTL does not Granger Cause
PROFIT_RETENTION -0,01 0,36 0,63 0,56 0,73 0,35 0,51 0,59 0,69 0,58 0,66 0,75 0,73 0,78
Table 6: Most Striking Granger-causes of the additional aggregated variables
Note: (Corr.) indicate the direction and strength of the relation between the variables; green coloured lags indicate that the lag is significant on a 95% confidence interval, while yellow coloured lags indicate the lag is significant on a 90% confidence interval.
In the table above the Granger-causes of the additional aggregated variables are discussed. The profit of acquisition/revenue variables are revenues of
acquisition/retention minus costs as presented in table 3 on page 28. The ATL (above-the-line) and BTL (below-(above-the-line) variables are the summed marketing
communication expenditures of ATL and BTL respectively.
In table 6 most striking is that the BTL variable seems to have the most and longest duration of effects. BTL positively influences acquisition, although is regrettably also positively influences churn. But the largest effects by far (highest correlation) are on the ‘profit of acquisition’ and acquisition, which is precisely as expected.
3.7 Unit Root
A time series can either be stationary or evolving. Stationarity implies that the time-series reverts to its mean, while evolving time-time-series show never ending changing behaviour. A unit-root test tests whether a time-series has a short-run changing processes (eventually mean-reverting) or long-term changing process, i.e. evolving. An evolving time series is said to have a unit root, the root of their characteristic equations lie outside of the unit circle. In practice this means that a time series will wander widely from any previously held position because its probability distribution changes over time. This shows itself in the time series by persistent, long-run
movement called trends or a time series can be unstable over time and display breaks. Either case of non-stationarity makes it impossible to accurately predict future values of it (Leeflang et al. 2000; Hanssens et al. 2003; Stock & Watson, 2011).
Augmented Dickey-Fuller Test
The test used to determine whether the variables in the data set are evolving (has a unit root) or are stationary is called augmented Dickey-Fuller (ADF) test. In the statistical program Eviews an overall indication of stationarity of the data set and the stationarity of individual variables is calculated.
As can be seen in the results of the ADF tests in appendices 4 and 5 the overall significance of the tests for the data set is significant while not all the individual variables are significant. The variables, 'acquisition', 'profit acquisition', 'BTL', 'online' and the PI The focal company (price index) are all evolving.
3.8 The Model
A vector-autoregressive (VAR) model is estimated using above-the-line and below-the-line marketing metrics (spending), customer metrics and returns. The models that will be utilized in this paper will be standard VAR and a slight modification from a standard VAR model, the VARX model. The ‘X’ stands for an added vector that incorporates exogenous variables which add to the otherwise endogenous variables of the model. A number of different models will be estimated to determine which has the best explanatory strength and best fit.
3.8.1 Endogenous vs. Exogenous
Usually endogenous variables are internal variables which can be controlled, while exogenous variables represent uncontrollable external factors. These terms are used differently in the context of VAR models. A regular VAR model treats all the variables in it as endogenous variables, this means that every variable influences all the other variables and is influenced by all other variables. A VARX model adds a vector that contains exogenous variables; these variables influence endogenous variables but are themselves not influenced. Thus all endogenous variables can be regarded as dependent variables while exogenous variables are independent variables.
This construct is very helpful in situations where one wants to incorporate variables that exert control over endogenous variables but are themselves not affected by the endogenous variables. Missing data points are less of a problem for exogenous variables since they only influence other variables and do not need to be explained themselves by other variables. Therefore they do not add as much coefficients that need to be estimated as endogenous variables (and therefore need less data points).
3.8.2 Notation
Equation 3 represents the model in simplified matrix form while the whole equation is shown in appendix 6.
T
t
U
Y
B
C
Y
t i i i t i t,
1
,
2
,...,
1=
+
+
=
∑
= −Equation 3: Model Notation
Yt represents the vector of endogenous variables. C represents the vector of intercept. Bi represent vectors of coefficients. Yt-i represents the vector of lagged endogenous
4. Model Selection
Since the number of relevant variables is quite large and the number of data points available per variables is quite small, a selection of variables had to be used in the estimation of the models. Especially endogenous variables require a large number of data points to be calculated correctly, thus limiting the models in the number of endogenous variables (or lags) that can be used. To tackle this problem a number of different models were estimated in the search of the best fitting models. These models vary in the number of lags included and which/how the different variables were included (endogenous or exogenous).
