The difference between customer initiated touchpoints versus firm
initiated touchpoints and their synergy effects towards purchase
probability
by
Daan Hagen
University of Groningen
Faculty of Economics and Business
MSc Marketing
January 2020
Lage Trijnweg 6 9909 TD Spijk GN 0626869325 d.m.hagen@student.rug.nl Student number: S3274675 Supervisor 1: Peter van EckAbstract
The purpose of this study is to contribute towards customer journey attribution research - specifically customer initiated versus firm initiated touchpoints research - in two directions. First, the study examines whether there exists a positive relationship between customer initiated and firm initiated touchpoints and purchase probability, measured in four different ways: all customer initiated and all firm initiated touchpoints aggregated to two independent variables, measuring all touchpoints individually and finally the same two tests but then only for customers who purchased a product. Alongside the question whether relationships are existent, it will also be measured whether the impact of one type of the touchpoints (i.e. customer initiated or firm initiated) on purchase probability is larger than the other one. Second, this study examines to what extend there exists positive synergy effects between customer initiated and firm initiated touchpoints in their impact on purchase probability. This study is one of the first attempts to look into the differences between customer initiated and firm initiated touchpoints and partly resulted in outcomes not supporting the presumed relationships. Both customer initiated and firm initiated were found to have a positive impact on purchase probability, but firm initiated touchpoint had a greater impact on purchase probability (at the focus brand) than customer initiated touchpoints in both parts of the study (i.e. purchase not required as the first part and purchased required as the second part).
However, when the touchpoint were measured individually – instead of all customer initiated touchpoints combined and all firm initiated touchpoints combined – customer initiated touchpoints had a greater impact on purchase probability (at the focus brand) than firm initiated touchpoints in both parts of the study. Finally, the synergy effect turned out to be negative instead of positive. These findings can be contributing to future research and managerial decision-making, as it gives managers a better insight in how to deal with the distribution of resources when it comes to customer initiated and firm initiated touchpoints.
1. Introduction
The way(s) how customers purchase products or services has changed a lot during the last decades. Looking at modern society, it is hard to believe customers used to visit brick and mortar stores (i.e. traditional, physical stores) to fulfil all their shopping needs and
journey of an individual customer is the process of using all touchpoints over all (online) marketing channels prior to a potential purchase decision leading to a(n) (online) store visit. Touchpoints are events of contact – both direct and indirect – a customer has with a brand or a company (Baxendale et al, 2015).
Logically, the large amount of studies relating to these subjects is a sign of an increased focus on customer journey and customer touchpoints research in literature. However, following Anderl et al. (2016), despite the introduction of advanced attribution models, research focussed on channel effectiveness in a multichannel setting and the interplay of channels is still lacking. Their work was a first step in the right direction, but there is still uncertainty regarding the effectiveness of channels and touchpoints in a multichannel settings. Baxendale et al. (2015) also recognized the trend of measuring touchpoint effectiveness separately. Their paper measured the effects in a relative way, but as their used touchpoints - brand advertising, retailer advertising, in-store communications, word-of-mouth, peer observation and traditional earned media are different from this study and their focus was on offline purchases instead of online purchases, there is still a gap to fill. Baxendale et al. (2015) chose their touchpoints following an accumulation of prior research and chose offline instead of online because they found it more contributing due to the complexity of measuring offline customer journeys as customers tend to forget parts of their journey.
This study uses an online approach due to the usage of a database which was created before the start of this study, which is also the reason for the difference is used touchpoints compared to prior studies. The option to conduct additional research to build a dataset containing both new touchpoints and earlier tested touchpoints was considered, but it would have been hard to obtain longitudinal touchpoint usage of customers in the Dutch travel market (i.e. the market used in the existing database). Thus, the final database would have been a database consisting both longitudinal touchpoint usage of Dutch travel customers and non-longitudinal touchpoint usage of any possible customers in any possible market, which was expected to decrease the clarity and overall reliability of the results.
where the advertiser initiates communication between the customer and the firm (Anderl, Becker, Von Wangenheim & Schumann, 2016; Bowman & Narayandas, 2001; Wiesel, Pauwels, & Arts, 2011). There is a great amount of research about individual customer and firm initiated touchpoints effectiveness, but research about the effectiveness of combined customer and firm initiated touchpoints is still lacking. Rosenbaum et al. (2017) stated that in existing research, there barely is research that distinguishes customers of a particular
organization in terms of their used touchpoints and their view – in terms of importance – of these touchpoints. They suggest gathering customer information and identifying importance of touchpoints as a solution by directly asking them, even though that approach is time consuming. Their own data collection was conducted through self-administered
questionnaires, meaning that they used that approach as well. This study uses a dataset of used touchpoints by customers over time, which is therefore passively measured data. Asking customers directly gives prescribed outcomes which might be different from actual customer behaviour, whereas this study uses data of actual customer behaviour.
