STATISTICAL ANALYSIS OF DATA ON THE WORK QUALITY OF CALL CENTER OPERATORS

STATISTICAL ANALYSIS OF DATA ON THE WORK QUALITY OF CALL CENTER OPERATORS

Meruyert Abdirakhmanova

Master student, Astana IT University,

Kazakhstan, Astana

ABSTRACT

Currently the so-called call centers that receive and process calls (requests) of potential customers of various companies have become widespread. It will not be an exaggeration to say that the call center is the "face of the company" since the quality of its work could be to draw conclusions about how correctly and efficiently the company will serve the client's requests in the future. In this article, we analyzed statistical data on the work quality of operators of call centers. The importance of calculating the main indicators in assessing the quality of the call center is discussed. The quality of the operator's work is considered as the criteria for the optimality of the call center.

Keywords: call center; indicators; operator; call service center.

INTRODUCTION

In conditions of high availability with customers, it becomes a guarantee of the activity of organization. The company tries to minimize temporary misunderstandings between them and satisfaction with customer loyalty, to strengthen and continue further mutually beneficial relationships. Companies strive not only to retain their customers, but also to increase the self-sufficiency ratio. Some call centers have quality control teams that monitor and evaluate the quality of their work. The quality of service provision is the main factor of economic efficiency in the info communication sphere. The expansion of the range of telecommunication services leads to the determination of the number of customers, that the need to improve the work of call centers, to the quality of the characteristics meet for ever higher requirements. The relevance of the article is to define the main indicators in assessing the quality of the call center.

The purpose of the article is to analyze the indicators of the quality of the work of operators. The article defined indicators that affect to quality and methods of theoretical research were used: literature review, observation, market research, statistical analysis.

LITERATURE REVIEW

Nowadays call centers pay attention to the quality of customer service. At the same time, indicators of the quality of service are the probability of refusal in needs, the average waiting time for service in the queue, the average queue duration, and the average service time as mentioned earlier [1]. These indicators characterize the efficiency of the technical means and software of the call center. But the quality of customer service is also influenced by the quality of operators' work.

Companies with their own call centers incur significant costs associated with the purchase of licenses and special equipment, maintenance of premises, organization of workplaces, recruiting, and remuneration of employees. In turn, the number of operators might amount to hundreds, and often thousands of staff units. In this regard, companies seek, to reduce the number of employees of the center if it is possible. At the same time, there is a risk of a decrease in the quality of the work of the call center, which could lead to a decrease in customer loyalty and their outflow, and hence a reduction in income. Under these conditions, the accuracy of planning the number of call center personnel is one of particular importance.

It is important to note that call service centers operate under conditions of a priori uncertainty in the behavior of the external environment, which is reflected, among other things, in the uncertainty of the intensities of applications in call centers with a limited number of operators as shown by Zarubin [2], as previously stated.

Some call centers have quality control teams that monitor the work of operators and evaluate the quality of their work. In addition, there are automated systems for monitoring the work of operators that record the conversation between the operator and the subscriber.

The quality indicators of the call center, based on the materials of the article [3, 4], could be divided into: indicators of the quality of customer service by the operator; indicators of the quality of service of the call center; degree of customer satisfaction. Participants of meeting processing metrics with subscribers and are associated with several characteristics: customer focus, credibility, performance, and efficiency.

There are such areas of activities where the use of call centers would be an effective solution. In the service sector modern organizations are working strive for round-the-clock and operational interaction with real and potential customers. Belskaya [3] states that features and organizational aspects of the activity of the call center operator provide a basis for building a group of specialists into the high-risk category in terms of monitoring the development of professional stress.

There are several works, as shown by Malov [5], that describe a method for monitoring the operation of the call service center (CSC) based on the use of mathematical modeling. According to [6, 7], the performance indicators are determined by the organization of the work of the call center and the patterns of the incoming flow of calls.

