Project for Business Analytics Class

Section 1: Introduction

This paper deliberates data regarding a bank's marketing campaign. The dataset obtained for this study belongs to a continuum of domains including finance, sales, and marketing. It will help banks develop better marketing strategies that will generate high revenue. The underlying rationale for the data is to increase the effectiveness of the marketing campaigns. Bank customers have varying needs and thus require different services from banks. Misguided marketing can cause a bank to spend a significant amount of resources in terms of money and time and still not attain the intended objectives of increased revenue. Hence, the marketing needs to be targeted to potential clients to be effective.Analysis of this data will generate insight that will guide future marketing campaigns. First, it will help create clusters of clients with similar characteristics, such as job description and education level. Banks can target these clusters using different techniques and amenities. By doing so, banks will not waste valuable resources targeting a group with irrelevant plans and services. For instance, blue-collar individuals save in a different format than white-collar individuals since they receive their money regularly. They do not have to wait for the end month to receive their salaries. Thus, their savings culture and needs are different. Banks need to tailor their marketing for this group to augur well with their earnings. Besides, the data will provide meaningful information on the relationship between marital status, loans, and bank balance.Besides, the data will generate strategic value for banks as they will manage their resources effectively to attain the optimum output. This value will enable banks to gain a competitive edge in the industry. The dataset used is obtained from Kaggle. It has been downloaded over 30,000 times and viewed over 250,000 times.

 Section 2: Information Management

The initial data set comprised more than 11,000 records which was too much. As a result, some transformation was made to the data to attain a reasonable number of entries.Data filtering was performed on pertinent columns including rating, reviews, and installs. Conditional formatting was applied on all the columns based on a predetermined measure that augured well with the aim of the research. Extraction through advanced filtering was conducted on several columns that had categorical data. The overarching lucidity for this action is that Excel does not allow a descriptive analysis of non-numeric data. Missing values were identified and populated with data before the analysis tests were conducted(Chatfield & Haipeng Xing, 2019). The filtering was intended to bring the records to a reasonable number above the required threshold of 400 records. The final pivot table comprised 564 records and was attained by filtering the balance category.

Section 3: Descriptive Data Analytics

A frequency distribution was run on Excel. This test returns a tabular or graphical output that displays the precise number of observations on the categories along a measurement scale. This descriptive technique is important as it allows the researcher to glance at the entire data conveniently. Besides, it displays whether the observations are high or low and whether they are concerted or spread across the measurement scale(Hogg et al., 2019). The frequency distribution was run on relevant categories in the data set. Other descriptive statistics such as mean, median, mode, standard deviation, and variance were run on numerical categories.

Frequency Distribution

 Table 1: Job

Most of the customers targeted in the latest marketing campaign by the bank were employees in management positions. This was a deliberate effort because they have sufficient disposable income and are paid via their accounts at the bank. Technicians were the second biggest group targeted in the campaign. The next in line were employees in blue-collar and admin positions. The latter receive payments regularly and do not have to wait till the end of the month to receive their salaries(Johnson & Wichern, 2019). Hence, their financial needs are differentiated from salaried employees receiving monthly payments.

 Marital Status

Figure 1: Marital Status

Most of the clients were married, implying that they had additional bills to support their families. Hence, they are more likely to take loans, which are banks' primary income sources. Single individuals were the second, followed by divorced persons.

 Table 2: Education

Most of the clients had attained tertiary education, which implies that they were knowledgeable of the terms of the loans marketed to them by the bank. This population could make an informed choice regarding the products marketed to them. These clients had better jobs, implying better salaries that could translate to higher loans. Their savings were also considerably higher compared to individuals that only attained secondary education.

Table 3: Loans

Most of the clients did not have an active loan. This implies that they could take new loans against their assets or salaries. Current loans often disqualify clients from taking additional loans until they are cleared. However, many clients with loans are likely to take more loans before or after repaying them. Hence, they form a good basis for a marketing campaign. Clients that are yet to take loans can be enticed by various strategies that will be discussed in the predictive analytics section.

Figure 2: Outcome from previous campaigns

The outcome of the previous campaigns was largely unknown. This implies that the bank did not set smart metrics to indicate whether the marketing campaign was successful. In the known cases, failure and success were approximately the same, meaning the marketing campaign was ineffective(Hays, 2018). Advancing technology has now made it possible to track the effectiveness of marketing campaigns. Banks can leverage this aspect to determine if they need to restructure their campaigns to attain augmented outcomes.

Table 4: Contacts

The majority of the clients could be reached via cellular. This population could thus be reached through email and internet marketing via their smartphones. Clients reachable via telephone could be reached through telephone marketing, whereas a combination of different marketing techniques could reach the unknown. Banks must use the preferred communication channels for the clients as this will increase the likelihood of engagement. Even though the clients may have several communication platforms, banks can identify the most frequently used. Subsequently, they can use this channel to carry out the campaigns.

