=Paper=
{{Paper
|id=Vol-3179/Paper_9.pdf
|storemode=property
|title=Optimization of Portfolio Marketing Strategy Based on Data Science Technologies
|pdfUrl=https://ceur-ws.org/Vol-3179/Paper_9.pdf
|volume=Vol-3179
|authors=Galyna Chornous,Yana Fareniuk
|dblpUrl=https://dblp.org/rec/conf/iti2/ChornousF21
}}
==Optimization of Portfolio Marketing Strategy Based on Data Science Technologies==
https://ceur-ws.org/Vol-3179/Paper_9.pdf
Optimization of Portfolio Marketing Strategy Based on Data
Science Technologies
Galyna Chornous and Yana Fareniuk
Taras Shevchenko National University of Kyiv, 90-A, Vasylkivska st., Kyiv, 03022, Ukraine
Abstract
The article contains the results of Data Science realization for pharmaceutical market. The
main goals of research are modelling the sales of three key brands of one of the Ukrainian
pharmaceutical companies based on regression analysis as a main method and to make
conclusions for portfolio marketing strategy optimization and effective brand management.
Estimation the influence of key elements of the marketing mix on brand’s share of market
make basis for ROI calculations and optimization of media budget allocation between brands
in portfolio. The article contains recommendations for analytical system construction.
Keywords 1
Marketing strategy, pharmaceutical industry, optimization, modeling, marketing mix, Data
Science, regression
1. Introduction
The pharmaceutical market is one of the key sectors of Ukraine's economy that is able to meet the
basic needs of consumers, especially in terms of COVID-19. Today, the Ukrainian pharmaceutical
market is extremely competitive and more and more companies are looking for ways to find effective
marketing strategy management to maximize market success. At the present stage, there is an urgent
need for business entities to ensure consistency between market management practices, marketing
activities and their effectiveness.
Under the influence of global trends, the conditions of the pharmaceutical business are changing and
companies are faced with the need to implement new market strategies, methods of product range and
pricing, media support, reassessment of principles and directions of economic development, radical
changes in approaches to enterprise management, formation of a new system management based on
marketing support [1].
Strategy is a general program of action that identifies the priorities of problems and resources to
achieve the main goal (goals). The strategy forms the key goals and the main ways to achieve them in
such a way that the company gets a single course of action. Development of strategy of the enterprise
is carried out in conditions uncertainty. The external environment is characterized by the instability of
the factors that force the organization to change. Marketing strategy is a system of enterprise action in
which the internal environment is balanced with the external. The analysis of the criteria on the basis
of which specific marketing decisions are made, showed that they are multifaceted, determined by a
significant number of indicators. Therefore, there is a need to develop and implement a marketing
strategy that can increase the efficiency of the enterprise [2].
Marketing support acquires special importance in the work of a pharmaceutical company. It is able
to bring to the activities an understanding of consumer needs and creative thinking, which makes it
possible to be customer-oriented and provide high profits. At the same time, marketing decisions must
be substantiated with analytical and commercial clarity, correspond to the corporate strategy, innovative
directions of development and promote the implementation of efficient, stable and profitable operation
of the enterprise. In today's highly competitive market, pharmaceutical companies cannot operate
without the use of Data Science technology and in-depth work with data to develop an effective
marketing program, well-planned and organized marketing activities, which makes this study actual
and practical [1]. The aim of the study is to optimize portfolio strategy through formation of effective
Information Technology and Implementation (IT&I-2021), December 01–03, 2021, Kyiv, Ukraine
EMAIL: chornous@univ.kiev.ua (G.Chornous); yfareniuk@gmail.com (Y. Fareniuk)
ORCID: 0000-0003-4889-1247 (G. Chornous); 0000-0001-6837-5042 (Y. Fareniuk)
©️ 2022 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
89
marketing strategy for each brand in accordance to return on investments (ROI) of each product, media
channels, marketing mix elements efficiency. Modeling and analysis based on Data Science
technologies help to prioritize brands in portfolio, which generate opportunities for brand performance
growth. The secondary goal is forming information-analytical basis for marketing strategy
development, implementation and its evaluation.
2. Literature review
Theoretical and practical researches of many experts in this field are devoted to aspects of
development of effective marketing strategies. Leading foreign scholars G. Armstrong, F. Kotler,
P. Drucker, W. Wong, J. Lamben, O. Walker, J. Ansoff, D. Saunders and others studied the issues of
strategic management in marketing, the formation of competitive marketing strategies, their separate
elements, marketing support, etc. The development of theoretical, methodological and practical bases
for the formation of marketing strategies for the development of enterprises in Ukraine are engaged in
such domestic scientists as O. Kendyukhov, N. Butenko [3], N. Kudenko, A. Starostina and others.
The use of marketing in the activities of enterprises is considered in the works of such authors as
N. Demchuk, I. Tubolets, O. Kuzmenko, V. Pavlova, V. Polyakh, V. Pisarenko, T. Zaporozhchenko,
E. Sergeeva and others. The leading position in the research of the above scientists is to justify the
feasibility and necessity of using marketing research and data analysis in the activities of enterprises to
harmonize modern market requirements, as well as the needs and behavior of consumers [4]. Marketing
capital and R&D capital have a direct and positive impact on maintaining sales leadership. A sales
leader that gradually spends 1% of its average five-year sales revenue on marketing and research
activities can increase the likelihood of maintaining sales leadership by 50% [5].
