Impact of direct trafficeffect
on online sales
Lukáš Kakalej
cík
Department of Applied Mathematics and Business Informatics,
Technicka univerzita v Kosiciach, Kosice, Slovakia
Jozef Bucko
Department of Applied Mathematics and Business Informatics,
Faculty of Economics, Technical University of Košice, Kosice, Slovakia, and
Jakub Danko
Department of Statistics and Probability, University of Economics,
Prague, Czech Republic
Abstract
Purpose – This study aims to analyze the impact of newly created brand awareness on customer’s buying
behavior in online environment.
Design/methodology/approach – The authors analyzed more than 280,000 online customer journeys
from four e-commerce stores based in Slovakia. Within the results of the interaction analysis of individual
customer journeys, the authors determined three criteria based on the level of theoretical brand awareness.
The purpose was to determine their occurrence in real-world data.
Findings – It was found that each of the specified criteria accounts for the significant share of the
company’s revenues. Based on these criteria and the level of their occurrence, the authors introduced the term
direct traffic effect.
Research limitations/implications – Because of the available Web analytics tools, the data might be
imprecise because of data collection issues. There is also ambiguity in the interpretation of the customer
journey.
Practical implications – The company can build awareness among prospective customers by offering
them a positive customer experience during the first interactions online. Data proved that customer will not
only repeatedly visit the website from the direct traffic source but also his customer journey will end with the
purchase of the company’s products.
Originality/value – This paper fulfills the need for further research on the impact of multi-channel
marketing on brand awareness and consumer behavior, respectively.
Keywords Online consumer behaviour, Marketing research, Customer analytics,
Multi-channel measurement
Paper type Research paper
Introduction
In 2004, a new stage of the Web environment began and the classic solid content of the
websites was replaced by a space for sharing and joint content creation (
Dinucci, 1999). This
development stage, known as Web 2.0, plays an important role in the functioning of
businesses. By eliminating time and space boundaries, it gives the business the ability to
reach out to its potential customers anywhere and anytime in real-time. The result of this
opportunity is a partial or complete shift of business focus to the internet or mobile
applications. Furthermore, internet users’ movement is well-measurable, allowing
Impact of
direct traffic
effect
Received 27 January 2019
Revised 7 August 2019
Accepted 28 January 2020
Journal of Research in Interactive
Marketing
© Emerald Publishing Limited
2040-7122
DOI 10.1108/JRIM-01-2019-0012
The current issue and full text archive of this journal is available on Emerald Insight at:
https://www.emerald.com/insight/2040-7122.htm
companies to analyze not only engagement, transactions and customer sales but also allows
them to analyze their potential customers – the source of their visits, their behavior, and also
how far they are to become customers of the company (Clifton, 2015).
More than 90 per cent of users are not ready to buy during their first visit to the business
website (
Vanden Heuvel, 2014). On the contrary, from the first visit until the purchase, the
users go through a process called the customer journey. The customer journey represents
the sequence of steps that users gradually pass through the awareness stages, evaluating
alternatives up to the actual purchase of the product (Roberge, 2015). Customer journey
mapping is a model, which describes all interactions with the intent to improve these
interactions, resulting in an increase in sales and customer satisfaction (Van Den Berg and
Pietersma, 2015). With digital advertising booming in rising amount of platforms and
formats, it is even more important to track consumer’s digital footprints at a detailed level,
enabling advertisers to get deeper insights into online consumer’s behavior, as well as an
image of what is the impact on conversion in case of the exposure of customer to individual
advertising channels (Ghose and Todri, 2015). As customers do not live in a silo, their
customer journey consists of interactions with both online and offline marketing channels.
This paper focus on the online part of the customer journey as online channels are
considered to be suitable channels for sales activation and with the current development of
analytical tools, it is possible to attribute the value to particular online channels (Binet and
Carter, 2018). The following marketing channels are among the most frequently measured
and reflected by Web analytics tools:
Direct traffic: represents a situation in which a user enters the URL of a website
directly into the browser window or he visits a website through a saved bookmark.
Visits from mobile apps or offline advertising sources (TVs, billboards, flyers, etc.)
may also be considered as a direct visit in case of inappropriately selected or
implemented tracking of traffic sources;
Organic traffic: represents a situation where a user enters a key phrase in the search
engine (Google, Bing, Yahoo and others), and clicks on search results to go to the
website of the business;
Referral traffic: represents a user’s visit by clicking on a link placed on another
website (it usually does not include social networks);
Social media: represents a user’s visit by clicking a link placed on social networks
(Facebook, Twitter, LinkedIn and others);
E-mail: represents the user’s visit by clicking the link in the e-mail delivered to his
mailbox;
Paid search: represents the user’s visit by clicking on paid search results (e.g. Google
Ads platform); and
Display advertising: represents the user’s visit by clicking on a banner ad placed, for
example, on the Google Display Network; and other, less frequently used marketing
channels for online promotion.