This study tries to answer the following research question: To what extend are there synergy effects towards the dependent variable of purchase probability (for purchases at any brand as well as purchases at the focus brand) when customer initiated touchpoints and firm initiated are used together? It will contribute to customer journey attribution research in a way that the research gap regarding the difference between customer initiated and firm initiated and the existence of synergy effects between them is filled in. Furthermore, this study can be the starting point of further research focussed on the integration of several touchpoints (i.e. customer initiated and firm initiated touchpoints) instead of measurement of effects of single touchpoints on outcomes as purchase probability. It will contribute towards managerial decision-making as it helps managers to make a choice of how to use their customer and firm initiated touchpoints in an integrative way to optimize their results.
2. Literature review and conceptual model
The focus on customer and firm initiated touchpoints is a delimitation of the customer journey concept. A step-wise literature review approach starting with the customer journey concept and working towards the delimitated parts of the customer journey concept used in this study will be used to come to the formulation of hypothesis.
As mentioned in the introduction, Anderl et al. (2016) define a customer journey of an individual customer as the process of using all touchpoints over all (online) marketing channels prior to a potential purchase decision leading to a(n) (online) store visit. The customer journey can also be defined as the journey a customer takes to achieve a shopping related goal, following the usage of a series of touchpoints (Stein & Ramaseshan, 2016). They view the typical customer journey as a sequence of four steps that can all be a shopping related goal itself: search for information, evaluate the option(s), purchase and post-purchase actions like retention. There are two important differences between the definitions of Ander et al. (2016) and Stein & Ramaseshan (2016). First, they mainly look at the customer journey from an online perspective. Second, the post-purchase stage as used by Stein & Ramaseshan (2016) is not used by Anderl et al. (2016). As this study also uses online customer journeys and the dependent variable is purchase probability, the definition of Anderl et al. (2016) is more applicable for this study.
As mentioned in the introduction, an increased focus on the customer journey emerged in the literature. Barwitz & Maas (2018) for example created a framework - based on prior research and novel findings following own research - on why customers choose for a certain pattern throughout their journey in an omnichannel setting. They distinguished two customer journey patterns: research shopping and impersonalization/interactivity reduction. The latter are mechanisms affecting the switch from purchase to post-purchase actions and are therefore not applicable to this study. The research shopping pattern concerns the combination of multiple channels to switch from pre-purchase to purchase. As they found that a combination of multiple touchpoints can lead to multiple outcomes (i.e. either increasing or decreasing the easiness of switching from pre-purchase to purchase) it can be a bridge from the general customer journey topic to the delimitated parts of the customer journey used in this study, especially the usage of synergy effects.
Touchpoints
Therefore, contrary to most prior research, this study will only focus on the integration of customer and firm initiated touchpoints instead the customer journey itself. Stein &
Ramaseshan (2016) already used the concept of touchpoints in their definition of the customer journey, implying the important linkage between the two concepts. In fact, there is no
any part of the product, service, brand or organization, via any channel at any point in time. These touches result in an experience perceived by the customer (Stein & Ramaseshan, 2016;Pantano & Milena, 2015;Zomerdijk & Voss, 2010).
Wiesel et al. (2011) found several cross channel effects of touchpoints, particularly between online and offline channels and that marketing communications have a direct effect
throughout the whole customer journey (i.e. both in early and later stages of the journey). There are also some more general interesting findings which suggest that optimizing touchpoints can lead to improved results, strengthening the presumption that optimizing touchpoints leads to improved results. For example, the first interaction with a new touchpoint is crucial for future usage of that touchpoint and a positive impression will result in a higher chance of usage, while a negative impression will result in a lower chance of usage
(Vakulenko et al., 2019). Baxendale et al. (2015) found somewhat similar results regarding positivity and added that touchpoint frequency also affects the consideration of touchpoints in a multi-touchpoint setting.
Based on the aforementioned presumption that optimizing touchpoints leads to improved results, the following hypotheses are formulated:
H1a. Online customer initiated touchpoints have a positive impact on purchase probability.
H1b. Online firm initiated touchpoints have a positive impact on purchase probability. Wiesel et al. (2011) also found that online customer initiated touchpoints have a greater impact on profit than offline firm initiated touchpoints have. Based on their finding, H1c is formulated:
H1c. Online customer initiated touchpoints have a greater impact on purchase probability than online firm initiated touchpoints have.
Synergy effects
Synergy effects is the last step in the delimitation of the customer journey concept. Synergy effects occur when the combined effect of multiple activities is larger than the sum of the individual effects of these activities (Naik & Raman, 2003;Belch & Belch 1998). Naik & Raman (2003) found evidence for the existence of offline synergy effects as there was a synergy effect between television and print advertising on the dependent variable sales. There is also evidence of synergy effects in the online settings, leading to higher purchase
for synergy effects between online broadcast media (e.g. banners and videos) and online interactive media (e.g. blogs and social media) on purchase intention (Dong et al. 2018). The latter can be linked to customer and firm initiated touchpoints as online broadcast media is a one-way communication like firm initiated touchpoints, whereas the role of the customer in the communication process is larger for both online interactive media an customer initiated touchpoints. Bartwitz & Maas (2018) report that both positive and negative synergy effects have been found, but that recent studies tend to confirm tend to confirm that channel
integration has a positive impact on outcomes such as sales growth. Furthermore, they found synergy effects between different channels as customer perceived benefits of previous used channels (for example for research purposes) when they purchased in another channel. What exact synergies exists between which channels is not elaborated by the authors, but is also not relevant for this study. The key message is that there is another signal that synergy effects between customer and firm initiated touchpoints would make sense. Combined with the aforementioned findings regarding synergy effects between touchpoints on sales, the following hypothesis is formulated:
H2. The combined effect of customer initiated touchpoints and firm initiated touchpoints on purchase probability is larger than the sum of the individual effects of customer initiated touchpoints and firm initiated touchpoints on purchase probability.