Given the need to justify the costs of doing a business and their proper distribution among products,
companies are very interested in measuring the effectiveness of marketing. Adopting effective
marketing strategies is a difficult task, especially in a highly competitive market, which includes
complex and multifaceted marketing planning, methods and mechanisms to determine the best plan of
action in the current environment and find optimal solutions to achieve the best results.
The effectiveness of the firm depends on many elements. Marketing opportunities and strategies for
brand portfolio development make a significant contribution to the company's efficiency. Research [6]
develops and tests a framework for understanding the relationships between marketing opportunities,
marketing strategy, and marketing implementation. Understanding the relationships between these
variables can help managers implement a marketing strategy to maximize effectiveness. The main
results of the study show that the effectiveness of the marketing strategy positively mitigates the impact
of marketing opportunities on market efficiency and financial results [6].
The article [7] argues for the relationship between investment in marketing innovation, i.e. the way
in which products are developed, priced, distributed and promoted, and the performance of a new firm
product. Market activity and competitors' strategies create a significant influence on the choice of the
appropriate marketing strategy. M. Abedian et al. investigate the development of a mathematical
methodology based on the game theory approach to plan the optimal strategies of the marketing
complex in dynamic competitive markets, taking into account strategic forecasting and interaction
effects [8]. R. Sellers-Rudio et al. [9] examine the analysis of the effectiveness of investment in
advertising in the media using a two-stage analysis of the coverage of data with double loading to the
resources allocated to various advertising tools. The authors then explore the determinants of
advertising effectiveness in terms of the number of brands in the portfolio and the combination of
advertising tools used. The results show a certain level of waste on advertising media spending, as well
as the fact that both the size of the brand portfolio and advertising have a positive effect on efficiency.
C. Ten Caten et al. [10] offer a functional classification of enterprise product portfolio for use in
business planning, containing four functional classes: concept products, special purpose products,
profit-generating products and products in the mix composition. Based on this approach, managers must
identify the main features of their products and outline strategies that are unique to the business and
marketing goals of both individual products and the product portfolio as a whole. Classification is based
on two aspects, namely the share in the structure and margin. Thus, business planning, which involves
matching the product range with the expectations of customer behavior, allows the company to
maximize its general results.
Expanding product and brand portfolios are effective marketing strategies to meet customer’s needs
and create a competitive advantage in the market. However, product and brand portfolios can easily
90
spiral out of control, leading to a loss of market concentration and market share. A. Kirca et al. [11]
explore how a product and brand portfolio and such decisions influence unit sales and market share as
brand performance using a dynamic panel generalized method of moments estimation. The authors find
critical management trade-offs between brand and product portfolio decisions, as brand and product
portfolio decisions are intertwined, and careful research is necessary to maximize brand effectiveness.
The article [12] describes the implementation of a category management tool known as the Category
Optimizer ™ in Foster's Wine Estates Americas, which has faced the challenge of managing a wine
portfolio to increase profitability, improve competitive position and protect against a new competitor.
The Category Optimizer has led to recommendations that initially seemed contradictory. The classic
response should be to reduce prices to maintain share and volume, but it has been shown that combining
price increases with the introduction of a new product that flanks volumes in the new channel will
increase profits, increase revenue, protect and improve the market. High results were achieved after
implementation. Profitability increased by 70% and revenue by 3%.
It is well known that consumer behavior can affect the development of new products and,
consequently, the profits of firms. Therefore, companies must take into account consumer behavior
when planning marketing strategies. In [13], the authors propose a new model of diffusion to create
products, taking into account the diversity of their consumers and two key marketing elements: price
and advertising activity. It is shown that the optimal trajectory of pricing is a concave curve and there
is a threshold of the optimal trajectory of advertising. When customers are loyal, the best response from
companies is to reduce advertising spending. In addition, the company is able to reduce advertising
spending when the influence of word-of-mouth (WOM) advertising is high.
The aim of the study [14] is to assess the impact of pharmaceutical marketing strategies on the
behavior of physicians in prescribing. The overall significant impact of the strategies of the
pharmaceutical marketing complex on the behavior of physicians in prescribing was 55.9% (more than
half of physicians noticed such an impact). The influence of advertising strategy, product, distribution
and price strategy, perceived by doctors in their behavior when prescribing drugs, was 61%, 52%, 52%,
59%, respectively.
Portfolio management decisions are a key function of any drug company. The article [15] discusses
what constitutes a sufficient portfolio of drugs for a firm and how it can be successfully managed. The
authors explore the decision-making process from a variety of perspectives, including the importance
of the drug portfolio, tools for portfolio analysis, and the optimal portfolio. Drug portfolio decisions
using modern decision-making methods should be used much more widely in modern practice.
Improvement of management of drug portfolio is key element of an overall success strategy.
