Analysis Of Performance In Contact Centre Using Qta

Analysis of Performance in Contact Centre Using QTA

Introduction

A centralized customer-service office that provides tele-services, where they answer incoming calls from customers via telephone contact for business organizations is called a contact centre or customer contact centre. Diverse range of businesses like large, medium and small organizations are establishing contact centre. Departmental stores, mail order catalogue firms, utility companies, insurance companies, airlines, banks, emergency road service operators and many others are linked with contact centre to get connected with their customers through telephones.

Customer enquiries regarding phone service billings, and the reporting of faulty services, ordering new features and benefits, changing customer profile and cancellation of services etc., in telecommunication service providers are done through contact centre. Staffing involves the major cost for a contact centre. The percentage of cost distributed in the contact centre is at the rate of 65 percent for staffing, 25 percent for networking and communication and remaining 10 percent for maintenance. Ongoing training programmes are often conducted by the contact centre due to more resignation of jobs which in turn increases the cost for the contact centre. This drastic resignation of jobs among employees is because of the poor contact centre management which increases the stress level of the employees. The quantitative model which is generally analytical helps to tackle the cost resulting from staffing levels. The standard queuing model is used to examine the quantitative aspect of the performance of a contact centre.

, The case of one of the Malaysian telecommunication service providers is taken in this study and the performance is evaluated and staffing level suggestions are given. In the past, the previous week’s call volume and pattern were used to measure the performance of the contact centre. But in this study, performance of the contact centre and staff planning is evaluated using scientific approach. Performance of the contact centre is investigated in the context of the Erlang C model (classical M/M/s model) and Erlang A model (extended M/M/s + M model) and the results are compared. Service quality is another aspect of performance, service quality goal of the contact centre is to answer 80 percent of the incoming calls without making the customers wait for more than 20 seconds) i.e. known as the 80/20 rule.

Telecommunication Service Provider’s Contact Centre

The contact centre operates 7 days a week, 24 hours a day and handles inbound and outbound calls, faxes, e-services and mail. Since two-third of the calls handled is incoming calls, this study is limited to incoming calls from the post paid customers through telephone without considering Interactive Voice Response (IVR). 

The three stages of the incoming calls are,

  1. Interactive Voice Response (IVR)
  2. Queue stage and
  3. Service stage.

Some calls directly move to service stage skipping the queue stage. The customer call is first connected to the Interactive Voice Response (also called Voice Response Unit or VRU) which enables the customer to complete the self-service transactions. Self-service transaction means asking the customer to press five in the telephone keypad for balance enquiry. The customer may end the call if he/she doesn’t want to speak to an agent or a Customer Service Representative (CSR). To talk with the agent the customer may be asked to press a specified number and the customer gets connected to the agent immediately or joins the tele queue till the agent is available. The call is recorded as received when it exits IVR and joins the tele queue. Some impatient callers leave the queue abruptly and they are categorized as abandonment calls. The calls that reach the agent and served are reported as answered. The tele queues are normally served on the first come first served (FCFS) basis. The contact centre by-pass few calls directly to the next available agent. They are categorized as priority customers. For example, the customers who spend more than RM 500 per month are priority customers.

Performance Models Results

a)      Erlang C or M/M/s

In this study, Erlang C or M/M/s queuing model is used to analyze the performance of the contact centre. Erlang C or M/M/s model assumes Poisson arrival and exponential service time. The contact centre has a total of 60 customer service agents called servers and receives a total of 954 calls in an hour at the rate of 15.9 calls per minute and a service rate of 16.2 calls per hours. Since the queue capacity and the population are infinite, the customers are served on first come first served (FCFS) basis. To prevent the queue from

growing indefinitely, the utilization factor should be below 1. The contact centre with 60 servers has 0.981 utilization factor (?). The fraction of time each server is busy in a queuing system is calculated by the utilization factor.

Erlang C model measures the performance indicator of the contact centre using Queuing Toolpak 4.0. The effect of the different number of servers on the performance indicators is analyzed first since the main objective is to equip the contact centre with the right number of servers.