As the focus will be on a combination between customer and firm initiated touchpoints, it is assumed that the effect of firm initiated touchpoints on purchase probability is moderated by the effect of customer initiated touchpoints and vice versa. As will be elaborated in the
Figure 1: Conceptual model
3. Methodology
Data collection
The data for this study is provided by Growth from Knowledge (GfK), a German market research company operating globally. GfK is currently operating in more than 110 countries, has over 11,000 employees and is the fourth largest research company worldwide. Important to mention is that the data collection process started and ended before the start of this study, meaning that although the data is sufficient to perform the required tests, the dataset was in the first place not created to serve this study. Another important data collection remark is that despite GfK being a German company operating globally, data collection for this study is limited to the Dutch market.
GfK followed the personal customer journey of 29011 (potential) customers for one of the largest Dutch travel agencies from the 1st of June 2015 to the 31th of September 2016 by tracking their internet behaviour. The tracking was performed by adding a web browser extension to the (potential) customer’s browser, which subsequently sends the data to the GfK- Server with Nurago technology: turning big data into smart (i.e. interpretable) data. The tracking created a longitudinal research setting where all touchpoint usage over time is
captured. Automatically captured online data can generate the display of a large part of the customer journey (Baxendale et al., 2015;Trusov, Bucklin, & Pauwels 2009). Ages are
ranging from 17 to 94 with a mean of 52. Gender is equally distributed throughout the dataset, implying that there are no large differences between the amount of males and females.
Customer initiated touchpoints Firm initiated touchpoints (only for focus brand) Accommodations website Affiliates
Accommodations app Banner
Accommodations search E-mail
Information/comparison website Pre-rolls Information/comparison app Retargeting Information/comparison search
Tour operator/travel agent website competitor Tour operator/travel agent app competitor Tour operator/travel agent search competitor Tour operator/travel agent website focus brand Tour operator/travel agent search focus brand Flight tickets website
Flight tickets app Flight tickets search Generic search
Table 1: customer and firm initiated touchpoints within the database
Measurement
As shown in figure 1, this study’s dependent variable is the purchase probability (at the focus brand), whereas the customer and firm initiated touchpoints are the independent variables, which also moderate each other. The measurement of these variables will be discussed separately.
have a positive impact on this study’s validity. With the aggregated approach, it might be possible that – for example – one firm initiated touchpoints has by far the largest positive impact on purchase probability of all customer and firm initiated touchpoints together, whereas all other firm initiated touchpoints barely have a positive impact, resulting in firm initiated touchpoints ‘outscoring’ customer initiated touchpoints. This hypothetical situation would raise the question whether the statement ‘firm initiated touchpoints have a larger positive effect on purchase probability than customer initiated touchpoints’ would be valid. The same goes for the moderation analysis. The (lack of a) possible synergy effects and their direction could be overly affected by one or a minority of touchpoints. Finally, as will be elaborated in the ‘analysis’ section, to prevent biased outcomes, all touchpoints will be a binary variable implying whether the touchpoint was used or not. However, due to data
aggregation, customers can have touchpoint usage values higher than one. If – for example – a customer used a specific touchpoints four times, the touchpoint usage is four instead of one (whether it was used or not). As a result, the touchpoint usage variables changes to a continuous variable.
DV: Purchase probability
The focus brand as mentioned in table 1 is the aforementioned Dutch travel agency. The data consists of both purchases in general - a purchase at the focus brand or a competitor – and purchases at the focus brand. 3674 of the 29011 customer journeys lead to a purchase, 192 of the purchases were at the focus brand. As this is a relative small number and they are included in the purchases in general as well, purchases in general will be used as the dependent
variable. Since the focus brand identified the whole database as a potential customer it is likely that there are no differences within the dataset in the likelihood that customers could have been confronted by a firm initiated touchpoint, meaning that this choice will not affect H1 and H2. Just like the independent variable, the dependent variable will be a binary variable implying whether the (potential) customer purchased or not.
Analysis
To draw conclusions regarding the issues stretched out in the conceptual model (H1 and H2), RStudio has been used to perform tests with the data provided by GfK. Considering the binary dependent variable, binary logistic regression is used to test H1 and H2. The latter also
initiated touchpoint will be the same as firm and customer initiated touchpoints (i.e. the other way around) only one moderation analysis has been performed. Logically, significance values will indicate the existence of effects, whereas the measures Pseudo R-squared and log
likelihood will indicate how well the used models fit the data.