The following domestic and foreign scientists have made a significant scientific contribution to the
study of the problems of effective formation and implementation of marketing strategy: I. Alekseev,
S. Garkavenko, G. Assel, V. Gerasimchuk, S. Ilyashenko, S. Kovalchuk, Z. Shershneva, G. Bagiev,
D. Day, R. Fatkhudinov. In their works the issues of marketing management of business processes of
the enterprise are considered; essence and content of the concept of "marketing strategy" and analysis
of its elements. One of the factors that determine the effectiveness of marketing activities of enterprises,
scientists recognize the in-depth analysis of market data to find hidden patterns in them and the
formation of effective marketing decisions [2]. M. Stone and others investigate the impact of modern
interactive marketing on customer information and marketing research, in particular how radically
changed the ways of collecting and using customer information to develop and implement marketing
strategies and individual elements of the marketing mix. Researchers note the development of platforms
that allow companies to manage their information and interaction with customers in radically different
ways through the development of business intelligence [16].
D. Johnson et al. [17] explore the experience of marketing departments that are completely guided
by data in decision-making. The survey data confirm that Big Data Analytics consists of four main
activities: gaining knowledge, improving the quality of data, testing and disseminating the results of
data analysis. This study shows that the shift to analytics improves the quality of resources available to
the marketing department and provides a model for improving the quality of marketing information and
improving data-based decision-making.
Marketing mix modelling is the most commonly used method, which helps to determine the
efficiency of marketing mix elements and consists of building a regression (econometric) model based
on historical data to describe business performance indicators (sales or market share) as a functional
dependence from different factors, such as distribution, price, advertising activity (number of
impressions, target ratings, etc.), seasonality, macroeconomic factors, competitor’s activity and so on.
The contribution of classic machine learning methods like regression analysis to marketing decision-
91
making is quite important, but there are alternative methods. J. Dawes et al. [18] research evidence-
based methods that have been shown to be useful for forecasting. Y. Jin et al. [19], S. Zhang and
J. Vaver [20] recommend using Bayesian hierarchical modelling. According to research D. Chan and
M. Perry [21], the potential of marketing mix modelling may be limited by the lack of detailed and
qualitative data. To solve this problem, they propose to develop better data and models, as well as test
them, using simulations as the main instruments of improvement for marketing mix modelling.
Pharmaceutical market has a lot of high-quality data, so marketing mix modeling is a method, which
show good performance in solving of business tasks and is extremely important.
R. Birn and M. Stone explore the evolution of marketing information systems and the need for
companies to use marketing information to understand the needs and behavior of potential customers,
as well as to monitor it. The authors provide advice on key issues related to how marketing research
helps companies make decisions, information on classic marketing research methods, and what methods
companies need to grow, solve problems, and manage their portfolios effectively. The issue of
forecasting marketing results is also covered. The authors argue that the development and management
of customer analysis systems and marketing information is a critical tool for managing the company's
portfolio, which is used to support marketing strategy and marketing solutions [22].
But many questions about the peculiarities of marketing strategies of enterprises remain
insufficiently studied, in particular the peculiarities of forming an effective portfolio marketing strategy
through determining the ROI of each product and their prioritization, as well as some issues of forming
information-analytical basis for marketing strategy development and implementation, its evaluation
needs further development [2].
3. Methodology
The main goal of the research is to find effective recommendation for marketing strategy of key
brands in portfolio of one leading pharmaceutical company in Ukraine.
As the aim of the study is to optimize portfolio strategy, the system analysis was used. To achieve
the objectives, general scientific methods were used in this scientific paper. Logical generalization – to
identify trends in the brand’s development; comparison – to analyze differences in marketing
performance by different brands in portfolio, compare Return on Investment (ROI) by brands; induction
– to define the main beneficial Data Science methods in marketing based on previous researches;
analysis and synthesis – to analyze the theoretical and practical models used in marketing researches
for pharmaceutical companies in terms of increasing its profitability; methods of multicriteria analysis
and modeling - econometric modeling - for understanding and quantifying the impact of marketing mix
elements and market conditions on sales level of key brand of one of the Ukrainian pharmaceutical
companies based on regression analysis (Data Science and machine learning technologies in general),
as well as the recommendations formation for improvement the efficiency of marketing activities of the
companies on the pharmaceutical market and optimizing the business results depending on the chosen
marketing strategy of a separate product and for all product line in general.
To achieve this goal were used regression analysis and economic and mathematical modelling (based
on machine learning technologies) of the dependence of volume and value market share (sales level) of
certain medications on the such factors as: penetration of pharmacy stores (number of pharmacy stores,
where such drug is available for potential brand consumers, and share of these stores in general
pharmacy network in Ukraine), price policy (the rate of the drug’s price compared to its market/category
competitors), brand’s and its competitors’ advertising activity in all communication (media) channels
(television, digital, radio, out of home advertising and press), recommendations of doctors and
pharmacists and others [23]. Pharmaceutical companies have specific marketing mix (in comparison to
Fast Moving Consumer Goods market) as include work with doctors and pharmacists and realize
necessary penetration of pharmacy network.
Data on all indicators for the period from 2017 to 2021 were collected in a weekly breakdown for
all drugs in the relevant categories (database was collected on the basis of data of the pharmaceutical
company from Proxima Research [24], as well as the Industrial Television Committee [25] and Nielsen
Ukraine [26]). Data about media activity and other marketing variables (sales, penetration, price, etc.)
was collected from special monitoring softwares Markdata and Morion respectively on the basis of SQL
requests for each specific category. Due to confidentiality, all data in the article will be indexed from 0
to 1. All data was collected in excel file, which was integrated in R-Studio software for modeling.