Endogenous variables have more interconnected relations with other endogenous variables, therefore each included endogenous variable needs more data points than an exogenous variable. Due to the limited number of data points in the data set, this leads to a restriction on the number of lags that can be estimated for a model that
incorporates more endogenous variables. An important drawback of using exogenous variables is that without added interaction variables any interactions between
marketing variables are not shown in the model. This can be remedied by adding additional interaction variables.
4.1 Statistical Problems
In the original intended model all the different marketing communication channels were included separately and compared to the average revenues and costs of
The value of acquisition and retention variables were included in many different versions, as totals, averages and percentages (costs and revenues were included for retention and acquisition separately, see page 28). The quantity of acquisition and retention variables were straightforward the number of acquired and retained
customers. Lastly the marketing communication channel variables were included as expenditures per channel.
The first models that were estimated suffered of statistical violations and counter intuitive effects. The signs of multiple variables were opposite of what was expected for example, this can of course be explained by the statistical violations of the models. Therefore a number of adjustments needed to be made in the variables and models.
4.2 Solutions
The value of acquisition and retention variables. To bring back the number of variables used in the model only one 'profit' variable for acquisition and retention is used instead of the a revenues and costs variable for acquisition and retention. The quantity of acquisition/retention variables. Instead of including the variable 'retention', the number of churned customers was included in the model since the 'retention' variable was causing major statistical problems.
The marketing communication channel variables. Due to the low amount of data points available using all the marketing communication channels as separate variables was unwieldy and prone to statistical violations. This has lead to the decision to sum all the above-the-line marketing communication channels into one variable: 'ATL'. Similarly all the below-the-line marketing communication channel variables were summed into one variable: 'BTL'.
4.3 Lag specification
Since model 5 was deemed the most appropriate model to estimate the influence of the different marketing communication channels on the various acquisition and retention variables, this model is used for the estimations. Next we examine the optimal lag length.
1 Lag 2 Lags 3 Lags
Akaike information criterion 146.179 146.578 146.839
Schwarz criterion 147.550 149.145 150.621
Table 7: Information criterions overview
Note: Low numbers indicate a better fit
5. Estimation Results
The selected model is completely endogenous, therefore the estimation and all the accompanying tests for statistical validity can be done in EViews.
5.1 System relations
In the table 8 the relations between the endogenous system is presented. A plus sign indicates a positive relation while a minus indicates a negative relation; the green filling indicates it is significant. The complete results of the estimation of the VAR model can be found in appendix 8.
Churn Acquisition Profit retention Profit acquisition ATL BTL Churn - + + + + + Acquisition - - - + + + Profit retention - - - - Profit acquisition - - + - - - ATL - - + - - + BTL + + + + - -
Table 8: Accumulative impulse response functions overview
Note: Green coloured signs indicate significance at a 95% confidence interval
As the table above shows, acquisition (in total customers) has a negative effect on churn (in total customers) or a positive effect on retention as you will. This result is contrary to the result found with the correlations (see Granger-causes), this is because the correlations are at level. Incorporating the first lag into the model changed the sign. A negative relation could be due to the fact that current customers react on campaigns designed for new customers, this is a way for them to get cheaper energy contracts. It is also not uncommon that customers who indicate that they are
-100000 -50000 0 50000 100000 150000 200000 250000 300000 1 2 3 4 5 6 7 8 9 10
Accumulated Response of D(PROFIT_RETENTION) to Cholesky One S.D. D(ATL) Innovation
lines of what was expected. It clearly shows that above-the-line marketing
communication channels have a positive influence on retention and increasing the profitability of retained customers.
Below-the-line marketing communication channels on the other hand have a positive influence on the total number of customers that The focal company acquires and the profitability of those customers.