Furthermore, although no outliers have been detected (and therefore were not removed) several adjustments have been implemented to the original dataset provided by GfK in order to increase the reliability and validity of the study. The first important adjustments are made regarding the data structure. The dataset provided by GfK presented each step of a specific customer journey, meaning that the longer a specific journey was, the more rows were used. In order to ensure that the dependent variable was a binary variable, the data has been aggregated to one journey per row with a maximum of one purchase, meaning that length of the customer journey and the amount of purchases per person have not been taken into account. Subsequently, the structure of touchpoints usage was changed from one column containing all touchpoints (for example: touchpoint 1 indicated the customer initiated touchpoint ‘accommodations website’, whereas touchpoint 20 indicated the firm initiated touchpoint ‘email’) to twenty columns containing one touchpoint each. Within those columns, touchpoint usage was indicated by dummies, a one indicating that the touchpoint was used, whereas a zero indicates that the touchpoint was not used. Transforming the data structure allowed the aforementioned switch to continuous variables. Furthermore, maintaining the pre-allocated touchpoint usage structure could result in biased outcomes as a customer who – for example – used both touchpoints 2 and 10 four times would have a total touchpoint usage of respectively eight and forty times, whereas the dummies ensure that both touchpoint usages are indeed four. This was especially important for the aggregated customer initiated
(touchpoints 1 to 10 and 12 to 16) versus firm initiated (touchpoints 18 to 22) touchpoints as the pre-allocated numbers to firm initiated touchpoints were higher in general (compared to customer initiated).
Important to mention is that, contrary to original plans based on the work of Baxendale et al. (2015), no threshold value has been set to the amount of touchpoint(s) usages, meaning that both customers with only one touchpoint usage and customer with relatively high
this study’s purpose, it would make sense to implement touchpoint frequency in this multi-touchpoint setting study when the dataset used for the study includes multi-touchpoint frequency as well. It would have been relatively easy to subset the data with only people with a specific touchpoint usage, but exploration of both existing theory regarding touchpoint usage frequency and the current dataset did not lead to an appropriate threshold value, whereas using a random threshold value would decrease the outcomes’ reliability. Although not taking frequency into account at all is also likely to decrease the outcomes’ reliability, it is expected to be of a lesser extend compared to using a random frequency threshold value, which is the reason it was taken out.
Finally, all data structuration and analysis used for the possible relationship(s) between customer initiated and firm initiated touchpoints – and their possible synergy effect – on the purchase probability are also used for the analysis in the subset of people who purchased, regardless of the company. This subset will be used to perform the same tests, but now with purchase probability at the focus brand being the dependent variable.
4. Results
In order to perform the aforementioned binary logistic regression tests, it is necessary to test whether certain required assumptions for binary logistic regression hold. First of all, the dependent variable should be binary, implying that the dependent variable has two contradicting outcomes. The dependent variable of the first part of the study – where a purchase is not required – has two possible outcomes: a purchase (regardless of at which brand) or no purchase. The dependent variable of the second part of the study – where a purchase (regardless of at which brand) is required – has two possible outcomes: a purchase at the focus brand or a purchase at any possible brand except the focus brand. As the dependent variable in both parts of the study always provides one of the two possible outcomes, the first assumption for binary logistic regression holds. The second assumption for binary logistic regression requires a linear relationship between the logit of the outcome and each predictor variable. The binary logistic regression tests are performed by using the mfx package in RStudio, which gives marginal effects for the average observation. As a result, the marginal effect is the slope of the line (i.e. the relationship between the independent variable and the dependent variable), which never changes. Mathematically, marginal effects for a continuous independent variable should be interpreted in a way that a marginal increase in the
binary outcome of the dependent variable (i.e. 1 instead of 0 so for example a purchase or a response to an e-mail instead of not purchasing or not responding) by a certain percentage point. However, in order to increase managerial contributions, it might also be interpreted as an increase or decrease by one unit that leads to the increased or decreased probability of observing the dependent variable. A change in a binary independent variable should be interpreted as a binary change from 0 to 1, which always leads to the same change of percentage points of observing the dependent variable. The same goes for continuous
independent variables. If – for example – an increase in the independent variable by one unit results in an increased probability of observing the dependent variable by five percentage points, a second increase in the independent variable by one unit would again result in an increased probability of observing the dependent variable by five percentage points. As this implies a linear relationship, the second assumption for binary logistic regression holds. Finally, as will be elaborated per section of the results, there is no multicollinearity within the dataset, meaning that the multicollinearity assumption holds as well.
The elaboration of the results itself will follow a step-wise approach, starting with the combined effect of all customer initiated versus the combined effect of all firm initiated touchpoints on purchase probability, followed by the possible synergy effects between customer initiated and firm initiated touchpoints. Subsequently, the same will be done for all twenty touchpoints separately and the 75 possible combinations of synergy effects between customer initiated and firm initiated touchpoints. Finally, all aforementioned analysis will be repeated within the subset of customers with a purchase, meaning that the dependent variable then changes to purchase probability at the focus brand.