The increase in potential customer touchpoints and the reduced control of the
experience require firms to integrate multiple business functions to create and deliver a
positive customer experience (
Lemon and Verhoef, 2016). Usually, businesses do not use
only one marketing channel to get the customer. These channels, in most cases, work in a
cohesive way that contributes to the customer’s acquisition. A merit value for acquiring a
customer should be assigned to each such channel. Attribution models are used to model
JRIM
this problem. Attribution modeling (multichannel attribution) is a set of rules that give the
individual marketing channel credit for obtaining a customer conversion (Shao and Li, 2011;
Clifton, 2015). Danaher and Van Heerde (2018) define attribution modeling as the science of
using advanced analytical methods to allocate sufficient credit for each contact/interaction
of the customer with marketing channels used by the business. In the previous studies that
we have carried out (Ferencová et al.,2015), a problem has been defined with the evaluation
of the utility of marketing channels during the sales cycle. In spite of the questionnaire
survey conducted on customers, it is often difficult for the customer to determine which
channels he interacted with prior to the purchase. This problem can be solved by attribution
modeling, which evaluates every customer interaction with the business. However, in this
article, attribution modeling is essential only to understand the complexity of the customer
journey in terms of interacting with the company before purchasing the product.
There have been several studies that offered data-driven approaches to the
attribution to overcome t he weaknesses of standard heuristic models.
Yadagiri et al.
(2015) and Nissar and Yeung (2015) use Shapley value in their non-parametric approach
to attribution as a game theory-based model. In his thesis, Rentola (2014) used two
models: binary logistic regression to classify customers to converters and non-
converters (purchasers/non-purchasers), as well as a logistic regression model with
bootstrap aggregation. On the other hand, Shao and Li (2011) used bagged logistic
regression and a probabilistic model in their study. In their study, Li and Kannan (2014)
used a hierarchical Bayesian model. Geyik et al. (2014) developed their attribution
algorithm MTA to solve two problems: spending capability calculation for a sub-
campaign and return-on-investment calculation for a sub-campaign [more in (Geyik
et al., 2014)]. On the contrary, Wooff and Anderson (2015) offer an attribution
mechanism based on the appropriate time-weighting of clicks using the sequential
analysis. Hidden Markov model was used in the studies conducted by Abhishek et al.
(2012) and Wang et al. (2015). M arkov chain model was proposed in several studies as
well (Anderl et al., 2014; Anderl et al., 2015, 2016). For the purpose of our study, we
adopted the Markov model with the GDL estimator used in the study by Kakalej
cík
et al. (2018) to determine the importance of online marketing channels.
Previous studies on attribution modeling (Anderl et al., 2016; Rentola, 2014; Li and
Kannan, 2014) have shown that a specific source – Direct Traffic – has been an important
marketing channel with the merit of generating purchases and sales. Direct Traffic can be
labeled as a brand awareness aspect. Brand awareness could be defined as the extent to
which consumers are familiar with the distinctive qualities or image of a particular brand of
goods or services. Awareness is distinguished in terms of two dimensions as follows:
intensity and extent. The intensity of brand awareness indicates how effortlessly consumers
recall a particular brand. The extent of brand awareness refers to the possibility of acquiring
and consuming brand services and products especially when the brand emerges in
consumers’ minds (Barreda et al.,2015). The extent of brand awareness can be understood
as physical availability defined by Romaniuk and Sharp (2015), which in the context of
online media (such as websites or social networks) refers to the possibility to purchase the
product online. Brand awareness, in accordance with established marketing theory and
practice standards, can take three forms:
(1) Top-of-mind: represents the gold standard of brand awareness. In this case, the
brand is the first to be remembered by the customer without any help/support;
(2) Spontaneous awareness: represents recognition of the brand by the customer, with
no help given; and
Impact of
direct traffic
effect
(3) Supported awareness: different help is given to the customer, and it is monitored,
which brands come to his mind in the given context (Kahn, 2013).
Brand awareness plays a key role in consumers’ buying decision-making process (Binet and
Carter, 2018). It possesses aspects such as individual recognition, the dominance of
knowledge and brand recall (Kim et al.,2008). When typing the name of the website to the
Web browser window, the customer usually have to type the brand name, and therefore,
remember or recall it. Consequently, Direct Traffic, in most cases, might be connected to
spontaneous awareness or even “top-of-mind” brand awareness – depending on whether the
website came to mind of the customer as the first or another alternative to solve his current
problem. Ash (2012) claim that Direct Traffic referral means that the person is specifically
aware of and looking for the company. It is usually achieved as a result of repeated exposure
to the brand’s diverse settings. Bones et al. (2019) discuss that the analysis of Direct Traffic
over time can help brands understand changes in brand awareness, especially rises in Direct
Traffic can provide an indication of increased brand awareness. In the context of brand
awareness building in the online environment, Barreda et al. (2015) discuss that the quality
of the system (navigation simplicity, good user experience and security), the quality of
information (information that help the user to make a better decision), rewards (users obtain
financial, psychological or membership privileges) and virtual interaction (the range in
which users can participate in changing the content of the website in real-time) are among
the bearers of building the brand awareness. This means that the user is transitioning from
the search of how to resolve the problem to a direct business visit, which can solve his
problem under the conditions of meeting these prerequisites through the business website.
Anderl et al. (2015) divide the sources of website visits into two categories as follows:
(1) Channels initiated by the customer, which are further divided into branded
(including Direct Traffic) and generic;
(2) Business-initiated channels: it represents the promotional activities of the business
(e-mail, affiliate, banner ads, etc.).