5.2 Impulse Response Function
In an impulse response function the constructed system (model) is shocked through increasing one variable with one standard deviation and calculating the responses of the other variables in the system. In this case we are interested in how the retention and acquisition variables respond to increases in the marketing communication channel variables (ATL and BTL). Figure 6 shows the accumulated responses for the effect an increase in spending on above-the-line marketing communication channels have on churn. Figure 7 shows the accumulated response of the 'profit of retention' to an increase of spending on above-the-line marketing communication channels.
Striking is the similarity between the effects in figures 6 and 7 are although mirrored in the x-axis. The effects in both cases seem to wear out around 6 months were it
-1200 -800 -400 0 400 1 2 3 4 5 6 7 8 9 10
Accumulated Response of D(CHURN) to Cholesky One S.D. D(ATL) Innovation
-200 0 200 400 600 800 1000 1 2 3 4 5 6 7 8 9 10
Accumulated Response of D(ACQUISITION) to Cholesky One S.D. D(BTL) Innovation -2000 0 2000 4000 6000 8000 1 2 3 4 5 6 7 8 9 10
Accumulated Response of D(PROFIT_ACQUISITION) to Cholesky One S.D. D(BTL) Innovation
This is in line with the findings of Leone (1995), Assmus et al. (1984) and Clarke (1976). They found that advertising effect wear out between 6 to 9 months and 3 to 15 months respectively.
In the figures 8 and 9 the impulse response functions of acquisition and the profit of acquisition to an innovation of one standard deviation in below-the-line marketing communication expenditures is presented.
Expenditures on below-the-line marketing communication channels has a positive effect on acquisition and the profit from acquisition. The effects are strongest in the first few periods, seemingly levelling off completely after around 6 months. Again in the line of the findings of Leone (1995), Assmus et al. (1984) and Clarke (1976).
In conclusion the effects of ATL and BTL marketing communication channels are in line of the expectations, both seem to have larger effect in the first few months while the effects dissipate completely after 6 months. For an overview of the non
cumulative impulse functions see appendix 9.
5.3 Model Validation
To test the statistical validity of the model a number of different tests are run. This is done to ensure that the data and residuals are 'normal', if this would not be the case than that would mean that incorrect conclusions would be drawn from other tests, or that wrong estimates of model parameters are obtained.
First of all the Jarque-Bera test, which is a goodness-of-fit test of whether sample data have the skewness (symmetry) and kurtosis (peakedness) matching a normal
distribution.
*Component Jarque-Bera df Prob.
1 1.987 2 0.370 2 4.071 2 0.131 3 0.879 2 0.644 4 2.436 2 0.296 5 1.443 2 0.486 6 2.392 2 0.302 Joint 13.207 12 0.354
Table 9: Jarque-Bera test
*the components indicate the variables
The Jarque-Bera test for normality should be insignificant to indicate that the kurtosis and skewness is normal, as table 9 indicates this is the case for out model. Thus we can conclude that the data is normally distributed.
Secondly the Breusch–Godfrey serial correlation Lagrange multiplier test is
Lags LM-Stat Prob. 1 37.304 0.409 2 38.873 0.342 3 36.287 0.455 4 46.343 0.116 5 36.056 0.466 6 39.590 0.313 7 29.411 0.773 8 61.211 0.006 9 29.028 0.789 10 20.682 0.981 11 43.001 0.196 12 28.049 0.826 Prob. from chi-square with 36 df. Table 10: Breusch–Godfrey test
The null-hypothesis of the the Breusch–Godfrey test is that serial correlation is not present. The results of the test in table 10 shows that there is no serial correlation present in the data at the 1 lag (serial correlation shows up at lag 8 which is irrelevant since the model does not incorporate that many lags).
Lastly the Portmanteau tests autocorrelation in the residuals of the model. Auto-correlation would indicate a cross-Auto-correlation in the residuals and thus an underlying pattern, which could indicate missing variables, lags or an incorrect form. The null-hypothesis of the test is that no residual auto-correlation is present.
Lags Q-Stat Prob.