Aggregated customer initiated versus firm initiated touchpoints and their synergy effects Table 2 shows that, in line with H1a and H1b, there exists a positive relationship between the usage of both customer initiated and firm initiated touchpoints and purchase probability, since an increase in the number of touchpoint usages results in an increase of respectively 0.004 percentage points for customer initiated touchpoints and 0.019 percentage points for firm initiated touchpoints of the purchase probability. As both effects are significant at a
firm initiated touchpoint have a greater impact on purchase probability than customer initiated touchpoints. The same goes for H2 as the results are significant at a 0.001 level, but the effect is in the opposite direction compared to the expectation stated in H2. Thus, there is enough evidence to state that the combined effect of customer initiated touchpoints and firm initiated touchpoints on purchase probability is smaller than the sum of the individual effects of customer initiated touchpoints and firm initiated touchpoints on purchase probability. Regarding model quality, the three Pseudo R-squared and the significant negative likelihood difference value imply that the model outscores the null model which only contains the intercept. Finally, since all the VIF-values are lower than the threshold value of 5, there is no multicollinearity within the model.
Measure Effect Significance VIF-value
Effect of IV’s and moderation Customer initiated touchpoints Firm initiated touchpoints Customer initiated * firm initiated touchpoints
Pseudo R-squared and log likelihood measures
0.00004 0.00019 -0.000000002 *** *** *** 1.246029 2.320689 2.223412 McFadden Cox and Snell Nagelkerke Log likelihood difference 0.0608284 0.0720374 0.1018280 -361.78 *** Significance levels: *** = 0.001, ** = 0.01, * = 0.05, . = 0.1
Table 2: Effect of customer initiated versus firm initiated touchpoints on purchase probability and moderation
before coming to a conclusion regarding the customer initiated versus firm initiated
touchpoint debate. First of all, as can be seen in table 3, six of the twenty touchpoints have a significant, positive effect on purchase probability at a significance level of at least ten percent. Five of them are customer initiated touchpoints, meaning that the other one is a firm initiated touchpoint. Therefore, 33.33 percent of the customer initiated touchpoints has a significant positive impact on purchase probability, whereas this is twenty percent for firm initiated touchpoint. As a result, customer initiated touchpoints do outscore firm initiated touchpoints, which is in line with H1c. Thus, there is enough evidence to state that the effects of customer initiated touchpoints is larger than the effects of firm initiated touchpoints on purchase probability. Regarding model quality, the three Pseudo R-squared and the significant negative likelihood difference value imply that the model outscores the null model which only contains the intercept. Finally, since all the VIF-values are lower than the threshold value of 5, there is no multicollinearity within the model.
Measure Effect Significance VIF-value
Effect of IV’s Accommodations website Accommodations app Accommodations search Information / comparison website Information / comparison app Information / comparison search Touroperator / travel agent website competitor Touroperator / travel agent app competitor Touroperator / travel agent search competitor Touroperator / travel agent website focus brand
Touroperator / travel agent search focus brand Flight tickets website Flight tickets app Flight tickets search Generic search Affiliates Banner E-mail Prerolls Retargeting
Pseudo R-squared and log likelihood measures 0.03278 -0.000004 -0.00004 0.00698 -0.00035 0.0031 -0.00462 0.00229 0.00132 0.00147 * *** *** 1.083697 1.002858 1.078864 1.261626 1.651845 1.006239 1.059634 1.012854 1.052080 1.145302 McFadden Cox and Snell Nagelkerke
Log likelihood difference
0.1229 0.1402 0.1982
-731.17 ***
Significance levels: *** = 0.001, ** = 0.01, * = 0.05, . = 0.1
Table 3: Separate effects of customer initiated versus firm initiated touchpoints on purchase probability
As shown in table 4, there exists synergy effects for 29 of the 75 possible combinations of customer initiated and firm initiated touchpoints. Although positive synergy exists, the majority (68.97 percent) of significant synergy effects are – again – negative.
Significant at 0.001 level Significant at 0.01 level Significant at 0.05 level Significant at 0.1 level Positive effect Negative effect 2 7 0 3 4 8 3 2
Table 4: Separate synergy effects of customer initiated and firm initiated touchpoints
As shown in table 5, in line H1a and H1b, there does exists a significant positive relationship between the usage of both customer initiated and firm initiated touchpoints and purchase probability, meaning that there is enough evidence to state that both customer initiated touchpoints (H1a) and firm initiated touchpoints (H1b) have a positive or negative impact on purchase probability. However, just like the first part of the study, the effect of firm initiated touchpoints is larger than customer initiated touchpoint, which is not in line with H1c. As a result, there is in fact enough evidence to state that firm initiated touchpoint have a greater impact on purchase probability than customer initiated touchpoints. Again, the same goes for H2 as the results are significant at a 0.001 level, but the effect is in the opposite direction compared to the expectation stated in H2. As a result, there is enough evidence to state that the combined effect of customer initiated touchpoints and firm initiated touchpoints on purchase probability is smaller than the sum of the individual effects of customer initiated touchpoints and firm initiated touchpoints on purchase probability.
Regarding model quality, the three Pseudo R-squared and the significant negative likelihood difference value imply that the model outscores the null model which only contains the intercept. Finally, since all the VIF-values are lower than the threshold value of 5, there is no multicollinearity within the model.
