Modernization of higher education and increased competition in the educational services market require new ways of improving the efficiency of higher education institutions (HEI) management. The generalized structure of forming the HEI video marketing strategy is developed and a series of actions for achieving goals is presented. An algorithm for the targeted advertising forming based on classification tree using R language is proposed. The research methodology is described and HEI targeting model is developed employing the classification tree and contextual advertising campaign data on Facebook. The results of the pilot case study are presented, which confirm the practical usefulness of the proposed targeting model, which allows forming target groups.
Today, marketing is a necessary attribute of every higher education institution (HEI).
The new Law on Higher Education [28] in Ukraine causes changes in the field of education: the competition between HEIs increases, the requirements and values of consumers (target audiences) have changed significantly. All these factors stimulate the need to organize effective HEI marketing advertising activities. Obviously, a similar problem is relevant for the world education system.
In order to promote education services advertising, HEIs must use modern information and communication technologies, integrated into a well-designed and strategically thought-out system.
One such tool is video marketing, which might be an important part of the overall
HEI marketing strategy. For example, recent Ericsson data [
Let us give you some statistics on how popular video marketing has become in recent years [36]: video is projected to claim more than 80% of all web traffic; adding a video to marketing emails can boost click-through rates by 200-300%; embedding videos in landing pages can increase conversion rates by 80%; 90% of customers report that product videos help them make purchasing decisions; 87% of online marketers are currently using video content in their digital marketing strategies; a single minute of video content is the equivalent of 1.8 million words.
In this regard, we can assume that the development of HEI marketing strategy should be focused on optimizing the work with the brand, effective use of advertising tools and communication elements, including social networks [34], synthesis of innovative marketing concepts in the field of education [27]. Thus, exploring ways to improve video marketing performance is one of the most promising areas in HEI marketing strategy. 2
Paper [33] focuses on the growing importance of online marketing, including research of the state of the art through analysis of the data provided by numerous surveys. It also contains brief description of the online marketing itself, basic strategies on the internet nowadays. The study [22] evaluated the main content of YouTube videos of Spanish brands that were included in the 2015 Interbrand rating. There were estimated 900 videos published by 30 Spanish brands belonging to different business sectors.
The authors [
The authors [
Paper [
The results of the study [
The purpose of [
The attention in [
The work [
Article [21] examines a technology adoption model to assist the inventor of an online video marketing platform in assessing the behavioral intention to use special online video marketing platforms for small and medium-sized businesses. Article [23] examines the problem of social media marketing of user-created instant music videos. A hybrid variable scale clustering algorithm (HVSC) is proposed to analyze user features using text and video content. A study [24] offers critical managerial perceptions of social media marketing on how to use both FGC and UGC in managing the marketing funnel and brand reputation.
Work [26] analyzed the general evolution of this modern advertising type, which conveys sociocultural values and creates school identity, examines different ways of use, norms, and diversity of the images offered by some educational institutions. The study [30] investigates how to enhance geographical targeting by a suite of other targeting strategies, including behavioral targeting, temporal targeting, and use of discount in an online-to-offline commerce context, to form a more comprehensive contextual targeting strategy. However, there is no specification of this study results.
Work [31] proposes a two-step method based on the Gaussian filter and decision tree (M-GFDT). The Gaussian filter corrects the business data distribution in the first stage, and classifies the decision tree. The decision tree is a widespread approach to identifying and visualizing of logical patterns in data. The decision tree construction algorithm, first proposed by Quinlan [37], operates on the principle of recursive partitioning of the dataset and incremental tree construction [29].
Multifaceted marketing ads include a look-alike model that reads data on user
behavior in networks. For example, in [
It should be noted the above-mentioned work mostly analyzes user actions in response to video marketing. Besides a number of similar works require relatively sophisticated tools for their implementations, that is, the question of simplifying the formation process and, accordingly, making targeted management decisions in the formation of video marketing strategy is still relevant.
In this regard, a goal of this paper is to develop targeting models for HEI video marketing based on the classification tree (dendrogram). Unlike analogues [30, 31], a model of video marketing targeting based on dendrogram allows making changes by dendrogram branches in the advertising campaign strategy by attributes, which the objective function depends on. Moreover, the use of the rpart library in the R programming language makes it possible to clean quickly and filter data, which makes it easier to target groups forming compared to [31]. The novelty of the work is the formation of the most favorable (in economic terms) target group for HEI video marketing, which will reduce the advertising campaign cost. 3
Based on the authors' experience in targeting the advertising of the admission campaign and recommendations [33-35], a generalized algorithmic structure for forming HEI video marketing strategy was developed (Fig. 1).