92
For the constructed multiple regression (econometrics) models for each brand, which were estimated
by the method of least squares using R-Studio software, main hypotheses about the adequacy of the
models, the significance of the coefficients for all factors, the presence of heteroskedasticity and
autocorrelation were tested. Also, for the formation of recommendations for optimal price policy, an
approach based on machine learning technologies (coefficients from constructed regression models and
additional data-based calculations) was developed to estimate the optimal range of price index (drug’s
price in comparison to the average category price of competitors) to maximize market share in volume
or market share in value depending on the company's goals.
Estimation the impact of media factors by each brand, comparison of such influence by brands in
terms of profitability help to develop recommendations for media strategy of these drugs, prioritize
brands by media support and marketing budget allocation and as a result increase ROI of
media/marketing investments for each brand separately and for portfolio in general [18].
For day-by-day integration of Data Science into marketing strategy planning we realize
informational and analytical system based on Power BI interactive dashboard with integration of
previously accumulated knowledge based on econometric modeling. Such informational and analytical
system was called “Brand’s Marketing Management System”.
Project was deploying accordingly to the most widely-used analytics model CRISP-DM [27, 28].
CRISP-DM describes the process through 6 stages: “Business Understanding, Data Understanding,
Data Preparation, Modelling, Evaluation and Deployment” [28]. In this project, the overall process was
implemented through step-by-step analysis and modeling, as shown in the Figure 1.
Figure 1: Proposed concept of modelling marketing mix project realization
The process includes the ability to flexibly change the order of the stages, going back when the need
arises. Data mining has cyclical nature, as the process of finding solutions continues throughout the
project and after it has been deployed. The main learnings, experience and knowledge of the previous
cycle can create new, deeper business issues that have a strong positive influence on future data mining
processes [28].
4. Results
Data Science is a relevant tool for searching effective business solution for portfolio marketing
strategy optimization. The research of marketing activities of several pharmaceutical brands in Ukraine
in three product categories will show how on the basis of economic-mathematical modelling (regression
analysis) and other Data Science technologies we can analyze what factors and to what extent affect the
business results of the company (sales or market share in the relevant category).
Understanding the influenced factors enables to calculate the elasticity of sales to each factor and as
a result to distinguish the optimal range for each of them taking into account their marginal utility (for
example, calculating price elasticity and finding the optimal value of the price index - the ratio of brand
price and average price of competitors or determining the optimal level of media pressure in each
communication channel). In addition, understanding the effectiveness of each communication channel
make it possible to develop an effective media strategy and tactics for each brand.At the first stage of
working with a business task, it is necessary to study in detail the features of the brand and the category
for which the analysis is implemented to determine all the factors that influence on the final performance
of the business. In this study, we examine three brands that are the one of the main players in the
93
categories of diaper dermatitis, pain and spasmolytics, tablets and lozenges for Cold&Flu respectively.
Brands actively use TV and Digital support to promote these products for potential consumers.
For each brand of drug, we will construct the multiple regression model, which looks like this:
𝑆𝑂𝑀 = 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡 + 𝑎1 ∗ 𝑃𝑒𝑛𝑒𝑡𝑟𝑎𝑡𝑖𝑜𝑛1 + 𝑎2 ∗ 𝑃𝑒𝑛𝑒𝑡𝑟𝑎𝑡𝑖𝑜𝑛2 + 𝑎3 ∗ 𝑃𝑟𝑖𝑐𝑒𝑖𝑛𝑑𝑒𝑥 + 𝑎4 ∗
𝐷𝑜𝑐𝑡𝑜𝑟𝑠 + 𝑎5 ∗ 𝑃ℎ𝑎𝑟𝑚𝑎𝑐𝑖𝑠𝑡𝑠 + 𝑎6 ∗ 𝐴𝑑𝑠𝑡𝑜𝑐𝑘(𝑇𝑉1 ) + 𝑎7 ∗ 𝐴𝑑𝑠𝑡𝑜𝑐𝑘(𝑇𝑉2 ) + … + 𝑎𝑛 ∗
𝐴𝑑𝑠𝑡𝑜𝑐𝑘(𝑇𝑉𝑛 ) + 𝑐1 ∗ 𝐴𝑑𝑠𝑡𝑜𝑐𝑘(𝐷𝑖𝑔𝑖𝑡𝑎𝑙_𝑣𝑖𝑑𝑒𝑜) + 𝑐2 ∗ 𝐴𝑑𝑡𝑜𝑐𝑘(𝐷𝑖𝑔𝑖𝑡𝑎𝑙_𝑑𝑖𝑠𝑝𝑙𝑎𝑦 ) + 𝑏1 ∗
𝐴𝑑𝑠𝑡𝑜𝑐𝑘(𝑇𝑉_𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑜𝑟1 ) + . . . + 𝑏𝑚 ∗ 𝐴𝑑𝑠𝑡𝑜𝑐𝑘(𝑇𝑉_𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑜𝑟𝑚 ), (1)
where
SOM – share of market in Volume (in %);
Penetration1 and Penetration2 – penetration level for SKU 1 and SKU 2 respectively (in %);
Doctors or Pharmacists – level of brand’s recommendations by doctors and pharmacists (in %);
TV1 or TV2 or TVn – TV activity of brand by different creative materials (TV1 is for creative 1 and
TVn is for creative n respectively in TRPs – target rating points in %);
Digital_video and Digital_display – digital activity of brand in video and display formats in
Internet respectively (in impressions);
TV_Competitor1 or TV_Competitorm – TV activity of competitor 1 or competitor 2 respectively
(in TRPs);
Adstock is the instant, prolonged and lagged effect of advertising activity on consumer purchase
behavior, which indicate influence of TV activity (in TRPs) during a time.
