Hot Topics
- Home
- /
- Contact Center
- /
- Insight through Analytics in Contact Center
Figure 1 – Identifying agents who need training is an operational activity performed in every contact center. This figure attempts to show a few of the parameters that reflect agent’s performance. However, the plots and their underlying report data are inadequate to analyze and display the many to many dependencies of these parameters in a manner that guides decision-making. It is almost impossible for Contact Center managers to decide on a training plan based on such two dimensional analysis.
Insight through Analytics in Contact Center
Introduction
It is essential for managers to have the ability to analyze and obtain insights from the huge volume of data generated by contact center transactions. While many managers are intimidated by analytics because of its use of advanced mathematics and statistics, technology and tools available today make it feasible for most organizations to reap value from its use.
Challenges
Performance management is a challenge for contact centers even after implementing reports on metrics.
While executives use reports to measure performance and set benchmarks, they lack ability to scientifically analyze the underlying data for cause and effect relationships. This is because their current ability restricts them to at best a two dimensional analysis. This is inadequate because such basic analysis cannot reveal all the dependencies that define the complex relationships between data elements in a contact center environment. The dependencies of measures in a contact center are many to many and their manual analysis is impossible.
Are your contact center reports effective?
Data quality in Contact Center impacts accuracy and completeness of performance reports. Comprehensive approach to data analysis considering diverse technologies involved will result in cleaner data.
So most Contact Centers lack the ability to reveal patterns and obtain insights in the data.
Here is a scenario a manager faced relating to preparing a training plan for his agents. Training agents is an important activity in a contact center that needs to be performed on a continuous basis. As shown in Figure 1, the manager attempts to analyze various performance parameters like Average Handle Time (AHT), First Call Resolution, % Calls Transferred, % Calls Transferred to Same Queue etc to arrive at an optimal training plan. However it is ambiguous if these are indeed the right parameters to analyze and what is the level of influence they have on agent training. Further these parameters have many to many dependencies, which cannot be envisaged and analyzed in a two-dimensional plot.
Solution
To successfully manage performance of Contact Center, executives need to ferret out meaningful relationships and patterns in the underlying data to identify issues and opportunities. While analytics cannot replace the knowledge and understanding of humans, it can aid in making effective decisions. Analytics uses the combination of statistics, machine learning, neural computing and artificial intelligence techniques to accomplish this purpose.
Derive the right Contact Center metrics
Metrics help spot issues, identify root causes and control factors affecting customers. However, complexity and proliferation of systems in a contact center makes it difficult to derive the right metrics.
Let us look at how the challenge shown in Figure 1, was addressed. An analytical solution was implemented using cluster and discriminant data analysis. Calls Answered, Average Handle Time, First Call Resolution and % of Calls Transferred were selected as parameters (variables) for this analysis. Cluster analysis was applied on these parameters, which grouped the data based on pattern similarity. Data on how classification was done historically for similar performing agents along with the results of cluster analysis is fed to discriminant analysis. First Call Resolution and % of Calls Transferred were found to have greater significance on training requirement than other parameters. These were then factored into 2 factor axes namely F1 and F2. Thus the complexity of a multidimensional data set was simplified using these analytical methods and presented in two-dimensions with the help of factor axis.
With the second-degree function obtained from discriminant analysis, a model was built to identify the training needs of agents based on their performance parameters. Figure 2 shows the visual output of the model. Now in order to classify the type of training required by agents, parameters like % Calls Transferred to Same Queue, % Calls Transferred to Tier 2 and First Call Resolution were selected. Agents were then grouped based on cluster analysis and the past data on how agents who require training were classified into technical and soft skill buckets. Discriminant analysis was applied on the output and a model was built from the resulting second-degree function. The visual output of this model is shown in Figure 3.
Figure 2 - Call data such as calls answered, transferred, average handle time etc. were subjected to cluster and discriminant analysis. The complex relationships were analyzed and plotted in two dimensions for simple visualization. First call resolution and % of calls transferred had significant influence on training needs than others.
So most Contact Centers lack the ability to reveal patterns and obtain insights in the data.
With this approach, the manager was able to identify the parameters and understand the level of significance they had on agent training. This reduced considerable overhead for managers and eliminated the need for any guess work.
A multi-disciplinary team is required to perform the tasks involved in implementing analytic projects. Understanding of Contact Center processes and data that drive them are essential for a successful implementation.
This data has to be integrated, transformed, filtered and segmented before regression and other statistical functions can be performed. Proper design of data infrastructure is critical in performing these operations and to apply statistical algorithms on the data. Data preparation is an important activity, which normally takes up to 25% of the total project time. However availability of an existing data infrastructure will considerably reduce data preparation time.
As the business environment changes in a Contact Center, the accuracy of models degrade. In the example described above on identifying training needs, the assumption is that the products and services are mature and a typical agent is able to resolve most of the calls. Now consider a scenario where a new product gets launched by the organization. As the knowledge of agents about the new product is limited, more calls may need to be transferred to the next level (tier 2). If the same model were used to identify the agent’s training needs during this period, it would fail. In order to ensure the accuracy of the models on an ongoing basis, the results have to be monitored and validated. Methods like hypothesis testing and neural nets can be designed to confirm that the model’s results are within acceptable limits.
Figure 3 - shows the classification of agents based on their training needs. Along with other call handling data, % of calls transferred to tier 2 and % of calls transferred to same queue were added. This helped Contact Center managers to effectively address agent's training needs.
Summary
It is important to use the power of analytics to obtain insights in the data to better manage the performance of your Contact Center. A successful analytical implementation requires focus on people, process and technology areas. EXILANT has built the analytics expertise to help clients implement their performance management projects.
Recent Links :
Benchmarking contact center performance
This case study is centered on how EXILANT helped the client to establish & improve performance management solution for their contact center despite having subsystems in diverse technologies.
Derive the right Contact Center metrics
Metrics help spot issues, identify root causes and control factors affecting customers. However, complexity and proliferation of systems in a contact center makes it difficult to derive the right metrics.
Contact center performance management
Implementing performance management in Contact centers is a challenge due to diverse technologies of its subsystems and poor data quality. Our multi disciplinary approach, helps clients manage such challenges.
