With digitization of almost all industries on the way, advanced technologies like machine learning are revolutionizing the way of work for most industries today. Many customer service centers are already thinking about adopting machine learning for their day to day operations and these techniques will soon be a part of industry standard best practices. It is therefore imperative for all call centers to adopt new age technology for improving their performance metrics and to stay competitive.
This post describes how traditional call centers can create a strategy for adopting machine learning capabilities by evaluating the technical capabilities offered against their KPIs.
Every customer service operations manager has the following business goals in mind-
To achieve these goals, most call centers will be tracking one or more of the below KPIs-
Customer service leads can leverage various machine learning capabilities to meet these KPIs. Below are some of the use cases of machine learning that can be applied to common traditional call center processes.
Machine learning capabilities for call centers-
Which capabilities are the most effective for an organization and give the best ROI will depend on various factors like the call center structure, KPIs tracked and implementation challenges. We should note that each KPI improvement drives improvement for other KPIs as well. E.g. If the AHT goes down it will also indirectly have a positive impact on CSAT as customer queries will be resolved faster. Below is a look at various machine learning capabilities and the impact that these will have on KPIs.
For understanding the ROI let’s assume a call center with 500 call center agents with a call volume of 15 per agent per day.
*Assuming issue classification by a manual gatekeeper takes 0.5 days on an average
**Assuming 2 minutes spent on the activity per agent per case (2 minutes x 500 agents x 15 calls per day = 250 man hours)
***Assuming like most organizations, 60% of the requests are driven out of customer queries that are simple in nature and do not need manual intervention like status updates, product information, tier 1 queries. Cost saving = (60% of (500 agents x 8 hours) = 2400 man hours)
**** CSAT is measured as number of customers who are satisfied out of the total list surveyed. On a 5 point scale and a benchmark of 4.1 or 80% satisfied customers a case volume of 1 lakh cases per month and a response rate of surveys as 5%, one of the top reasons for customer dissatisfaction is having to repeat information. Assuming reduction in dissatisfied customers by atleast 5-10%, the CSAT score will increase by 0.04)
The metrics that are important for a service organization will depend on a number of factors like the size of the organization (50-100 employees to >100 employees), the domain that the business caters to like finance, healthcare, retail, travel, hospitality etc., does the organization have an existing CRM implementation, does it have the data that is required for it to get started, channels used for support by the organization, is the call center outsourced or in-house amongst other factors.
Business factors (cost, domain) that define how various call centers are different-
Other factors that are decided by IT of the support organization with business
Taking a few examples below of typical call center setups-
SMB (Small and medium scale businesses)
For example, the support org of an SMB will be characterized by
An SMB will think about CRM implementation when the volume of queries becomes difficult to handle without it. ML and AI application KPIs that might be more relevant for an organization of this nature will be –
An insurance customer service org is a perfect example of a support organization that cares most about CSAT and repeat purchases with a few unique characteristics –
The ML/AI capabilities useful for a customer service center in insurance will be-
Once the service org decides as to what are the key areas of investment for implementing machine learning capabilities from a business and KPI perspective, they might need to look at pre-requisites in terms of data and systems along with implementation challenges. Below is an example of how an issue classification ML model can be implemented on a call center site.
Issue classification – Phases of implementation in a call center using available models:
The service org can decide to either create a model of its own from scratch or deploy an existing model from any of the products available in the market. Building a model from scratch will need expertise in the area of data science and machine learning and since this is generally not a skill set available in a customer service /IT organization, a more viable option will be to use an existing product from the market to meet the service organization needs. At this point it is important to choose a product that has capabilities that require minimal customization and no prior data science knowledge. In the above example of issue classification, the model has been coded and trained on a generic data set and incorporates advanced ML techniques like automatic feature extraction and automatic selection of appropriate data classification models to be chosen to give maximum accuracy. The steps required by the call center IT team to implement this model would be-
Once the model has been trained, it should be possible to use it in day to day operations to increase productivity and lower costs.
Capturing below implementation challenges that might exist while applying an ML solution to the call center.
Requires domain knowledge
After the customer service manager takes into consideration the cost benefit analysis, effort involved and data pre-requisites, he can decide on the next steps to implement machine learning based models for his call center and hence make it more productive.
- Neerja Rewal