Measure Effect Significance VIF-value
Effect of IV’s and moderation Customer initiated touchpoints Firm initiated touchpoints Customer initiated * firm initiated touchpoints
Pseudo R-squared and log likelihood measures
0.000004 0.000017 -0.0000000004 *** *** *** 1.377375 4.169049 4.211981 McFadden Cox and Snell
Nagelkerke Log likelihood difference 0.0721983 -37.354 *** Significance levels: *** = 0.001, ** = 0.01, * = 0.05, . = 0.1
Table 5: Effect of customer initiated versus firm initiated touchpoints on purchase probability at the focus brand and moderation
Customer initiated versus firm initiated touchpoints and their synergy effects per touchpoint for the subset of people who purchased
As shown in table 6, five of the twenty touchpoints have a significant effect on purchase probability at a significance level of at least ten percent. Four of them are customer initiated touchpoints, meaning that the remaining one is a firm initiated touchpoint. However, two of the four significant effects of customer initiated touchpoints are negative. Therefore, 13.33 percent of the customer initiated touchpoints has a significant positive impact on purchase probability. The significant effect of the firm initiated touchpoints ‘prerolls’ is also negative, meaning that zero percent of the firm initiated touchpoints has a significant positive effect on purchase probability, whereas this was 13,33 percent for customer initiated touchpoints, which is in line with the expectation stated in H1c. Thus, there is enough evidence to state that the effects of customer initiated touchpoints is larger than the effects of firm initiated
touchpoints on purchase probability at the focus brand. In line with the other parts of this study, the three Pseudo R-squared and the significant negative likelihood difference value imply that the model outscores the null model which only contains the intercept. Thus, this has positive impact on the quality of the model. Finally, all VIF-values were – again – lower than the threshold value of 5, meaning that there is no multicollinearity within the model.
Measure Effect Significance VIF-value
Information / comparison search
Touroperator / travel agent website competitor Touroperator / travel agent app competitor Touroperator / travel agent search competitor Touroperator / travel agent website focus brand Touroperator / travel agent search focus brand Flight tickets website Flight tickets app Flight tickets search Generic search Affiliates Banner E-mail Prerolls Retargeting
Pseudo R-squared and log likelihood measures
-0.00002 -0.000111 -0.00732 0.000165 0.00683 0.000019 -0.00007 0.000802 0.000058 -0.000099 -0.00131 0.000177 -0.01108 0.00000008 * *** . ** 1.204475 2.038095 1.067224 1.964762 1.956137 2.323132 1.169697 1.204655 1.434041 2.562060 1.030625 1.011916 1.013665 1.017325 1.402401 McFadden Cox and Snell Nagelkerke
Log likelihood difference
0.1810870 0.0627319 0.2085780
-120.66 ***
Significance levels: *** = 0.001, ** = 0.01, * = 0.05, . = 0.1
Table 6: Separate effects of customer initiated versus firm initiated touchpoints on purchase probability at the focus brand
As shown in table 7, there exists synergy effects for seventeen of the 75 possible
Worth mentioning is the fact that five of the twelve significant negative synergy effects occurred for touchpoint ‘Touroperator / travel agent website focus brand’. As a result, it is the only customer initiated touchpoint with significant synergy effects with all five firm initiated touchpoints. Significant at 0.001 level Significant at 0.01 level Significant at 0.05 level Significant at 0.1 level Positive effect Negative effect 0 8 0 0 0 3 5 1
Table 7: Separate synergy effects of customer initiated and firm initiated touchpoints for people who purchased
5. Conclusion and discussion
As discussed in the results, the first part of the analysis (i.e. where a purchase was not
required and thus purchase probability was measured for all brands) resulted in accepting both H1a and H1b as there was enough evidence to conclude that both customer initiated
touchpoints (H1a) and firm initiated touchpoints (H1b) have a positive impact on purchase probability. This means that an increase by one unit in customer initiated touchpoint usage in general (i.e. irrespectively of which specific customer initiated touchpoint is used) results in an increase of the purchase probability by 0.004 percentage points. For firm initiated
touchpoints an increase by one unit in firm initiated touchpoint usage in general results in an increase of the purchase probability by 0.019 percentage points. This is in line with both H1a and H1b as an increase in both customer initiated and firm initiated touchpoints lead to an increase in the purchase probability. From a managerial perspective the finding is insightful as a manager now knows (s)he can use both customer initiated and firm initiated touchpoints to improve his (or her) product’s purchase probabilities. Looking at the individual effects (i.e. all touchpoints measured separately), customer initiated touchpoints ‘accommodations website’, ‘Touroperator / travel agent website competitor’, Touroperator / travel agent website focus brand’, Touroperator / travel agent search focus brand’ and ‘flight tickets search’ have a significant positive effect on purchase probability. The firm initiated touchpoint ‘retargeting’ has a significant positive effect on purchase probability as well. This means that - for example - an increase by one unit in the usage of retargeting results in an increase of the purchase probability by 0.147 percentage points. Therefore, managers should use the six
products.