𝐴𝑑𝑠𝑡𝑜𝑐𝑘(𝑇𝑉𝑡 ) = 𝑇𝑉𝑡 + 𝑎 ∗ 𝐴𝑑𝑠𝑡𝑜𝑐𝑘(𝑇𝑉𝑡−1 ) (2)
The first product category is newborn health care. The brand of this pharmaceutical company has
strong position on the market (enter the TOP-5) and started media support since second half of 2020.
The main specific of this category is high level of target audience’s rotation on a monthly basis. Every
month around 24 thousand of moms enter to the category and there is marketing task for brand to recruit
it to the brand’s consumption [29].
In general target audience is quite wide and we segment it by their behavior and consumer journey:
Pregnant woman on third trimester;
Moms with kids 0-3 months old;
Moms with kids 4-6 months old;
Moms with kids 6-12 months old;
Moms with kids 13-18 months old;
Moms with kids 19-36 months old.
Pregnant woman on third trimester and moms with kids 0-3 months old are newcomers. Main
marketing tasks for this group of target audience are build awareness, drive product trial, so it relevant
to regular media reach and recruiting. Moms with kids 0-18 months old are core of target audience and
main marketing tasks are maintaining and switching from competitors. For group of moms with kids
18+ months old we need to drive product experience, establish credibility and support advocacy. This
group is category’s graduates and we need to turn it into brand ambassadors.
Based on econometric model for our brand we calculate different scenario forecasting to optimize
architecture of media campaigns taking into account memory decay [30, 31]. Memory decay is 1-a from
equation (2) and estimate memory decay of advertising message among target audience. As a result of
simulation four variants of media architecture (continuity – every week placement with low media
pressure, pulse – one week on air and one week of break with high media pressure, pulse 2.1 – 2 weeks
on air and 1 week of break and burst with 5-weeks flights) we recommend for this brand burst strategy
with 5 weeks on air and 2 weeks of break, which helps to increase advertising effect (AdEffect) by 5%
comparison with continuity strategy (Figure 2).
After the realization of such recommendation, we estimate the current efficiency from media activity
based on econometric modeling and calculate ROMI (return on media investments) for this brand
according to equations (3)-(5) [32].
𝑅𝑂𝑀𝐼 𝑔𝑒𝑛𝑒𝑟𝑎𝑙 = 𝑆𝑎𝑙𝑒𝑠 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑓𝑟𝑜𝑚 𝑚𝑒𝑑𝑖𝑎 𝑖𝑛 𝑈𝐴𝐻 / 𝐺𝑒𝑛𝑒𝑟𝑎𝑙 𝑀𝑒𝑑𝑖𝑎 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡 (3)
𝑅𝑂𝑀𝐼 𝑇𝑉 = 𝑆𝑎𝑙𝑒𝑠 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑓𝑟𝑜𝑚 𝑇𝑉 𝑖𝑛 𝑈𝐴𝐻 / 𝑇𝑉 𝑀𝑒𝑑𝑖𝑎 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡 (4)
𝑅𝑂𝑀𝐼 𝐷𝑖𝑔𝑖𝑡𝑎𝑙 = 𝑆𝑎𝑙𝑒𝑠 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑓𝑟𝑜𝑚 𝐷𝑖𝑔𝑖𝑡𝑎𝑙 𝑖𝑛 𝑈𝐴𝐻 / 𝐷𝑖𝑔𝑖𝑡𝑎𝑙 𝑀𝑒𝑑𝑖𝑎 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡 (5)
94
Results of modeling show that media activity has strong positive influence on sales level and share
of market. More than 25% of sales was accumulate in period after media support start in 2020-2021
years (Figure 3). In period of 2021 media boost is 22,3%.
Figure 2: Result of scenario forecasting
Figure 3: Model decomposition for brand 1
The next brand is one of the key players in pain and spasmolytics category, which is a quite clattered
as presented on the market by more than 60 brands. Taking into account such highly competitive
environment we need to find effective media strategy based on all available data on the market and Data
Science technology [33]. We also construct econometric model for this brand and realize scenario
forecasting for optimal media campaigns architecture and strategy. Since 2020 we change the media
strategy from burst to pulse and there was a significant growth of media efficiency. Brand is in TOP-3
by brand awareness in this category and “week on – week off” strategy shows high efficiency for sales
growth. In such conditions one week of break is not critical as a result of low memory decay for this
brand. After such change of media strategy, we save more than 40% of investments and ROMI of media
activity show significant growth. The contribution of media activity to sales doubled from 8% in 2018-
2019 to 15% in 2020-2021 with simple level of investments (Figure 4).
The third brand is one of the key players in tablets and lozenges category, which is competing with
others subcategory of Cold & Flu market. The main media task is to optimize marketing mix to increase
return on marketing investment in general. We continued to construct econometric model for this brand
also. The main specific was that we determine based on Data Science technologies strong synergy effect
between two key elements of marketing mix (price and promotion). This brand use media activity on
2017 and 2019-2021, but media contribution in 2017 was very low (Figure 5).