In (Block 1), the HEI determines video marketing strategy for computer science (CS) specialty. Block 2 involves evaluation of the state and potential of the existing HEI CS video marketing, and then strategy objectives are specified (Block 3). In Block 4 planning actions take place. It is necessary to take into account the specificity of HEI advertising, when client’s behavior is especially important. Strategy quasi-experiment takes place in Block 5 verifying a proposed strategy in some practical case. Next, we need to evaluate the expected efficiency (Block 6) because it is important for budget redistribution. If this evaluation doesn’t satisfy then it is necessary to simulate video marketing targeting for the advertising campaign to make changes during the formation of the target group at the planning stage (Block 4).
No
4
Next, we need to evaluate the efficiency (Block 6) of the expected new video marketing strategy for the computer science specialty, to perform automation at this stage, because it takes a long time to process and is important for budget redistribution.
When evaluating the efficiency of the expected strategy (Block 6), it is necessary to simulate video marketing targeting for the advertising campaign of higher education institutions, which will allow changes to be made during the formation of the target group at the planning stage (Block 4).
Data from Facebook advertising campaign results are used as a basis for the experimental research methodology. As mentioned above, HEI targeted advertising should be focused on customer behavior. Therefore, in the first step of the technique, it is advisable to select the following four targeting options for Age, Sex, r, Vaverage (Table 1).
1
Data preparation: as.factor(Age, Sex, r) Installation and run of libraries
library(rpart) library(rpart.plot)
library(rattle) library(RColorBrewer)
library(caTools) Creation of control sample. split <- sample.split(face$r,
SplitRatio = 0.6) 3 4 5
Return of vectors subset Train <- subset(face, split == TRUE) Test <- subset(face, split == FALSE) Building of cp table.
printcp(FaceTree) View graph of error.
plotcp(FaceTree)
with(FaceTree, {lines(cptable[
Error)+SE"), lty = c(1, 1, 2), col = c("red", "black", "black"), bty = "n") })
End Fig. 2. Algorithm for targeting video marketing based on dendrogram using R language F<-Age+Sex+Vaverage
Model building.
FaceTree <- rpart(r ~ F, data = Train,
method = "class", control=rpart.control(minsplit =1, cp =
0.0005)) Graphic representation of decision trees prp(FaceTree, box.palette=c("Greys"))
fancyRpartPlot(FaceTree, palettes=c("Greys"), type=2) 6
In block 1, the data must be prepared and cleared for analysis (file with the extension “*.xlsx”) (block 2), where the model parameters values are located in columns. Parameters are converted into data with factors value of (Age, Sex and r).
Next, the libraries (Block 3) are run to build the model, control samples (Block 4) are created and the vector r data is split into two groups in a predetermined ratio, keeping the relative ratios of different labels in r. In addition, a subset (Block 5) of vectors, matrices, or data frames that meet certain conditions are returned. In block 6, a targeting model is constructed based on the recursive partition and regression tree, taking into account the settings (Block 7) and the graphical representation of the dendrogram (Block 8). Block 9 builds a cp table based on the tree Complexity Parameter, and block 10 maps the model errors, details of which are described below in Section 3.3.
For implementation of the algorithm and the program (see Fig. 2) on the example of the admission campaign for “Computer Science specialty” of Ternopil National Economic University (Fig. 3) more than a thousand of indicators of advertising campaign on Facebook according to the parameters (see a Table 1) were used.
The resulting model of targeting based on dendrogram is presented in Fig. 4 and a graphical representation of the model error is in Fig. 5. As it can be seen in Fig. 4 most videos were viewed by male clients in the age of 18-25, 35-55.
The ordinate axis (see Fig. 5) shows the X-val Relative Error and the abscissa axis parameter of the complexity of the tree cp, which starts from the lower boundary point (inf). Size of tree is determined by the number of branches 1… 9. As it can be seen in Fig. 5, starting from branch 7 of the tree the result is stabilized and the subsequent addition of branches has little effect on the tree learning error.
The calculations showed that the parameter of the tree complexity, with a minimum of relative error (min (SV error) + SE) at cross-checking is determined by the value of cp = 0.0033 between branches 7 and 8 (see Fig. 5), where SV is Cross Validated and SE – Standard Error for cross validation error. To check the modeling results based on dendrogram (see Figs. 4, 5), a repeated simulation of targeted advertising was conducted, where we select the following target group for the ad: male in the age group 18-25, 35-55. As can be seen from the results of the re-simulation (Fig. 6), the conversion cost for video ads decreased from $ 0.99 to $ 0.08 as of July 13, 2019 (with 113 conversions), which is about 24% better than the first ad variant campaigns.