Testing H1c lead to conflicting results. Contrary to the expectations stated in H1c and the work of Wiesel et al. (2011), when the touchpoints were measured together (i.e. all customer initiated versus all firm initiated touchpoints) the effect of firm initiated touchpoints on purchase probability was larger than the effect of customer initiated touchpoints on purchase probability, instead of the other way around. Thus, there was not enough evidence to conclude that customer initiated touchpoints have a greater impact on purchase probability than firm initiated touchpoints (i.e. H1c was rejected), but there was enough evidence to conclude that firm initiated touchpoints have a greater impact on purchase probability than customer
initiated touchpoints have. This means that a manager should prefer firm initiated touchpoints over customer initiated touchpoints when (s)he is deciding whether (s)he should use customer initiated or firm initiated touchpoints to increase the purchase probability of his (or her) products. A possible explanation might be that the five customer initiated touchpoints used in this study (affiliates, banner, e-mail, prerolls and retargeting) are highly effective (compared to other possible firm initiated touchpoints). However, important to mention is that they are effective in increasing purchase probability in general, meaning that there are spill-over effects to competitors as well.
However, for the individual effects, customer initiated touchpoints outscored - in terms of impact on purchase probability – firm initiated touchpoints in the first part of the analysis (i.e. where a purchase was not required). Although the measurement of the touchpoints
individually provides managers insights into which touchpoint to use, more research is required to find out whether the conclusion (i.e. firm initiated touchpoints have a greater impact on purchase probability than customer initiated touchpoints have) based on measuring all customer initiated touchpoints combined and firm initiated touchpoints combined holds.
customer initiated versus all firm initiated touchpoints), nine positive interactions effects were found, but the majority of the interaction effects (twenty) were – again – negative. As a result, there was not enough evidence to conclude that the combined effect of customer initiated touchpoints and firm initiated touchpoints on purchase probability is larger than the sum of the individual effects of customer initiated touchpoints and firm initiated touchpoints on purchase probability, but there was enough evidence to conclude that the combined effect of customer initiated touchpoints and firm initiated touchpoints on purchase probability is smaller than the sum of the individual effects of customer initiated touchpoints and firm initiated touchpoints on purchase probability. This is partly in line with Bartwitz & Maas (2018) findings as they report - although the tendency is more towards positive effects – the finding of existence of both positive and negative synergy effects between customer initiated and firm initiated touchpoints. Thus, for managerial decision making, it is advised to use either customer initiated touchpoints or firm initiated touchpoint as using them in an integrative way is likely to lead to a decreased purchase probability.
The second part (i.e. where a purchase was required and thus purchase probability was measured for the focus brand) of the study also lead to the finding that there was enough evidence to conclude that both customer initiated (H1a) and firm initiated (H1b) touchpoints have a significant positive impact on purchase probability. This means that an increase by one unit in customer initiated touchpoint usage in general (i.e. irrespectively of which specific customer initiated touchpoint is used) results in an increase of the purchase probability by 0.0004 percentage points. For firm initiated touchpoints an increase by one unit in firm initiated touchpoint usage in general results in an increase of the purchase probability by 0.0017 percentage points. This is in line with both H1a and H1b as an increase in both customer initiated and firm initiated touchpoints lead to an increase in the purchase
unit in the usage of Touroperator / travel agent search focus results in an increase of the purchase probability at the focus brand by 0. 683 percentage points. From a managerial perspective it is interesting and satisfying to know that the usage of their own focus brand channels (i.e. website focus and search focus) increases the purchase probability at the focus brand. From that point of view, the significant negative effect ‘Touroperator / travel agent search competitor’ can also be explained as this is a channel owned by a competitor, meaning that a negative effect of this touchpoint on the purchase probability at the focus brand makes sense. Concluding, managers should use the aforementioned Touroperator / travel agent website focus and Touroperator / travel agent search focus touchpoints in order to increase their product’s purchase probabilities.
Furthermore, measuring the customer initiated and firm initiated touchpoints combined (i.e. all customer initiated versus all firm initiated touchpoints) lead to firm initiated touchpoints partly outscoring customer initiated touchpoints. Thus, there was not enough evidence to conclude that customer initiated touchpoints have a greater impact on purchase probability than firm initiated touchpoints (i.e. H1c was rejected), but there was enough evidence to conclude that firm initiated touchpoints have a greater impact on purchase probability than customer initiated touchpoints have. Just like the first part of the study, this means that a manager should prefer firm initiated touchpoints over customer initiated touchpoints when (s)he is deciding whether (s)he should use customer initiated or firm initiated touchpoints as this is more likely to increase the purchase probability of his (or her) products. As the dependent variable in the second part of the study is purchase probability at the focus brand, the aforementioned (i.e. at the first part) spill-over effects are not applicable here.