Anderl et al. (2015) in their taxonomy model describe that after several visits to the website
through the business-initiated channels, the customer moves to the stage when the visits are
initiated by himself, which is a shift in customer decision-making about the product during
the purchasing process. The shift that has just been mentioned, is the subject matter of this
part of the paper. Li and Kannan (2014), in this context, list the concepts of spillover effect
and carryover effect. A carryover effect occurs when a user visits a website through a single
source of traffic, and subsequently, visits (and eventually) buys during a visit from the same
source. On the contrary, a spillover effect occurs when a user visits a website through a
single source of traffic, and subsequently, visits (and eventually) buys during a visit from
another source. Within our analysis, we will assume that both these events will occur in the
context of Direct Traffic as a brand awareness indicator. Linking brand awareness to the
source Direct Traffic is the research content of the presented study.
Sample and methods
The presented study aims to analyze the impact of online brand awareness on customer’s
purchasing behavior, based on the current state of knowledge. By decomposition of this
objective, we determined the following partial objectives:
The analysis of the current state of discussed issue in areas of multichannel
attribution and online brand awareness;
JRIM
Determine the importance of the source Direct Traffic by using Markov chains for
multichannel attribution and analyzing transition probabilities, as well as removal
effects in case of four Slovak e-commerce stores and compare it with the results
from the previous studies (Anderl et al., 2016; Rentola, 2014; Li and Kannan, 2014);
and
Observe the incidence of three defined criteria connected to online brand awareness
and determine their impact on the business performance of the analyzed e-
commerce stores.
Primarily, we focused on “website stickiness” after first customer visits, based on the
taxonomy model defined by
Anderl et al. (2015). For the purpose of our study, the data from
four e-shops (companies) were collected, one of which is focused on the sale of electronic
components, two of which are focused on the sale of sportswear, the latter focusing on the
sale of nutritional supplements [Table I displays characteristics of businesses based on
Finstat (2017)]. The names of the companies remain anonymous in agreement with their
representatives. As was mentioned in the introduction of this paper, customers do not live in
a silo and interact with both offline and online communication activities of the companies.
To prevent offline communications to have an effect on our results, we selected companies
that did not execute any offline communication three months prior to and during the period
of data collection.
The data on customer journeys of e-shop customers of the analyzed companies were
obtained from the
Google Analytics (2018) platform that companies use to measure the
performance of their websites (e-shops). The description of input data in terms of the
number of customer journeys and the volume of generated revenue is shown in Table II.
To perform the analysis, the following customer journeys were excluded from the
available customer journeys data sets in the first step:
Customer journeys beginning with Direct Traffic that indicate previous brand
awareness or brand experience (re-purchase, offline promotion of the business – are
in contradiction with study goals). There are several limitations to this step: when
inactive cross-device tracking of users, the customer can initiate the exploration
phase on the mobile and later purchase the product on the desktop device by
remembering the URL of the page. In this case, in Web analytics software, the
customer journey starts with Direct Traffic; users and their customer journeys are
tracked through cookies. If a customer commenced a product survey, deleted his
cookies and then purchased the product by visiting the website after entering the
URL address in the browser, the shopping journey recorded by the analytics
software starts with Direct Traffic; in customer journeys, it was not possible to mark
search queries from organic search related to the brand as Direct Traffic. So, if a
user searched for a store name through a search engine, visited the website and later
purchased, his customer journey starts with the organic search source instead of
Direct Traffic; and
Customer journeys that do not contain the Direct Traffic source.
Subsequently, three monitored criteria were defined, on the basis of which we conducted the
analysis:
Criterion 1: the last source (step) in the customer journey is Direct Traffic. In this
case, we do not monitor the spillover or carryover effect because both cases may
occur (customer journey may end Direct Traffic > Direct Traffic and Social
Impact of
direct traffic
effect
Company 1 Company 2 Company 3 Company 4
Subject of activity Distribution of industrial
electronic components for
industrial production
Retail sale of sporting goods of
a wide range
Retail sale of sporting goods
with a focus on running and
triathlon
Retail sale of food and
nutritional supplements
Revenues in 2016 in this e 15,561 16,018 308 4,993
Number of employees 50-99 200-249 3-4 20-24
Monitored period of customer
journeys
April 1, 2016- August 31, 2016 July 1, 2016-June 30, 2017 July 1, 2016-June 30, 2017 December 4, 2016-
December 4, 2017
Table I.
Characteristics of the
analyzed companies
JRIM
networks > Direct Traffic). We are only interested in the impact of Direct Traffic
(and therefore, brand awareness) on the buying behavior of the users;
Criterion 2: in the customer journey, the Direct Traffic source is at least twice in a
row at any point in the customer journey. In this case, we monitor the spillover
effect from another channel into the channel we are monitoring, as well as the
transfer effect in the Direct Traffic source that indicates brand awareness. In this
case, we are not interested in what channel the customer journey ends with. Anderl
et al. (2015) discuss that there is a positive interaction effect when using customer-
initiated channels and the following business-initiated channel visits, which is
greatly influenced by the company’s ability to use remarketing strategies.