The main reason of such changes is high level price sensitivity in category and strong influence of
price on media response. In conditions of high level of price index (ratio of price per one brand’s
package to average price on the market in this category per one package), price become a barrier for
consumer to make purchase and interested potential consumers after media support switch on cheaper
competitors in pharmacy stores [34]. As a result, we stimulated only brand awareness, which didn’t
convert to sales. The absolute influence of price index also quite significant during a modeling period
and shown as negative factor on the Figure 6.
95
Figure 4: Model decomposition for Brand 2 and media impact in 2017-2021 years
In general, for brands in category there are strong influence of price to sales and media response.
So, the Figure 6 shows the price sensitivity influence on media efficiency. Brand 1, brand 2 and brand
3 are the cheapest brands and have the highest efficiency from media activity. Brand 8, 9, 10 are the
most expensive and their media activity generates the lowest effect on sales. It is necessary always
analyze current price sensitivity to optimize brand’s marketing strategy [35].
On the basis of the constructed economic and mathematical model of the drug’s sales, we can
estimate the curves of value share of market (money) and volume share of market (packages) depending
on the level of price index, as the coefficient of model at the price index indicates how share of market
in volume will change with rising price index by 1 unit (the character of the connection is linear or non-
linear in the case of a linear or non-linear regression respectively).
Figure 5: Model decomposition for Brand 3 and media impact by 2017-2021 years
The share of market in money and the share of market in packages are linked by a price index. Value
market share is volume market share multiplied by price index. Due to the fact that the rate of change
of share of market in volume (linear in our case) doesn’t coincide with the rate of price index change
(also linear, but with another growth rate), there is a non-linear relationship between the price index
and share of market in monetary terms (value market share), which leads to an optimization zone
depending on business goals. On the case of our brand of drug, the optimal value of the price index to
maximize value market share is from 1.8 to 2.3 (Figure 7), which means that the drug should be more
expensive up to +80-130% to average prices on the market to obtain the maximum level of profit.
Depending on the business’ objectives, this methodology becomes a flexible and convenient tool for
96
the price policy specialists, as it is possible to form a recommendation for price level to achieve the
goals of both value share of market and volume share of market.
Figure 6: Price sensitivity in category and their influence on media response
Figure 7: Optimal level of price index in terms of growth of share of market in value for Brand 3
In terms of media factors contribution, TV activity is a main sale generated channel with highest
ROMI. High performance shows Out of Home advertising (OOH), which brand used at the end of 2020
to regional sales support. General media contribution to sales in 2020-2021 is 12%, which is a little bit
higher than average market benchmarks. In period of 2021 media boost is 18,9%. Among media
channels contribution is distributed by the next proportions: 79% of media effect is accumulated by TV,
11% by digital and 10% by OOH advertising.
For these three brands the results of portfolio strategy are summarized in the Table 1, which shows
the media boost in percentages and Brand Priority Index in company portfolio. Media boost indicates
the share of sales generated by media activity. After converting such indicator into sales in monetary
terms and comparison with realized media investment via equations (3)-(5), we may to prioritize brands
in portfolio according to their general media performance indicated by ROMI.
Table 1
Brand Portfolio Strategy Results
Parameter Brand 1 Brand 2 Brand 3
Media boost, % 22,3% 16,2% 18,9%
2021 YTD
Brand Priority 2 1 3
in Portfolio
To maximize the brands’ efficiency, we recommend to implement Data Science instruments into
day-by-day brand management. The effective concept of such implementation of modelling marketing
mix project is cyclic process, which accumulates next stages:
Obtaining data on the sales and marketing mix elements and forming a single database;
Updating the model, evaluating the effectiveness of previously made decisions and analyzing
current results;
97
Formation of recommendations for the strategy and forecast of business indicators for the next
periods;
Optimization of investments taking into account business KPI / checking the implementation of
sales plans (forecasting);
Implementation of media and business solutions.
In case of regular support of marketing mix modeling projects for company’s portfolio we may to
determine business tasks depending on different time intervals (weekly, monthly quarterly). Main tasks
for each time intervals are structured in the Table 2.
The next steps of Data Science integration into marketing strategy planning are to realize
informational and analytical system with integration of previously accumulated knowledge based on
econometric modeling. For these brands was developed such informational and analytical system called
“Brand’s Marketing Management System” with structure, which presented in the Table 3.
Table 2
Marketing Mix Modeling and Data Science tasks in accordance to time intervals
Weekly Monthly Quarterly
1. Forecasting of business 1. Monitoring the 1. Media placement strategy
KPIs implementation of sales (media mix, period and weight),
2. Reporting and plans taking into account minimizing
information exchange 2. Tracking the dynamics of costs while achieving goals and
3. Fast response to any indicators maximizing efficiency
changes and risks in order 3. Recommendations based 2. Evaluating the effectiveness of
to manage business results on the current market previous campaigns, evaluating and
situation comparing creatives
Table 3
Structure of informational and analytical system “Brand’s Marketing Management System”
Business review Media review Econometric modeling Portfolio strategy
Market and brand TV, Digital, Modeling results Portfolio
dynamics OOH, Radio, Price elasticity analysis
Market structure analysis Press ROI by media channel Media Boost by
Sales vs media analysis Seasonal ROI brands
Regional analysis Forecasting
5. Conclusions
Such results for three main brands in portfolio of pharmaceutical company become a basis for
portfolio strategy optimization as we clearly understand and quantitatively estimate the current and
potential effect from all elements of marketing mix complex. Portfolio marketing strategy optimization
based on data-driven decisions has strong impact on business performance due to high quality and great
validity of decision-making process.
