Influence of Car Configurator Webpage Data from Automotive Manufacturers on Car Sales by Means of Correlation and Forecasting

Round 1

Reviewer 1 Report

I insist on adding information to the content of the publication:

1. why a BTS strategy is used instead of a BTO strategy - even more so in a post-pandemic reality and economic crisis

2. there is no relevant information from the point of view of the research problem: how are visits to a customer's website and the choice of a car colour correlated with the history of sales of this colour to the same person? If such a correlation is not included in the research we can talk about a large margin of error and supply creating demand and not the other way round

3. I suggest expanding the description of the similarity of the real estate market to the study in question in terms of the validity of the choice of research tools

4. Twitter - how does the age of Twitter users relate to the age of people making purchases - where is the denominator of the study found here?

5. how were search motives, which are not being researched, linked to observed search data, which is being researched?

6.Once again, I emphasize the necessity to describe the correlation of data on people who have gone through the entire path of configuration on the website and to verify whether the exactly configured model was ordered by the same person

7. In addition, I would ask you to describe the order in which the surveys are filtered: is the mechanically configured car first? Is the first criterion the choice of colour?

 

Once this information has been added and clarified in the body of the article, I recommend it for publication.

 

I wish you much scientific success and further publications. 

Author Response

Dear reviewer,

Thanks for your suggestions. Next, you can find my comments to clarify your doubts. 

1. BTS strategy are consequences of the suppliers' lead times. It would be even more difficult to change to BTO with the actual crisis of limited access to semiconductors.
2. Subsections Correlation between sales and CC data from Methodology and Results explain the discoveries done in this area. It is not possible to matching per person online car configurator and sales. 
3. We do not have access to Twitter data. Moreover, car configurator does not require user to log in, so no data about sex, age, etc. is available.
4. Order of car configurator is as follows: car model > car trim > engine > color and wheels > upholstery > additional packages. 

Reviewer 2 Report

This paper presents a case study of using internet and sales data to forecast color mix sales of cars. The SEAT automotive manufacturer is selected as an example. Based on some historical sales data, the authors use several existing models to forecast the color mix sales and finally present some discussions on improving future sales.

Although the topic of this study may be useful, this paper has the following limitations. Thus, it is not suitable to be accepted as a journal paper in its current form.

L1. This paper is simply the application of existing forecast models. There is no novel idea or algorithm proposed by the authors. To my understanding, such a study might be more suitable as a lab project in a data-mining course.

L2. The research results are not meaningful to other manufacturers. If the authors think that the results are of general meaning, it is necessary to show your evidence.

L3. The features considered in this study are not convincing. Why did you only concern about the webpage visits of users? In order to reflect users' interests in color mix, more features, e.g., price, news, and reviews, could be used to improve the accuracy of prediction. It is necessary to explain why you do not include other features because we all know feature is a critical factor impacting the performance of a prediction model.

L4. The writing of the paper needs improving. In the introduction part, it is better to state the unique contributions and novelty of this study clearly. The title is suggested to be "... of automotive manufacturers: A case study of SEAT". 

Author Response

Dear reviewer, 

Thanks for your suggestions. I list the changes I introduced in my paper with the intention of clarifying your doubts. 

L1 and L2. Leitmotiv of the paper has been well explained and can be extended to other automotive OEMs. The note shows the new procedure to prove the influence of car configurator data in demand prediction by means of correlation and forecasting. 

L3. Explanation is provided in 2.2 Research gap

L4. Introduction states clearly contributions and novelty of the study. New title is proposed. 

 

Reviewer 3 Report

see separate file

Comments for author File: Comments.pdf

Author Response

Dear reviewer, 

Thanks for your suggestions. New paper version corrects English typos you pointed out. 

Reviewer 4 Report

The paper examines whether the inclusion of car configurator webpage information to weekly color mix sales improves the predictive power of the research object. The paper employs various forecasting methods to achieve the research objectives, and provides a comparison of forecast accuracies to underpin agruments for the best technique/model.

In my opinion, research target setting, literature review, data collection, and method selection altogether meet the requirements of the journal, however, I have the following recommendations to improve the quality of the paper:

-  Supplement the Introduction section by positioning the paper in academic literature, and clarify its added value to literature. In addition, summarize the key empirical results of the paper, before introducing the structure of the rest of paper.

- Provide a more detailed analysis of the collected database. How many observations have been examined in what breakdown...

- Improve the Methodology section by providing more details of the applied methods (such as essence of methods, statistical assumptions, advantages and drawbacks). Nothing has been written about the XGboost method, which arrived at the best predictive power. Elaborate the features of ensemble learning and its application to the current research.

- ARIMA and VARMAX are not machine learning methods, but traditional parametric statistical procedures.

- Although correlations and forecast performance indicators seem to be spectacular in the Results section, however, in themselves are not enough to evaluate the results of the performed empirical research. Nothing is disclosed about the model development procedure, inter alia modeling steps, parameter and hyperparameter settings, in particular when applying the XGboost method. Such details of empirical model development cannot be hidden in secret at this journal level. In its current form, results are not transparent, and give the feeling to the readers that authors simply imported the database into a modeling software, do not know what happened in between, and intend to present correlation and accuracy statistics as being the results of empirical research, which are in fact not results.

-  A classic Discussion section is missing from the paper. Compare your results to that of other papers, even dealing with similar problems but in different context.

- Conclusions section is rather limited and needs to be extended to reflect the findings of the paper. Since this is a very applied subject, managerial insight needs to be clearly staded making the technique(s) impactful as well.

Author Response

Dear reviewer,

Thanks for your suggestions. I list the changes I introduced in my paper with the intention of clarifying your doubts.

1. Introduction part now positions the paper in academic literature, and clarifies its added value to literature.
2. Section Dataset description is supplement with plots referring sales and car configuratior visits distribution and time series.
3. Methodoloy includes, in subsection Forecasting Techniques, description about the different forecasting techniques (ML algorithms and statistical procedures) together with more detailed explanation about the forecasting process for each car variant. 
4. Results section is now complemented with Methodology section and figures and explanations in every step of the procedure are provided.
5. Discussion added references to papers to a similar context already mentioned in Related Works section. 
6. Conclusions section attemps to focus on the managerial insights and prove the novelty and originality of the research.

Round 2

Reviewer 2 Report

The authors have made significant revisions to the article, which has been well improved. Now all my concerns have been well explained, and I would suggest acceptance of this article after grammar edits.

Reviewer 4 Report

Authors have made efforts to increase the quality of the manuscript. Since all my recommendations have been covered, and the paper have been adequately revised, I recommend the acceptance of the paper.