Remarketing (also referred to as behavioral targeting) is a way to promote a product
to the people who have previously visited a business website (Marshall and Todd,
2017). In the case of a user’s reaction to a remarketing campaign, the customer
journey can end with, for example, the source banner ad or paid search; and
Criterion 3: in the customer journey, the Direct Traffic source is at least three times
in a row at any point in the customer journey. This number indicates even stronger
brand awareness than Criterion 2. At the same time, it represents either the
formation of positive brand preferences or the user’s decision-making among the
available alternatives to the product offered by other businesses as well. In this case,
we again monitor the spillover effect from another channel into the channel being
monitored by us, as well as the transfer effect in the Direct Traffic source that
indicates brand awareness. As in the previous case, we are not interested in what
channel/source of traffic the customer journey ends with.
Before the analysis of the criteria was conducted, we used Markov chains to determine the
importance and value of Direct Traffic. Formally, a sequence of random variables X
t
fg
1
t¼1
,
X
t
[ S :¼ {s
1
, ...,s
m
}, is a Markov chain of order r if, for all (a
1
, ..., a
t þ 1
) [ S
tþ1
, P(X
tþ1
=
a
tþ1
|X
1
= a
1
, ..., X
t
= a
t
)=P(X
tþ1
= a
tþ1
|X
trþ1
= a
trþ1
, ..., X
t
= a
t
) and r is the smallest
integer to satisfy it. Essentially, this represents that the probabilities related to X
tþ1
depend
only on the last r events, for all t.
In this context, S is referred by the state space, a particular sequence (a
1
, a
2
, ...) [ S
1
is
called by a trajectory, the size of S is the length of state space or number of states,
represented by m, and the probabilities of X
t þ 1
= a
t þ 1
considering that (X
trþ1
, ..., X
t
)=
(a
trþ1
, ..., a
t
) are called the transition probabilities represented by the notation p(a
tþ 1
|a
t rþ1
, ..., a
t
:¼ P(X
tþ1
= a
tþ1
|X
trþ1
= a
trþ1
, ..., X
t
= a
t
). A particular state b is
absorbing if the probabilities to leave the state are “0”,i.e.p(c|a
trþ1
, ..., b)=0,V c = b, and
consequently, p(b|a
t rþ1
, ..., b)=1.
A Markov chain can be represented by an initial probability distribution for the first r
steps and the m
rþ1
transition probabilities. When r=1, it is possible to have a graphic
Table II.
Characteristics of
input data on
customer journeys
Company 1 Company 2 Company 3 Company 4
Number of conversions 6,304 21,119 2,118 255,034
Total amount of purchases e1,579,778.00 e976,514.82 e219,719.49 e4,889,682.05
Average order value e250.60 e46.24 e103.74 e19.17
Customer journey duration (mean) 23.92 15.72 15.73 20.20
Customer journey duration (median) 14 9 6 10
Impact of
direct traffic
effect
representation for the Markov chain. For more details about Markov chains, we recommend
(
Karlin and Taylor, 1975).
Anderl et al. (2014) propose the use of Markov chains on channel attributions,
considering the state space S as the states “start” and “conversion” combined with the set of
marketing channels. In this case, the process {X
t
} represents the possible customer journeys
through these channels. They suggest using a removal effect for attribution modeling. The
removal effect is defined as the probability to achieve the conversion from the “start” state if
some of the states (s
i
) are removed from the model. As the removal effect reflects the change
in conversion rate if the given state s
i
is removed, the value (or importance) of the given
marketing channel can be determined. If N conversions are generated without the particular
channel (compared to the number of conversions in the full model), the removed channel
determines the change in the total number of conversions (Bryl, 2016).
In addition to the above methods and procedures, elements of descriptive statistics and
characteristics of variables (average, median and quartiles) were used for data analysis.
Data were analyzed using the statistical platform
The R Project for Statistical Computing
(2016).
Results
To vindicate the importance of this study, it was necessary to examine the importance
of Direct Traffic and its impact on sales. In the first part of the study, we analyze all of
the buyer journeys with the use of Markov chains in accordance with the theory used in
the previous section of the paper. We are particularly looking at the transition
probability from any other channel to Direct Traffic and the transition probability from
Direct Traffic to purchase. In addition, we are interested in measuring the r emoval
effect, which is the direct indicator of the importance of the marketing channels.
Afterward, we proceeded with the analysis based on the three criteria we determined in
the methodology section of this paper.
The initial step of the analysis was the generation of transition diagrams and also
the generation of transition matricesforeachanalyzedcompany.Figure 1
shows the
customer journeys transition diagrams for all the analyzed businesses. The individual
points of the graphs represent specifi cstates– marketing channels. It can be noticed
that the transition d iagram starts with the state (start) that represents the start of the
customer journey, and e nds with the state (conversion) that represents the conversion or
transaction. Individual states are linked by nodes, each n ode containing information
about the transition probability from a particular state to another particular state. The
nodes between the two marketing channels m and n show two probabilities – the
probability of transition from state m to state n and the probability of transition from
state n to state m. The nodes that connect states (start)and(conversion) contain only
one probability because no customer journey is heading to the state (start). Likewise, no
customer journey i s heading from the state (conversion) toward the marketing channels
used. This is a logical state because, after the performed transaction, it does not make
sense to monitor what marketing channels the customer uses at the point of further
interaction with the business. In the s tate (conversion), you can see a loop with a
pictured probability of 1. This loop originated from computational reasons during the
implementation of all the customer journeys.Asallthecustomerjourneyshavetogo
from the state (start)tothestate(conversion), the loop serves to complete each further
interaction until the moment when all the customer journeys from the data file get
through the state (conversion).