When the number of servers is 59 the contact centre will be operating at full capacity because the utilization factor is found to be 1. Performance indicators were not available when the contact centre uses servers below 59, stating that the system is not stable. The relationship between the utilization factor and the number of servers is negative linear in nature. Increase in the number of servers decreases the utilization factor. This relaxes the contact centre environment by reducing the workload.

The relationship between the utilization factor and average time in queue (queue in seconds) are examined next. The waiting time for a call remains close to 0 seconds until the utilization of capacity reaches 80 percent; beyond 80 percent, the waiting time in queue increases rapidly. This can be tackled in three ways,

1.      Number of servers can be increased.

2.      Time spent in each can be reduced or

3.      Number of calls arriving at the contact centre can be controlled.

Negative relationship is found between the number of servers and the probability of an arriving call that has to wait to be served. Probability is 0 of having to wait to be served when the number of servers in the contact centre is 80. In other words, no calls wait in queue when the number of servers in the contact centre is 81. Different number of servers leaves different impact on the service provided.  In the beginning, performance indicator increases when the number of server’s increases, but after reaching an optimal point, the impact of the additional servers reduces tremendously. The service time, which is the amount of time spent on each call and the server utilization rate influences the service rate of the contact centre. Calculating the number of servers required for various service times is very important because service time varies according to the complexity of the calls received. This can be calculated using Erlang C or M/M/s queuing model for service time ranging from 120 seconds to 360 seconds. When the agent utilization rate is low, more servers are required for the same amount of service time. For example, while 136 servers and 70 percent agent utilization are required for answering the call which takes 6 minutes, 106 servers is required for 90 percent agent utilization; the difference between 70 and 90 percent is 30 servers. This indicates the weakness of the contact centre; the contact centre operates with 60 servers by utilizing the full capacity leaving a lot of room for improvement in the level of service provided to customers.

b) M/M/s + M or Erlang A model

Here, M/M/s + M or Erlang A model is used to evaluate the performance indicator of the contact centre. Abandonment factor is included in this model. Average wait in queue in seconds is calculated by the percentage of abandonment calls. This helps to derive the average caller’s patience (ACP). Average caller’s patience is represented by the symbol

(?-1). On Monday, 7th February 2005, 21 percent of total calls received are abandoned during 11 am to 12 noon and average wait in queue at the contact centre is 32.4 seconds. Based on this, ACP is assumed to be exponential distribution and ACP at the contact centre is 154.28 seconds or 2 minutes and 34 seconds. Thus, in addition to Poisson arrival and exponential service time M/M/s + M or Erlang A model assumes exponentially distributed patience time.

For example, the average caller’s patience is 2 minutes and 34 seconds when the number of servers is 60, the arrival rate is 954 calls per hour and the average service time equals 3 minutes and 42 seconds. Abandonment feature is not available in queuing toolpak 4.0, so, the 4callcenters version 2.23 is used to measure the performance indicator of the contact centre. Linear relationship is found between the percentage of abandoned calls and average time in the queue in seconds for calls arriving at the contact centre. Percentage of abandonment calls increases when the average waiting time increases.  

Average time in the queue is 123 seconds and 80 percent of the calls will be abandoned when the server equals 12. From the report of 4callcenters it was gathered that performance indicators were not made available and that the contact centre was tremendously overloaded, when the servers were reduced below 12. Violation of the assumption that the utilization rate is below 1 is the important factor to note in this model.

Negative exponential relationship is found between the number of servers and the number of abandoned calls. The relationship between the following parameters of percentage of abandoned calls, percentage of answered calls, percentage of utilization and varying number of servers is tested. Percentage of abandoned calls decreases until it reaches 0 when the number of servers is increase and after 71 it remains at 0. In response to the increase in the number of servers, the percentage of calls answered increases steadily. The percentage of calls answered remains at 100 percent at 71 servers and onwards. According to this, all the calls are answered if the contact centre starts using 71 servers.