Adj
Q-Stat Prob. df
1 10.890 NA* 11.053 NA* NA*
2 44.165 0.165 45.336 0.137 36 3 78.900 0.27 81.674 0.204 72 4 122.307 0.164 127.794 0.094 108 5 155.699 0.239 163.836 0.123 144 6 191.836 0.259 203.470 0.111 180 7 219.02 0.430 233.774 0.194 216 8 272.156 0.1830 293.994 0.036 252 9 298.035 0.330 323.821 0.072 288 10 316.048 0.614 344.940 0.203 324 11 353.786 0.582 389.960 0.133 360 12 377.408 0.741 418.644 0.208 396 Table 11: Breusch–Godfrey test
6. Market Simulations
The estimations of the VAR-model in the previous chapter are used for making a market simulation. Table 12 and 13 present overviews of the market simulation for investments made in above-the-line and below-the-line marketing communication channels. The returns on investment in ATL and BTL marketing communication channels are based on an investment of one standard deviation (inter- and
extrapolations are made to 75% and 125% of one standard deviation investment). To calculate the revenues an investment of one standard deviation in ATL and BTL would produce, the value of customers that were prevented from churning or gained plus the overall increase in the acquisition and retention profit variables is calculated. The number of customers lost or gained is multiplied with the estimated value of a single customer, an estimated retention rate and depreciated for future years. The total value of these acquired or retained customers is summed over 10 years, which seems like reasonable horizon (within the focal company an horizon of 25 years is normally used).
All these factors combined give an indication of the net impact of an investment in one marketing communication channel. It should be noted that the total effects calculated are ceteris paribus, which is of course not a realistic view of reality.
[confidential information: can be requested from the author]
Table 12: Market simulation outcomes for investment in ATL
[confidential information: can be requested from the author]
As can be seen in the tables above ATL delivers the greatest return on investment, the initial investment is higher but it also retains more customers than BTL acquires. In total an investment of 1 standard deviation in ATL would deliver a factor [confidential information: can be requested from the author] return on investment; and an investment of 1 standard deviation in BTL a factor [confidential information: can be requested from the author]
7. Conclusion
This paper set out to find out “How above-the-line and below-the-line marketing efforts of The focal company influence the total value of their customer portfolio”. How acquisition and retention interact; if any synergies exist between different
marketing communication channels; and what the result of an increase/decrease of x%
/ €x in any of the marketing investments on the customer value, and vice versa?
In the definitive model that was estimated the interactions between acquisition and retention were mostly insignificant, except for the negative effect acquisition had on churn. This is contrary to the correlation found earlier. The negative relation could be explained by fact that current customers react on campaigns designed for new
customers, this is a way for them to get cheaper energy contracts. It is also not uncommon that customers who indicate that they are contemplating switching to be offered deals meant for new customers.
Of all the marketing communication channels, telemarketing, direct mail, door-to-door and online seem to deliver the best results judging by the Granger causes. But due to statistical problems with the earlier models individual marketing
communication channels could not be modelled. So the focus shifted to the overall impact of above-the-line and below-the-line marketing communication channels.
As table 8 (page 42) and the figures 6 to 9 (pages 43-44) show, spending on above-the-line (ATL) marketing communication channels lowers churn and increases the profitability of retained customers. While spending on below-the-line (BTL) marketing communication channels increases acquisition and the profitability of acquired customers. The effects of the expenditures seems to be concentrated in the first few months for both ATL and BTL, with the effect largely dissipating after around 4 months for both.
churn. The market simulation showed that an investment in ATL would deliver a factor [confidential information: can be requested from the author] return on investment over 10 years ([confidential information: can be requested from the author]); and an investment of 1 standard deviation in BTL a factor [confidential information: can be requested from the author]
8. Managerial Implications
Management should be aware of the fact that their company resources are limited, especially when it is deliberately limited by the management in this case. This has very direct consequences on the viability of programs focused on the retention and acquisition of customers. The tension between acquisition and retention is very real and both aspects of the business should be approached holistically and not separately. The focus for either retention or acquisition should be on the value and not the
number of customers.