JRIM
When analyzing the transition probability, two events can be observed in all four
companies:
(1) If we do not consider completing a purchase (conversion), with almost every
marketing channel, the most likely next step is to visit the website from the Direct
Traffic source. This means that customers will either remember the page (and will
write a URL directly to the browser on the next visit) or they will save the page as
a bookmark to access it later; and
(2) When visiting a website whose source is labeled as Direct Traffic, there is the
highest probability that customers will purchase the product.
When analyzing the selection of the appropriate order for the Markov chain, changes in the
removal effect were also in the center of our attention. It is established that the higher
the removal effect of a given marketing channel, the more important a marketing channel for
the business because excluding it from the marketing portfolio would greatly reduce the
number of transactions (conversions) achieved.
Table III shows the removal effects using
the first-order of the Markov model in terms of conversions (C), as well as in terms of
revenue generated (R).
Figure 1.
Transition diagrams
of the customer
journeys
Impact of
direct traffic
effect
From Table III, it is obvious that Direct Traffic is reaching the highest removal effects,
both in terms of conversions and revenue. This knowledge directly supports the findings
from the previous part of the analysis, which concluded that visits from the Direct Traffic
channel have the greatest chance of ending up with a purchase. These findings are
corresponding with the results obtained in previous studies (Anderl et al.,2016; Rentola,
2014; Li and Kannan, 2014). These results lead us to analyze the impact of Direct Traffic on
sales more in detail – by examining results in accordance with the criteria set in
methodology.
Table IV illustrates the occurrence of the given criteria in the customer journeys of the
analyzed companies. It can be noticed that the tightening of the criteria also decreases the
share of the number of customer journeys in their total number. When comparing Criterions
1 and 3, the number of customer journeys decreases in all analyzed companies by more than
a half. The results of the analysis carried out are of the highest importance for Company 4
because, in both absolute and relative numbers, customer journeys according to established
criteria occur the most in case of this company in comparison with other analyzed
companies.
Table V shows the relative frequency of the number of interactions (website visits) that
precede a visit from the Direct Traffic source and end with a purchase (Criterion 1). Looking
at the table, it is possible to see that the values of the four analyzed companies are very
similar. Approximately 30 per cent of the Criterion 1 customer journeys require only one
prior visit from another source to follow a visit from the Direct Traffic source that ends with
the purchase of the product. This phenomenon can represent a high “likeness” of a product
or more precisely brand that results in spontaneous brand awareness and subsequent
purchase with a low number of visits to the website (the entire customer journey can also
Table IV.
Removal effects
(Markov model of the
first-order)
Company 1 Company 2 Company 3 Company 4
CRCRC RCR
Direct traffic 0.95 0.92 0.70 0.73 0.58 0.61 0.86 0.89
Organic search 0.30 0.32 0.39 0.38 0.40 0.37 0.41 0.40
Reference resources 0.10 0.12 0.17 0.17 0.36 0.40 0.13 0.14
Social networks 0.02 0.02 0.06 0.06 0.03 0.03 0.21 0.21
E-mail 0.16 0.17 ––0.02 0.02 0.06 0.05
Paid search 0.21 0.22 0.41 0.40 0.38 0.36 0.41 0.39
Banner advertising 0.01< 0.01 0.01 0.01 0.02 0.02 0.03 0.03
Other ––––– –0.03 0.03
N/A ––0.18 0.18 0.01< 0.01< 0.11 0.11
Table III.
The number and
proportion of
customer journeys
according to
established criteria
Company 1 Company 2 Company 3 Company 4
Total number of conversions 6,304 21,119 2,118 255,034
Criterion 1 1,310 5,347 465 80,271
Criterion 1 (%) 20.78 25.32 21.95 31.47
Criterion 2 880 3,402 259 51,497
Criterion 2 (%) 13.96 16.11 12.23 20.19
Criterion 3 632 2,284 167 34,656
Criterion 3 (%) 10.03 10.81 7.88 13.59
JRIM
last only a single day). Relative frequency of the number of interactions from 1 to 4
interactions prior to the Direct Traffic source visit decreases by 30-40 per cent with each
additional interaction. Concerning the users who end up their customer journey by
purchasing while visiting a website from the Direct Traffic source, it is possible to see high
decisiveness. This is also evidenced by the cumulative proportions of the length of customer
journeys for each analyzed company, which are shown in
Figure 2. It says that
approximately 90 per cent of the customer journeys are composed of 10 or fewer
interactions. However, as the last interaction from the Direct Traffic source is not included in
this number, this statement needs to be modified as follows: approximately 90 per cent of
customer journeys ending in a visit from the Direct Traffic source consists of 11 or fewer
interactions. The conclusion is that brand awareness has a positive impact on shortening the
customer journey before the purchase made by customers online.
The goal of Criterion 2 was to analyze the number of interactions with a website that
triggers spontaneous brand awareness so that users (customers) visit the website twice in a
row through the Direct Traffic source, regardless of what source they make the purchase
itself when visiting. Thus, the customer journeys can end with the Direct Traffic source but
they can also end with any other source from the spectrum used.