M/M/s (Erlang C) and M/M/s + M model (Erlang A) Comparisons

Shorter waiting time and shorter queue is produced by Erlang A and lower utilization rate is enjoyed by the agents. The main objective of the contact centre is to answer 80 percent of the incoming calls within 20 seconds. It is found in Erlang C that 64 servers were required to answer the incoming calls within 20 seconds on average. In Erlang C, only with 65 servers, we can achieve the goal of contact centre i.e. answering 80 percent of the incoming calls within 20 seconds.   Erlang A requires only 53 servers to reduce the time in queue to 20 seconds and below. To achieve 80/20 Erlang A requires 60 servers which is less than Erlang C. When comparing to achieve both the objective of 80/20 and 80 percent agent utilization rate at least 73 servers are used in Erlang A model when compared to 74 in the Erlang C model.

Erlang C requires more server than Erlang A to achieve the goal of the contact centre i.e. answering all the incoming calls within 20 seconds. Over staffing is required in the model which ignores abandonment factor and under staffing is required in the models which include the abandonment factor. Table 4 confirms the above statement. Agent utilization plays an important role in influencing the number of servers required, even while taking  the abandonment factor into consideration and while comparing, there is only a very small difference in number of servers required.

Erlang C does not take the abandonment rate into consideration, so abandonment rate is unique to Erlang A. Erlang A seems to be stable at all times(except when below 12 servers are used) but Erlang C becomes unstable when the utilization factor hits 100 percent and more. The agent’s occupancy rate decreases for both the models when the number of servers increases, to achieve certain level of agent occupancy; number of servers required differs for both the models. For example Erlang A needs 65 servers and Erlang C requires 67 servers to achieve 90 of agent occupancy. Adding 2 or 3 servers in Erlang C model would result in Erlang A, since in operating contact centers personnel cost are the major operational costs.

Conclusion

In this study, we built a queuing model to measure the performance of one of Malaysia’s leading telecommunication service provider to assist its staffing levels, to achieve progress in service quality while keeping costs down. The data used to build a performance model in the context of two models i.e. the Erlang C and the Erlang A are the data collected in the contact centre between 11 am to 12 pm on Monday, 7th February 2005. Erlang C model shows the result that with the current number of 60 servers is not performing the standard required. Only 24 percent calls are answered within 20 seconds but the goal is to achieve 80 percent of the calls within 20 seconds. The average number of calls in the queue was 44 and 2 minutes and 44 seconds are spent by each calls waiting in the queue. To improve the service to the required standard of 80/20, 65 servers are needed in Erlang C model in which abandonment factor is also not included, this leads to under or over staffing resulting in economical consequences.

Thus, Erlang A model including the abandonment factor is used to measure the performance of the contact centre. Average caller’s patience (? ?1) is the addition parameter in Erlang A model. Compared to Erlang C, Erlang A performs better with 60 servers. The average time in queue (7.27) is also quicker, the queue is shorter (1.9 calls), and the agent utilization rate is lower in Erlang A model. Arrival rate, service rate and number of servers are widely explained and the inter-relationship between them is analyzed in this study. Overstaffing of 5 servers to achieve the goal of the contact centre is indicated by the Erlang C model. The conclusion is limited to peak hour of Monday since the data collected is in the peak hour between 11am to 12pm on a Monday. In future the contact centre may go for the changes mentioned below.

  1. The mean service rate which is currently at 16.2 customers per hour can be increased by creative design change with the application of technology.
  2. The callers to have a clear idea of the purpose of the call and to enable the servers more questions can be added in IVR, this enables the servers to focus on the problem solving rather than spending time on understanding the problem.
  3. Number of arrivals can be reduced.
  4. To encourage the callers to call after office hours, off-peak hour promotions activities can be taken by the contact centre, but this should not increase the cost and it should improve customer satisfaction.
  5. To thus reduce the burden on the customer service representatives, more customers can be encouraged to use the IVR.

Last, but not the least, employing a team of analyst to mange the ample data available in the contact centre to reap the benefit from this data is very important