The purpose of this article is to analyze the statistical data of the call center, which are quantitative indicators of the quality of the operator's work. As mentioned in [8] the main assessments of the quality of functioning of the call centers are the following indicators:

-the average waiting time for a response by served subscribers, that is, how long, on average, the subscriber waits on the operator's answer line, listens to music and the announcer's voice;

-the maximum waiting time for the operator's response (the subscriber waited for the operator's response longer than other subscribers);

-percentage of dialing (percentage of serviced calls from the total number of calls received by the center):

                                                  (1)

where A – serviced calls and B - lost calls.

-percentage of lost calls;

-level of service, % of subscribers who expected an answer from the operator for no more than 30 sec, while 80% of subscribers must receive the answer within 30 sec.

To improve the quality of work of call center operators, monitoring is necessary. As seen in [9] monitoring the quality of the work of operators is conditionally divided into two types: regular (scheduled) control of all operators; control of specific operators, carried out if necessary. The decision to monitor individual operators is made, as a rule, based on statistical data on their performance. For example, the average call processing time of one operator is significantly worse than that of other center operators.

Truth, here it is necessary to differentiate by operators, for example, the average call processing time for a beginner should be compared with the average time of other operators hired in one time, and not with the average time of experienced operators. When scheduling monitoring sessions, the specifics of the work of each specific call center are considered. Factors affecting the schedule include hours and days of the greatest load, time of change of operators, qualification of operators (for more details, see [3]).

The classic example of a call center is a structural one, designed to receive and process telephone calls from subscribers. However, result of the widespread use of Internet technologies in the entire sphere of human activity, new possibilities for using their services, such as e-mail, web chat etc., have appeared.

SOLUTION METHODS

The article provides exploratory data analysis using data visualization tool. The method of exploratory data analysis includes the data that contains the information that was analyzed to answer the research question.

The data from “city of Cincinnati” website was chosen as the object of the study, in which dataset contains the record of all Citizen Service Request (CSR) calls received at the call center at the Department of Public Services. Data is generated by incoming calls received and refresh frequency is daily, renewal of data we can find in [10]. The downloaded file contains data from 2014 to 2022 about 786430 rows. The initial view of the data is shown in Figure 1.

Figure 1. Dataset

Before stage of data cleaning, our dataset has main initial indicators on the following values: AGENTDISPCOUNT – count of operators, AGENTDISPID - when an agent has completed their call, they are required to choose the disposition of the call and how it was completed, CALLSTARTDT - date the incoming call occurred, CALLACTIONID - a code for internal actions that occurred as the call completed, CALLID - unique to a leg of the call, TALK_TIME_SECS - the total amount of time in seconds the call lasted in seconds, ABANDONED - an abandoned call is a call that is ended before any conversation has occurred, this field is a 1/0 attribute on if an incoming call to the DPS call center was abandoned, ANSWERED - this attribute shows whether the call was answered, if the call = "0" that means there was a technical error and the call was dropped before a representative could answer the phone.

Since the data type is a time series, we can analyze weekly, monthly, seasonally, and yearly. Before estimating the average service time and making an analysis of the factors influencing this value, it was necessary to collect statistical data on the following values: date the incoming call occurred, number of incoming calls; number of unserved/served calls; average call answer speed in sec; number of operators (Figure 2).

Figure 2. Pivot table

Based on selected data we created dashboard and presented it in Figure 3 below.

Figure 3. Call center data dashboard

In this article we analyzed data received in December 2020 and December 2021. Before estimating the average service time and making an analysis of the factors influencing this value, it was necessary to collect statistical data on the following values: number of incoming calls; number of served calls; number of lost calls; average call answer speed in sec; amount of operators. Based on selected data we created dashboard and presented it. Figure 4 shows incoming traffic for December 2020 and December 2021.

Figure 4. Incoming traffic for December 2020 and December 2021

The data about answered calls and the number of missed calls for each year demonstrated in Figure 5 and Figure 6 respectively. As could be seen from the graphs, the largest number of calls comes on Monday. Then the load decreases somewhat and keeps approximately at the same level and falls at the end of the week. The least load is for Sunday. A sharp decline in calls is coming closer to the New Year holiday and in 2021 only 1 call was received.