Table VI states that in
customer journeys where this criterion was met, 66-72 per cent of customers had only one
visit to build relatively strong brand awareness, they were in more frequent interaction with
the website of the companies and completed their customer journey with a purchase. The
conclusion is that in the online environment, a part of the customers’ needs just one
interaction with the website to remember the name of the brand or the URL of the website, to
buy on this website later. However, the number of these purchases is low, as the monitored
customer journey must consist of a specific number of interactions so that the set criterion
can be reached.
As mentioned, very strong brand awareness is expressed by set Criterion 3. Within it,
customer journeys have been analyzed, showing a visit from the Direct Traffic source three
times in a row, which means a very strong engagement and absolute brand awareness or
website as such. Looking at
Table VII it can be noticed that the values of 1 and 2 of the
previous interactions are almost identical to those of Criterion 2 (however, they are slightly
lower with the exception of Company 1 concerning the two previous interactions). However,
from the number of interactions three and more, we can notice a slight increase in the
number of frequency, which may mean that a larger number of previous visits from other
sources results in an increase in brand awareness so that the customer starts visiting the
website only through the Direct Traffic source.
Based on the analysis of Criterions 2 and 3, it can be concluded that according to the Li
and Kannan (2014) study, there is a strong spillover effect of transmitting website visits
Table V.
Relative frequency of
the length of
customer journeys
(Criterion 1)
Relative frequency
No. of interactions Company 1 (%) Company 2 (%) Company 3 (%) Company 4 (%)
1 27.18 31.34 33.76 32.18
2 18.24 19.64 19.78 19.90
3 11.83 13.63 13.12 12.40
4 7.02 9.67 8.82 8.39
5 5.88 6.25 6.24 5.96
6 4.58 4.36 3.66 4.22
7 2.98 2.92 3.66 3.32
8 and more 22.29 12.19 10.97 13.6
Impact of
direct traffic
effect
from the previously used website traffic sources toward a specific source Direct Traffic,
which results in the carryover effect when the customer further realizes his interaction with
the website from the Direct Traffic source. The combination of these effects also has an
impact on ending the customer journey in the form of a realized purchase.
At the beginning of this part of the work, a range of customers of the monitored
companies, whose criteria for the Direct Traffic effect are related, was defined. However, its
Table VI.
Relative frequency of
interactions before
reaching Criterion 2
Relative frequency
No. of interactions Company 1 (%) Company 2 (%) Company 3 (%) Company 4 (%)
1 69.32 68.72 71.43 66.34
2 13.30 12.73 12.74 13.21
3 7.95 8.64 6.56 8.74
4 3.30 4.29 4.25 4.47
5 2.39 2.44 1.16 2.78
6 and more 3.75 3.17 3.86 4.46
Figure 2.
Number of
interactions –
cumulative frequency
(Criterion 1)
JRIM
impact on company financial indicators – in this case, on generated revenue gained by the
customers whose conversion journeys met the set criteria – also reflects the importance of
the observed effect. The overview of the services generated on the basis of defined criteria is
provided in
Table VIII.
Table VII shows that for Companies 1, 2 and 3, the proportion of revenue in total sales is
higher than the share of monitored customer journeys. This means that the value of these
purchases is higher than the proportion of those purchases in total purchases. The
purchases sorted out according to the set criteria are the most important for Company 1
because they account for about one-third of all purchases in their e-shop. Concerning
Company 4, proportionately lower earning shares were recorded for purchases filtered
based on Criterions 1 and 2. However, the absolute value of the sales of these purchases is
clearly the highest among all the companies. Based on the monitored values, it can be said
that the Direct Traffic effect deserves attention as a field of further research, as there is a
possibility that optimization of brand awareness in the online environment can have a high
added value for the company.
Conclusions and limitations
Customers’ shopping decisions force marketing professionals to look at the customer
journey beyond their last interaction before purchasing. The presented study aims to
analyze the impact of brand awareness created online on customer’s buying behavior, based
on the current state of knowledge. Within the results of the interaction analysis of individual
customer journeys focused on the Direct Traffic source, the term Direct Traffic effect was
introduced. The source (marketing channel) Direct Traffic was placed on the level of
the creator and the result of brand awareness in the online environment. During the
customer journey, the customer moves from company-initiated interactions to interactions
initiated by himself (
Anderl et al.,2015), which also results from the study carried out by
Table VII.
Relative frequency of
interactions before
reaching Criterion 3
Relative frequency
No. of interactions Company 1 (%) Company 2 (%) Company 3 (%) Company 4 (%)
1 65.66 66.29 70.66 61.73
2 14.87 12.26 10.18 12.69
3 8.07 9.11 8.38 10.16
4 4.11 5.17 4.19 5.60
5 2.53 3.15 1.20 3.73
6 and more 4.75 4.03 5.39 6.09
Table VIII.