Let’s keep in mind that the average of an ad is click-through rate – CTR increased from 2.78 to 3.4.
Unlike analogues [30, 31], the Targeted advertising model, based on dendrogram, allows the tree branches making changes in the advertising campaign strategy for the attributes, which the target function depends on. Moreover, the use of rpart library in the R programming language makes it possible to quickly clean and filter data, which simplifies the formation of target groups compared to [31]. 4
A generalized algorithmic structure for HEI video marketing strategy forming is proposed, which is focused on Targeted Advertising, to improve the effectiveness of the admission campaign.
A targeting model based on dendrogram and target group allocation has been developed, which allows using tree branches to make changes in advertising campaign strategy based on the attributes dependent on the target function. Moreover, the proposed targeting model makes it possible to quickly clean and filter data, which simplifies targeting.
The results of the experimental research conducted on the example of HEI admission campaign for computing major confirmed the effectiveness of the developed targeting model, which is of practical importance in the formation of the Targeted Advertising budget, as part of the formation of video marketing strategy. 21. Olowolayemo A., Saleh N., Adewale N.T., Shatar F.: Exploring attitude toward online video marketing in Malaysia. In: Saian R., Abbas M. (eds) Proceedings of the Second International Conference on the Future of ASEAN (ICoFA) 2017, volume 2. Springer, Singapore. pp. 341-350 (2018). 22. Costa-Sánchez, C.: Online video marketing strategies. Typology according to business sectors. Communication & Society. 30(1): 17-38 (2017). 23. Wang, A., Gao, X.: Hybrid variable-scale clustering method for social media marketing on user generated instant music video Tehnički vjesnik. 26(3): 771-777 (2019). https://doi.org/10.17559/TV-20190314152108 24. Colicev, A., Kumar, A., & O'Connor, P.: Modeling the relationship between firm and user generated content and the stages of the marketing funnel. International Journal of Research in Marketing. 36(1): 100-116 (2019). 25. Sedej, T.: The role of video marketing in the modern business environment: a view of top management of SMEs. J. for International Business and Entrepreneurship Development. 12(1): 37-48 (2019). Doi:10.1504/jibed.2019.103388 26. Viñao, A., & Ruiz-Funes, M. J. M.: Advertising, marketing and image: Visual representations and school modernity through postcards. Sisyphus: Journal of Education IV, (1): 42-66 (2016). 27. Yeveitseva O. S., Novikov D. V.: Features of development of marketing strategy of higher education institution. Economy and the State. 3: 77-79 (2017). (in Ukrainian) 28. Law of Ukraine “About higher education”. no. 1556-VII (2020). (in Ukrainian) 29. Breiman, L., Friedman, J. H., Olshen, R. A., Stone, C. J.: Classification and regression trees.
Monterey, CA: Wadsworth & Brooks/Cole Advanced Books & Software. (1984). ISBN 978-0-412-04841-8. 30. Lian, S., Cha, T., & Xu, Y.: Enhancing geotargeting with temporal targeting, behavioral targeting and promotion for comprehensive contextual targeting. Decision Support Systems. (117): 28-37 (2019). 31. Wang, H., & Hong, M.: Online ad effectiveness evaluation with a two-stage method using a Gaussian filter and decision tree approach. Electronic Commerce Research and Applications. (35): 100852 (2019). DOI:10.1016/j.elerap.2019.100852. 32. Facebook for Business – https://business.facebook.com/business/help 33. Schwarzl, S., & Grabowska, M.: Online marketing strategies: the future is here. Journal of
International Studies. 8(2): 187-196 (2015). 34. Bomba, A., Kunanets, N., Pasichnyk, V., & Turbal, Y.: Process modeling of message distribution in social networks based on socio-communicative solitons. International Journal of Computing. 17(4): 250-259 (2018). 35. 5 Steps to Develop a Video Marketing Strategy on a Startup Budget – https://digitalmarketinginstitute.com/blog/5-steps-to-developing-a-video-marketingstrategy-on-a-startup-budget 36. https://www.forbes.com/sites/forbesagencycouncil/2017/02/03/video-marketing-thefuture-of-content-marketing/#505b54546b53 37. Quinlan, J. R. Decision trees as probabilistic classifiers. In: Proceedings of the Fourth International Workshop on Machine Learning. pp. 31-37 (1987).