The management of the focal company should focus more on raising the value of customers and less on increasing sales. This presents a certain conundrum since the above-the-line and below-the-line marketing communication channels seem to do this. ATL increases the profitability of retention while BTL increases the profitability of acquisition. In the light of the marketing simulation ATL marketing communication channels are found to add the most value, therefore these channels should be used more prominently in the marketing communication mix of the focal company.
9. Limitations and Future Research
The most obvious limitation of this research is the small data set; the current data set used data from 2005 to 2010. This can be easily remedied in future researches since the data is available from 2005 onwards. Another important point is the manner in which the expenditures on the various channels is recorded, this is obviously done from an accounting perspective rather than from an marketing perspective. It would help further research if the expenditures on individual channels were better recorded, instead of multiple channels lumped together on the basis of their campaign. Also perhaps ‘real’ expenditures could also be registered besides just the budgeted expenditures; this would solve the problem of ‘negative’ expenditures.
Due to statistical problems, perhaps related to the issue addressed in the paragraph above the marketing channels could be modelled individually, which is a significant loss for the focal company since it would have given a great indication which marketing communication channels are most effective. To make the model more practical competitors should also be modelled, due to the limited data set it was not possible to model competitors accurately in the final model in this paper.
10. References
Anderson, J. & Narus, J.A. (1999), Business Market Management. Upper Saddle River, NJ: Prentice Hall
Arora, R. (1979), “How Promotion Elasticities Change,” Journal of Advertising Research, Vol. 19, pp. 57-62.
Baidya, M.K. & Basu, P. (2008), "Effectiveness of marketing expenditures: A brand level case study", Journal of Targeting, Measurement and Analysis for Marketing, Vol. 16, Issue. 3, pp. 181–188.
Berger, P.D. & Bechwati, N.N. (2001), "The allocation of promotion budget to maximize customer equity", Omega, Vol. 29 Issue 1, pp. 49-61.
Blattberg, R.C. & Deighton, J. (1996), “Manage Marketing by the Customer Equity Test,” Harvard Business Review, Vol. 74, pp. 136–44.
Blattberg, R.C. Getz, G. & Thomas, J.S. (2001) "Customer Equity: Building and managing relationships as valuable assets", Harvard Business School Press, Boston Massachusettes.
Bronnenberg B., Mahajan & Vanhonacker, V.W. (2000), “The Emergence of Market Structure in new Repeat-purchase Categories: the Interplay of Market Share and Retailer Distribution”. Journal of Marketing Research, 37(1) 16-31.
Carat (2011), “Mediafeitenboekje Nederland 2011”.
Coppett, J.I. & Staples, W.A. (1993), "Telemarketing: the Dark Continent", Journal of Marketing Theory & Practice, Vol. 1 Issue 3, pp.1-10.
Dekimpe, M.G. & Hanssens, D.M. (1995), “The Persistence of Marketing Effects on Sales”, Marketing Science, Vol. 14, Issue 1, pp. 1-21.
Dertouzos, J.N. & Garber, S. (2006), “Effectiveness of Advertising in Different Media: The Case of U.S. Army Recruiting”, Journal of Advertising, Vol. 35, Issue 2, pp. 111-122.
Enders, W. (2004), “Applied Econometric Time Series”, Wiley, NY.
Geyskens, I Gielsen, K. & Dekimpe, M.G. (2002), ”The Marketing Valuation of Internet Channel Additions”. Journal of Marketing, Vol. 66, pp. 102-119.
Granger, C.W.J. (1969), “Investigating causal relationships by econometric models and crossspectral methods”, Econometrica, 37(3) 424-438.
Hanssens, D.M. Parsons, L.J. & Schultz, R.L. (2003), Market Response Models: Econometric and Time Series Analysis, Kluwer Academic Publishers Group, Dorderecht. e.d.: Second Edition.
Johansen, S, Mosconi R. & Nielsen, B. (2000), “Cointegration analysis in the presence of structural breaks in the deterministic trend”, Econometrics Journal, 3 216–249.
InSites Consulting (2011), Social Media Around the World 2011, Available:
http://www.slideshare.net/stevenvanbelleghem/social-media-around-the-world-2011