 Table 5: Month

The above table displays the number of clients who responded to the marketing campaign each month. Most of them responded in May, November, and August. May is just after the Easter celebrations, and August is before the start of a new semester. During both periods, the clients needed money to pay various bills. Particularly, they may need this money to pay for bills such as school fees or vacations.

Descriptive Statistics 

Table 6: Age

The mean age of the clients was 44.4, and the median was 42 years. This age represents a group of individuals that have worked for over fifteen years. They have likely made investments or are about to make investments. Thus, they need bank loans to fund these investments. Besides, they have acquired significant savings that they can combine with their loans to fund their desired projects.

Table 7: Balance

The average balance of the clients was $11,516.8, whereas the mode and median were 8304 and 8851.5, respectively. The range of the balance was large, which signifies that the clients have varying financial abilities and thus require customized services(Johnson & Wichern, 2019). Correspondingly, the amount of loans they can take is different. Customers with higher balances can take bigger loans due to the transactions that offer them a higher threshold.

Table 8: Campaign

The average time for the previous campaign was 2.5 days, while the median and mode were 2 and 1, respectively. Likewise, the range was large, indicating that the marketing campaigns varied significantly. Some were simple to design and implement and took fewer days, whereas others were intricate, taking more days. For instance, telephone marketing can take place in a day because it is simple to call clients that meet a predetermined criterion. On the other hand, internet marketing may take longer since the clients have to be divided into clusters and different avenues used. Campaigns that last for extended periods are likely to reach a bigger audience.

 Section 4: Predictive Business Analytics

 The forecasting of the future value of balance is a pertinent aspect of this analytical exercise. The purpose is to utilize the existent historical data in the prediction of future value for appropriate planning. Forecasting provides an estimate to any organization to be able to develop a reliable budget and strategy to mitigate the expectations given the information obtained from history (Johnson & Wichern, 2019). Therefore, this section will circumspectly outline the forecasting of balance using two methods namely exponential smoothing and moving average.

Moving Average

The simple moving average avails the forecast value based on the formula below: -

 Whereby t is the current period, is the forecast for the next period, and n is the forecasting window also known as in n -point moving average. Therefore, if n is equal to 2, the moving average aggregates two observations back and finds the average within that window (Roderick & Rubin, 2019). A is the actual figures given for each period.This report employs a three-point moving average to obtain the forecast of the Sales figures. The equation is as above with n being equal to 3 for every window throughout the dataset.

The plot of the moving average is outputted below.

Exponential Smoothing

Exponential smoothing is also a forecasting technique that is slightly different from the moving average function. The equation of the exponential smoothing looks like the one below: -

 The moving average expresses equal bias on each observation on a particular window. The exponential smoothing function contains a smoothing constant which expresses how much the forecast will react to the observed differences(Chatfield & Haipeng Xing, 2019). The higher the value of the smoothing constant the higher the reaction to differences. Choosing a smoothing constant of 0.2 yields the forecast below.

The plot is outputted below: -

Regular exponential smoothing will always lag behind the trend and sometimes it ought to be incorporated into the forecasting using the technique. Therefore, the forecast for July and August based on the data is given by the model, and the values are shown above. This state, therefore, indicates that the prediction is based on, historical data

Conclusion

The paper therein has entailed the analysis of the marketing campaign of a bank. The variables associated therein inclusive of the outcome of the campaign, the bank balance of the customers as well as the time have been instrumental. Moreover, the marital status, and the time required for previous campaigns. The average age of clients is also pertinent for the understanding of the impact of the marketing campaigns. The descriptive statistics associated with the variables have been comprehensively exposed to facilitate the potent understanding of the contents of the dataset about marketing in banks.

The predictive analytics furthermore has built on the descriptive knowledge assembled from the descriptive analytics to understand any projections. The variable under analysis was the client’s bank balance regarding the marketing campaign’s impact. Two-time series methods have been employed for forecasting. The moving average model considered the average of the bank balance over the three-year window and the projection stood at 6293 with a standard deviation of 2557.92. the range of this projection is also subject to the addition or deduction of the standard deviation from the projection. The exponential smoothing forecast stands at 6481.34 with a standard error of 5990.94. this latter forecast is a wide range of values as the minimum can go to below a thousand dollars in bank balance while the maximum can go to more than eleven thousand dollars. These forecasts indicate the wide range of values for possible bank balance for any customer. Therefore, the descriptive and predictive analytics of this dataset has exposed a variety of information about the balance of customer deposits in the bank about the marketing campaign.

References

Chatfield, C., & Haipeng Xing. (2019). The analysis of time series : an introduction with R. Crc Press,Taylor & Francis Group.

Hays, W. L. (2018). Statistics. Wadsworth/Thomson Learning.

Hogg, R. V., Tanis, E. A., & Zimmerman, D. L. (2019). Probability and statistical inference. Pearson.

Johnson, R. A., & Wichern, D. W. (2019). Applied multivariate statistical analysis. Pearson.

Roderick, & Rubin, D. B. (2019). Statistical Analysis with Missing Data. John Wiley & Sons.