Revenues generated
by defined criteria
and their share in the
total revenues
Company 1 Company 2 Company 3 Company 4
Total revenue e1,579,778.00 e976,514.82 e219,719.49 e4,889,682.05
Criterion 1 e481,133.74 e260,403.89 e61,144.36 e1,295,538.59
Criterion 1 (%) 30.46 26.67 27.83 26.50
Criterion 2 e542,035.20 e180,311.11 e36,184.32 e947,862.87
Criterion 2 (%) 34.31 18.46 16.47 19.38
Criterion 3 e531,638.96 e128,507.20 e25,788.08 e686,486.70
Criterion 3 (%) 33.65 13.16 11.74 14.04
Impact of
direct traffic
effect
Li and Kannan (2014) concerning both the carryover effect and the spillover effect. There are
three criteria that support the conclusions of the previous studies. It was found that each of
the specified criteria (in absolute values) accounts for a significant share of the company’s
revenues. In the case of customer-initiated visits, a company does not pay for customer
interaction. In conjunction with the theory of Barreda et al. (2015), the company can build
awareness among prospective customers by offering them a good customer experience
during the first interactions. This will brand on the customer’s memory so that he will not
only repeatedly visit the website from the Direct Traffic source but also his customer
journey will end with the purchase of the company’s products. Based on the Direct Traffic
effect results, it is also possible to see that for more than 60 per cent of customers, only one
previous interaction with the company’s website is sufficient.
The results of this analysis can be influenced by factors that could not be taken into
account during its implementation. The limitations are as follows:
Data collection at the cookie level: as mentioned in the previous sections of the study,
the user data might be considered to be data regarding the user’s single device
because of cookies. Thus, if a user uses more than one device, the Web analytics
software [unless cross-device tracking is set (Alhlou et al., 2016)] will record him
multiple times as a different user. Additionally, if a user deletes cookies in a Web
browser, the Web analytics software will record him as a new user. Cookies are,
however, according to Flosi et al. (2013) standard for tracking in multichannel
analytics;
Customer journeys lasted up to 30 days: some customer journeys could last longer
than 30 days. This could cause the first interaction of the actual customer journey
that might occur in the past and were not recorded, which could have been the cause
of distortion of the attribution modeling results. The Direct Traffic effect analysis
could also filter customer journeys that started with other channels than Direct
Traffic. Because these interactions took place earlier (as monitored 30 days long
window), the analytics system could record the Direct Traffic source as the first
source of the visit;
Customer journeys represented interactions with the website: the customer could also
come into contact with the marketing communications of companies elsewhere than
on the company’s website. For example, he could see an ad and not click on it, look
through a page on social networks (like the Facebook page) and not click on a
website, etc. Such behavior was not included in customer journeys; and
The ambiguity of the Direct Traffic source: the Direct Traffic source could represent
one of the other marketing resources, e.g. a visit from a mobile app (Facebook and
Messenger), a browser bookmark or an offline ad such as billboards, leaflets or
catalogs. In addition, each of the analyzed companies has a bricks-and-mortar shop.
However, by our selection of the companies, we have tried to eliminate the impact of
offline advertising.
Future research should focus on the elimination of the abovementioned limitations.
Moreover, as the brand awareness, if not supported, decays over time (
Binet and Carter,
2018), we would like to examine the potential difference in the contribution of Direct traffic
into generated profit in various timing conditions. Additional research should also focus on
finding the particular elements of the website that drive customers into remembering the
brand/website name while visiting the website. These elements might affect the future
JRIM
online revenue of the companies, and therefore, by knowing its impact, companies are able
to improve customer’s experience in accordance with Lemon and Verhoef (2016).
References
Abhishek, V. Fader, P. and Hosanagar, K. (2012), “Media exposure through the funnel: a model of multi-
stage attribution”, available at:
http://poseidon01.ssrn.com/delivery.php?ID=7191050920240
971251121070170800081180380690810830390911210921260900861070980670280661230291190
220591210270200640940870950940660170780600770290750980650950041240651130850420320
0506400908210409200 30220310 9107209 8029 02307612 70991231 0600207 90100120 8712 4078&
EXT=pdf (accessed 24 September 2018).
Alhlou, F., Shiraz, A. and Fettman, E. (2016), Google Analytics Breakthrough: From Zero to Business
Impact, John Wiley and Sons. Hoboken, p. 625. ISBN 978-1-119-14401-4.
Anderl, E., Schumann, H.J. and Kunz, H.W. (2015), “Helping firms reduce complexity in multichannel
online data: a new taxonomy-based approach for customer journeys”, Journal of Retailing,
Vol. 92 No. 2, pp. 185-203.
Anderl, E., Wangenheim, F. and Schumann, H.J. (2016), “Mapping the customer journey: lessons
learned from graph-based online attribution modeling”, International Journal of Research in
Marketing, Vol. 33, pp. 457-474.
Anderl, E. Becker, I. Wangenheim, F.V. and Schumann, J.H. (2014), “Mapping the customer journey: a
graph-based framework for online attribution modeling”, available at:
http://ssrn.com/abstract=
2343077 (accessed 24 September 2018).
Ash, T. (2012), Landing Page Optimization: The Definitive Guide to Testing and Tuning for
Conversions, John Wiley and Sons. Indianapolis, ISBN 978-0-470-61012-1.
Barreda, A.A., Bilgiham, A., Nusair, K. and Okumus, F. (2015), “Generating brand awareness in online
social networks”, Computers in Human Behavior, Vol. 50, pp. 600-609.
Binet, L. and Carter, S. (2018), How Not to Plan: 66 Ways Not to Screw It Up, Matador. Leicestershire,
ISBN 978-1789014-501.
Bones, C., Ham Mersley, J. and Shaw, N. (2019), “Optimizing digital strategy: how to make informed”,
Tactical Decisions That Deliver Growth, Kogan Page. London, ISBN 978-0-7494-8781-2.