Thus, the modelling and continuing process of deeply data analysis become a convenient and an
effective tool for making effective strategic and tactical marketing and media decisions:
regularly evaluation an optimal price taking into account the level of competitors' prices and its
effect on sales;
monitoring the efficiency of the advertising campaign in all media channels;
forecasting of general sales level and sales level by separate brand at different scenarios of
advertising activity;
forecasting of sales level for different combinations of marketing factors (realizing of scenario
forecasting);
calculation of optimal level of each factor taking into account the marginal efficiency of each of
them and ROMI comparison by channels, brands, periods, etc.;
effective allocation of marketing budgets between brands, instruments, periods, etc.;
analysis of the influence of factors on brand sales, evaluation of winnings and losses.
98
Taking into account information from the Table 1, the Brand 2 have the highest priority as generates
the largest ROMI. The next one is Brand 1 and third priority is for Brand 3. The marketing task is to
allocate marketing budgets accordingly to this priority, which helps to increase general profit of
pharmaceutical company. TV is a main sales driver for all pharmaceutical brands, but Digital and OOH
also show good performance with positive ROMI.
Technologies of Data Science is an effective tool for sales management, as they create the
opportunities to quantify the influence of each factor on sales, estimate their optimal mix for
achievement of business KPIs and improvements the company's position in the market, effective
marketing or media budgets allocation and scenario forecasting. Regular model support makes it
possible to increase the return on each factor, improve ROMI and ensure the achievement of business
goals in the most efficient way. Data Science creates the basis for making effective marketing solutions
and forming an effective business development strategy.
Implementation of Data Science projects has significant economic effect on business efficiency as
helps to improve return on marketing investment by more than 5% and forecast future business KPIs
with average error less than 5%. Such models and informational and analytics system are recommended
to applicate in real business practice of Ukrainian and international pharmaceutical companies for
finding hidden insights in category and product development, optimization of allocation of marketing
budget per portfolio and increasing of business KPIs. Data-driven marketing improve the efficiency of
marketing decision on day-by-day basis as we currently work in very dynamically developing
conditions on the market. The next stages of scientific research in such area are deeply analysis and
mathematical modeling of business results strategy of main competitors for each brand in portfolio
depending on marketing and media for finding additional insights, which can be effectively
implemented into future marketing activities.
6. References
[1] M. Bagorka, Y. Putinova, Marketing support of the industrial enterprise, Notes of Tavriya National
University: "Economics and Management" (2019), Т. 30(69), № 3, p. 59-66. (In Ukrainian).
[2] L. A. Kvyatkovska, Mechanism of evaluation of marketing strategy, Marketing and innovation
management (2011), № 4, Vol. II, p. 98-104. (In Ukrainian).
[3] N.A. Butenko, V. Kachalay, Brand dynamics approach to the product portfolio management on a
dairy products market, The Global Challenges For Environmental And Resource Economics In
Central And Eastern European Countries: Safety, Security, And Sustainability: 2d Intern. сonf.
(Kyiv, 2014), p. 56.
[4] I.A. Abramovych, Marketing Strategy, Scientific Bulletin of Kherson State University: Economics
and Management of Enterprises (2016), volume 21, p. 50-53. (In Ukrainian).
[5] G. Vadakkepatt, V. Shankar, R. Varadarajan, Should firms invest more in marketing or R&D to
maintain sales leadership? An empirical analysis of sales leader firms, Journal of the Academy of
Marketing Science (2021), p. 49. DOI: 10.1007/s11747-021-00774-2.
[6] L. Charles, G. Frankwick, Marketing capability, marketing strategy implementation and
performance in small firms, Journal for Global Business Advancement (2017), Volume 10, p. 327-
345. DOI: 10.1504/JGBA.2017.084612.
[7] C. Grimpe, W. Sofka, M. Bhargava, R. Chatterjee, R&D, Marketing Innovation, and New Product
Performance: A Mixed Methods Study, Journal of Product Innovation Management (2017), 34(3),
p. 360-383. DOI: https://doi.org/10.1111/jpim.12366.
[8] M. Abedian et al., A game theory approach to selecting marketing-mix strategies, Journal of
Advances in Management Research (2021). DOI: http://dx.doi.org/10.1108/jamr-10-2020-0264.
[9] R. Sellers-Rubio, A. Calderon-Martinez, Brand strategy scope and advertising spending: The more
the better? Tourism Economics (2021), vol. 27(1), p. 70-85, DOI:
https://doi.org/10.1177/1354816619882135.
[10] C.S. Ten Caten, O.S. Paleo, Functional classification of the product portfolio of companies,
Espacios (2012), Volume 33 (9), p. 12.
[11] A. H. Kirca, P. Randhawa, M. Berk Talay, M. Billur Akdeniz, The interactive effects of product
and brand portfolio strategies on brand performance: Longitudinal evidence from the U.S.
automotive industry, International Journal of Research in Marketing (2020), Volume 37, Issue 2,
pp. 421-439. DOI: https://doi.org/10.1016/j.ijresmar.2019.09.003.