However, the fact that no individual significant positive effects of firm initiated relationships on purchase probability at the focus brand were found, whereas two customer initiated touchpoints were found to have a significant positive effect on purchase probability at the focus brand is – just like the first part of the study – contradicting with the results of the measurement of all customer initiated touchpoints combined and firm initiated touchpoints combined. As a result – again – more research is required to find out whether customer initiated touchpoints have a greater impact on purchase probability than firm initiated touchpoints have or vice versa.
probability. However, these results were – again – not in line with the expectations stated in H2 and existing research (i.e. the work of Dong et al. (2018)), meaning that it lead to a rejection of H2 as there was a significant negative instead of a significant positive interaction effect between customer initiated and firm initiated touchpoints, implying that separate usage of customer initiated and firm initiated touchpoints is more effective than using them
integrative. When all 75 possible interaction effects were measured separately instead of aggregated (i.e. all customer initiated versus all firm initiated touchpoints), five positive interactions effects were found, but the majority of the interaction effects (twelve) were – again – negative. Thus, there was not enough evidence to conclude that the combined effect of customer initiated touchpoints and firm initiated touchpoints on purchase probability is larger than the sum of the individual effects of customer initiated touchpoints and firm initiated touchpoints on purchase probability, but there was enough evidence to conclude that the combined effect of customer initiated touchpoints and firm initiated touchpoints on purchase probability is smaller than the sum of the individual effects of customer initiated touchpoints and firm initiated touchpoints on purchase probability. As mentioned in the first part of the results, this is – although the tendency is more towards positive effects – partly in line with Bartwitz & Maas (2018) findings that there exists both positive and negative synergy effects between customer initiated and firm initiated touchpoints. As a result, for managerial decision making, it is advised to use either customer initiated touchpoints or firm initiated touchpoint as using them in an integrative way is likely to lead to a decreased purchase probability. This is especially the case for the aforementioned customer initiated touchpoint ‘Touroperator / travel agent website focus brand’ as this touchpoint has significant negative synergy effects with all five firm initiated touchpoints, meaning that using this otherwise effective touchpoint in an integrative way with any of this study’s firm initiated touchpoints would seriously harm this touchpoint’s effectiveness.
further research is still required and another contribution of this paper is therefore that it can also be used as a starting point for further research. It contributes to managerial decision-making in a way that managers now know that both customer initiated touchpoints and firm initiated touchpoints are likely to increase their sales, but that firm initiated touchpoints are (probably) more likely to increase their sales than customer initiated touchpoints do.
Especially the discovery of some individual touchpoints that had a significant positive effect on the purchase probability (at the focus brand) give managers a guidance which touchpoints they should use. Finally, they now know that using customer initiated and firm initiated touchpoints together will probably decrease their sales.
6. Limitations and further research
As mentioned in the conclusion and discussion section, this study’s findings might be a foundation for further research regarding customer journey attribution research and customer initiated versus firm initiated research in general. The results are contributing to both research and managerial decision-making, but due to its novelty, somewhat contradicting results and possible improvements, further research is recommended.
First of all, the methodology section made clear that original plans for this study included a threshold value for touchpoint usage frequency, but exploration of both existing theory regarding touchpoint usage frequency and the current dataset did not lead to an appropriate threshold value and using a random threshold value would decrease the reliability of the outcomes. Therefore, it is recommended to include frequency in further research. Secondly, the contradicting research regarding whether customer initiated touchpoints outscore firm initiated touchpoints or the other way around need additional research. Additional research might also find that the reason for the contradicting results in this study is that there is no type of touchpoint (i.e. customer initiated or firm initiated) that is better in general and that is dependent on the specific touchpoints in the study. Thirdly, this study only used the customer journeys from the Dutch travel market, meaning that repeating the study in another country and/or in another industry could lead to different results. Finally, the study used a total number of twenty touchpoints (fifteen customer initiated and five firm initiated), but it is possible that these twenty are only a small part of all touchpoints within the travel industry, let alone other industries. Thus, repeating the study with other customer initiated and firm
References
Anderl, E., Becker, I., Von Wangenheim, F., & Schumann, J. H. 2016. Mapping the customer journey: Lessons learned from graph-based online attribution modelling. International
Journal of Research in Marketing. 33: 457-474.
Barwitz, N., & Maas, P. 2018. Understanding the Omnichannel Customer Journey: Determinants of Interaction Choice. Journal of Interactive Marketing. 43: 116-133.
Baxendale, S., Macdonald, E. K., & Wilson, H. N. 2015. The Impact of Different Touchpoints on Brand Consideration. Journal of Retailing. 91(2): 235-253.
Dong, X., Chang, Y., Liang, S., & Fan. X. 2018. How online media synergy influences consumers’ purchase intention: A perspective from broadcast and interactive media. Internet
Research. 28(4): 946-964.
Naik, P. A., & Raman, K. 2003. Understanding the Impact of Synergy in Multimedia Communications. Journal of Marketing Research. XL(4): 375-388.
Rosenbaum, M. S., Otalora, M. L., & Ramírez, G. C. 2017. How to create a realistic customer journey map. Business Horizons. 60: 143-150.
Stein, A., & Ramaseshan, B. 2016. Towards the identification of customer experience touch point elements. Journal of Retailing and Consumer Services. 30: 8-19
Vakulenko, Y., Shams, P., Hellström, D., & Hjort, K. 2019.Service innovation in
e-commerce last mile delivery: Mapping the ecustomer journey. Journal of Business Research. 101: 461-468.