Bryl, S. (2016), “Attribution model with R (part 1: Markov chains concept)”, available at:
http://
analyzecore.com/2016/08/03/attribution-model-r-part-1/ (accessed 24 September 2018).
Clifton, B. (2015), Successful Analytics: Gain Business Insights by Managing Google Analytics, Advanced
Web Metrics, West Sussex, p. 336. ISBN 978-1-910591-00-0.
Danaher, P.J. and Van Heerde, H.J. (2018), “Delusion in attribution: caveats in using attribution for
multimedia budget al.location”, Journal of Marketing Research, Vol. 1, pp. 1-65.
Dinucci, D. (1999), “Fragmented future”, Print, Vol. 53 No. 4, p. 32, available at:
http://tothepoint.com/
fragmented_future.pdf
Ferencová, M., Jelenõová, I. and Kakalej
cík, L. (2015), “Social media usage in product information
searching”, Applied Mechanics and Materials, Vol. 795, pp. 69-76.
Finstat (2017), Dostupné na Internete,
https://finstat.sk/ (accessed 27 January 2018).
Geyik, S.C., Abhishek, S. and Dasdan, A. (2014), “Multi-touch attribution based buget al.location in
online advertising”, Proceedings of the Eighth International Workshop on Data Mining for
Online Advertising: ACM. New York, NY, pp. 1-9, ISBN 978-1-4503-2999-6.
Ghose, A. and Todri, V. (2015), “Towards a digital attribution model: measuring the impact of display
advertising on online consumer behavior”, MIS Quarterly, Vol. 1, ISSN 2162-9730.
Google Analytics (2018), Dostupné na Internete:
https://analytics.google.com/ (accessed 27 January
2018).
Impact of
direct traffic
effect
Kahn, B.E. (2013), Global Brand Power: Leveraging Branding for Long-Term Growth, Wharton Digital
Press. Philadelphia, p. 134. ISBN 978-1613630266.
Kakalej
cík, L., et al. (2018), “Multichannel marketing attribution using Markov chains”, in: Journal of
Applied Management and Investments, Vol. 7 No. 1, pp. 49-60.
Karlin, S. and Taylor, H. (1975), A First Course in Stochastic Processes, Academic Press, Cambridge.
Kim, H.K., et al. (2008), “Brand equity in hospital marketing”, Journal of Business Research, Vol. 51
No. 1, pp. 75-82.
Lemon, K.N. and Verhoef, P.C. (2016), “Understanding customer experience throughout the customer
journey”, Journal of Marketing, Vol. 80 No. 6, pp. 69-96.
Li, H. and Kannan, P.K. (2014), “Attributing conversions in a multichannel online marketing environment:
an empirical model and a field experiment”, Journal of Marketing Research, Vol. 51 No. 1, pp. 40-56.
Marshall, P. and Todd, B. (2017), Ultimate Guide to Google AdWords: How to Access 100 Million People
in 10 Minutes, Entrepreneur Press, Irvine, p. 380. ISBN 978-159918612-2.
Nissar, T. and Yeung, M. (2015), “Purchase conversions and attribution modeling in online advertising:
an empirical investigation”, EMAC 2015, Leuven.
Rentola, O. (2014), Analyses of Online Advertising Performance Using Attribution Modeling: diplomová
Práca, Aalto University, Aalto, p. 71.
Roberge, M. (2015), The Sales Acceleration Formula: Using Data, Technology, and Inbound Selling to Go
from $0 to $100 Million, John Wiley and Sons, Hoboken, p. 224. ISBN 978-1-119-04707-0.
Romaniuk, J. and Sharp, B. (2015), How Brands Grow: Part 2: Emerging Markets, Services, Durables,
New and Luxury Brands, Oxford University Press, Oxford: ISBN 9780195596267.
Shao, X. and Li, L. (2011), “Data-driven multi-touch attribution models”, Proceedings of the 17th ACM
SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM New York,
NY, pp. 258-264, ISBN 978-1-4503-0813-7.
The R Project for Statistical Computing (2016), “The R project for statistical computing”, available at:
www.r-project.org/. (accessed 19 August 2016)
Van Den Berg, G. and Pietersma, P. (2015), 25 Need to Know Management Models, Pearson. Harlow,
p. 224. ISBN 978-1-292-01635-1.
Vanden Heuvel, A. (2014), “5 Ways to nurture leads through the sales process, eCommerce style”,
Hubspot, 15 April, updated February 01 2017, available at:
https://blog.hubspot.com/insiders/
lead-nurturing-tips
Wang, Q., Li, B., Wang, P. and Yang, J. (2015), “When structural models meets big data: examining
multi-device attribution for native ads using TB-sized data”, Proceedings of Workshop on
Information Systems and Economics, available at:
https://research.yahoo.com/mobstor/
publication_attachments/LargeScaleStructualAnalysis.pdf
Wooff, D.A. and Anderson, J.M. (2015), “Time-weighted multi-touch attribution and channel relevance
in the customer journey to online purchase”, Journal of Statistical Theory and Practice, Vol. 9
No. 2, pp. 227-249.
Yadagiri, M.M., Saini, S.K. and Sinha, R. (2015), “A non-parametric approach to the multi-channel
attribution problem”, 16th International Conference, Springer International Publishing, Basel.
Corresponding author
Jozef Bucko can be contacted at:
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
JRIM