99
[12] A. Sinha, A. Sahgal, S.K. Mathur, Practice prize paper Category Optimizer: a dynamic-assortment,
new-product-introduction, mix-optimization, and demand-planning system, Marketing Science
(2013), 32(2), pp. 221-228, DOI: https://doi.org/10.1287/mksc.1120.0746
[13] S. Najafi Ghobadi, J. Bagherinejad, A. Taleizadeh, Modeling the diffusion of generation products
in the presence of heterogeneous strategic customers for determining optimal marketing-mix
strategies. Computers & Industrial Engineering (2021). Volume 160. DOI:
https://doi.org/10.1016/j.cie.2021.107606.
[14] A.D. Hailu, B.D. Workneh, M.H. Kahissay, Influence of pharmaceutical marketing mix strategies
on physicians' prescribing behaviors in public and private hospitals, Dessie, Ethiopia: a mixed
study design, BMC Public Health (2021), 21(1):65. DOI: 10.1186/s12889-020-10063-2.
[15] A. Jekunen, Decision-making in product portfolios of pharmaceutical research and development –
managing streams of innovation in highly regulated markets. Drug design, development and
therapy (2014), volume 8, pp. 2009-2016. DOI: https://doi.org/10.2147/DDDT.S68579.
[16] M. Stone, E. Aravopoulou, N. Woodcock, P. Laughlin, R. Stott, Interactive marketing, customer
information and marketing research, The Routledge Companion to Marketing Research (2021), p.
38-54. DOI: 10.4324/9781315544892-4.
[17] D.S. Johnson, D. Sihi, L. Muzellec, Implementing Big Data Analytics in Marketing Departments:
Mixing Organic and Administered Approaches to Increase Data-Driven Decision Making,
Informatics (2021), Volume 8, p.66. DOI: https://doi.org/10.3390/informatics8040066.
[18] J. Dawes, R. Kennedy, K. Green, Forecasting advertising and media effects on sales: Econometrics
and alternatives, International Journal of Market Research (2018), Vol. 60, No. 6, pp. 611-620.
DOI: https://doi.org/10.1177/1470785318782871.
[19] Y. Jin, Y. Wang, Y. Sun, D. Chan, J. Koehler, Bayesian Methods for Media Mix Modeling with
Carryover and Shape Effects, Technical report, Google Inc. (2017).
https://research.google/pubs/pub46001/.
[20] S. Zhang, J. Vaver, Introduction to the Aggregate Marketing System Simulator, Technical report,
Google Inc. (2017). https://research.google/pubs/pub45996/.
[21] D. Chan, M. Perry, Challenges and Opportunities in Media Mix Modeling, Technical report,
Google Inc. (2017). https://ai.google/research/pubs/pub45998.
[22] R. Birn, M. Stone, Key issues in managing marketing research and customer insight, The
Routledge Companion to Marketing Research (2021), p. 219 – 243.
[23] D. Bowman, H. Gatignon, Market Response and Marketing Mix Models: Trends and Research
Opportunities, Foundations and Trends® in Marketing (2010), Vol. 4, No. 3, pp. 129-207. DOI:
http://dx.doi.org/10.1561/1700000015.
[24] Website of Proxima Research, https://proximaresearch.ua/en/.
[25] Website of Television Industry Committee, http://www.itk.ua/en/root/index/.
[26] Website of Nielsen Ukraine, https://www.nielsen.com/ua/uk/.
[27] C. Shearer, The CRISP-DM model: the new blueprint for data mining. J Data Warehousing (2010),
volume 5, p. 13-22.
[28] O. Chernyak, P. Zaharchenko, Data mining: Textbook, Znannya, Kyiv, Ukraine, 2014.
[29] Website of State Statistics Service of Ukraine, http://www.ukrstat.gov.ua/.
[30] A. Pergelova, D. Prior, J. Rialp, Assessing advertising efficiency, Journal of Advertising (2010),
Vol. 39/3. DOI: https://doi.org/10.2753/JOA0091-3367390303.
[31] M. Kizim, O. Geiman, Scenario modeling of development of social and economic systems:
directions, features and mechanisms, Regional economics (2009), №4, pp. 16-23. (In Ukrainian).
[32] J. Büschken, Determinants of Brand Advertising Efficiency: Evidence from the German Car
Market, Journal of Advertising (2007), Vol. 36, No. 3, pp. 51-73. DOI:
https://doi.org/10.2753/JOA0091-3367360304.
[33] R. Batra, K. Keller, Integrating marketing communications: New findings, new lessons, and new
ideas, Journal of Marketing (2016), 80(6), pp. 122–145. DOI: https://doi.org/10.1509/jm.15.0419.
[34] A. Farm, Pricing in practice in consumer markets, Journal of Post Keynesian Economics (2020),
Vol. 43:1, pp. 61-75. DOI: https://doi.org/10.1080/01603477.2019.1616562.
[35] S. Osypenko, T. Romanchyk, S. Pisarevsky, Substantiating the Prices for Enterprise Products on
the Basis of Optimization Models, Business Inform (2020), №6, p. 145-151. DOI:
https://doi.org/10.32983/2222-4459-2020-6-145-151.
100