Figure 5. Answered calls and the number of missed calls for December 2020

Figure 6. Answered calls and the number of missed calls for December 2021

The average indicators for call center are below in Table 1: the number of incoming calls per month, answered calls, abandoned calls, average waiting time, average talk time, the number of operators, % of received calls.

Table 1.

Statistical data of call center

DISCUSSIONS

Comparing the data from Table 1, it might be seen that there are results for the universal operation of the call center, which is the result of an analysis of the number of calls served and lost calls. In December 2020, call center receives 192 times lower calls, compared to result of December 2021, but the percentage of serviced calls seems similar (»80%). Lost calls in both years are equal. The average waiting time in the queue is 32.9 times longer in December 2020. The average talk time is 7.8 sec longer in December 2021. According to dashboard, the number of calls for 2021 shows a downward trend from Monday, December 27, 2021 to Friday, December 31, 2021, the total drop was 416. The number of calls for 2020 shows a downward trend from Monday, December 14, 2020 to Wednesday, December 23, 2020, the total drop was 38. The count of ‘Not answered’ calls was unexpectedly high on Wednesday, December 2, 2020. It was 326, which is outside the expected range of 235-306. The count for ‘Not answered’ had a downward trend between Thursday, December 17, 2020 and Thursday, December 24, 2020 with a fall of 63, but had a significant change in trend and was up by 36 since Sunday, December 27, 2020. Based on the obtained statistical data, it might be seen that the existing call processing system in call centers requires the development of a new distribution algorithm of calls, considering the priorities of served calls since the number of lost calls has not decreased compared to 2020.

The results obtained might later be used in solving the problem using machine learning algorithms to minimize resource costs while improving the efficiency of operators considering the correct distribution of the number of operators.

CONCLUSION

Collection and analysis of statistical data of call center were carried out. The main indicators in assessing the quality of the call center is defined, and the controlled parameter of the systems was defined the number of operators in the call service center. As criteria for optimality, the call service center should bring the maximum possible profit with a minimum number of operators considering the correct distribution of the number of operators. The obtained results might later be used in solving the problem of optimization of call center mode along the changing flow of customers using machine learning algorithms to minimize resource costs while improving the efficiency of operators.

References: 

1. Goldstein, B. S., & Freinkman, V. A. (2006). Call-centers and Computer telephony. St. Petersburg, Russia: BHV.
2. Zarubin, A. (2003). Call and contact centers: the evolution of technological and mathematical models. Vestnik svyazi, (8), 85-88.
3. Belskaya N.M., (2012). Research and development of algorithms for monitoring and analyzing the quality of work of call center operators. Abstract of the thesis for the degree of candidate of technical sciences-Samara.
4. Rosljakov, A. V., Samsonov, M. J., & Shibaeva, I. V. (2002). Center of service of calls (Call Centre). M.: Eko-Trends, 272.
5. Malov A.V. (2010). Methods and means of ensuring fault tolerance and call centers based on IP telephony.
6. Gans, N., Koole, G., & Mandelbaum, A. (2003). Telephone call centers: Tutorial, review, and research prospects. Manufacturing & Service Operations Management, 5(2), 79-141.
7. Bernett, H. G., Fischer, M. J., & Masi, D. M. B. (2002). Blended call center performance analysis. IT professional, 4(2), 33-38.
8. Ding, S., & Koole, G. (2022). Optimal call center forecasting and staffing. Probability in the Engineering and Informational Sciences, 36(2), 254-263.
9. Goldstein B.S., Isaev V.I., Mamontova N.P., Frankman V.A. (2006). Analysis, synthesis and quality management of service centers infrastructure. Educational guide.
10. Tyler Data & Insights. (2018, November 29). Citizen Service Request (CSR) Call Center Calls. Retrieved from https://data.cincinnati-oh.gov/Efficient-Service-Delivery/Citizen-Service-Request-CSR-Call-Center-Calls/k2qr-ck2v