Göran Svensson, Research Lead at Teleopti in the field of Optimization and Systems Theory, recounts the key factors explored in his mathematical research into multi-skilled call centers. How can Quality of Service be maintained when scheduling multiple skills with a limited budget and number of available agents for certain skills?
This year I attended International Conference on Operations Research and Enterprise Systems (ICORES) 2018 in Portugal, where I presented my research results on resource allocations for a system of multi-class, multi-server queues. The ICORES conference covers topics in Operations Research (OR) and Systems Engineering. WFM is a subfield of these more general disciplines. The days were filled with interesting talks, many occurring in parallel sessions. One of the keynote speakers was a professor from Technion, which is one of the more prestigious universities for OR in general and WFM in particular. There were several presentations on optimization methods, resource management and decision analysis. The talks spanned most of the more mathematical parts of OR, from military logistics to fairness in healthcare scheduling.
I presented my work in the field of multi-skilled call centers and the corresponding decision processes. The problem is solved as a multi-objective optimization problem via the Marginal Allocation Algorithm with constraints on the total budget and availability of agents (customer-facing employees).
That a problem is multi-objective simply means that one tries to optimize several factors at the same time. Normally such a problem produces what is known as an efficient front or a pareto solution. This means that one gets a set of solutions where one goal cannot be improved without obtaining worse outcomes for the other goal(s). It is then commonly up to an expert to choose which of these solutions best serves their specific needs. An example would be the tradeoff between service quality and the cost of using more agents.
The first objective (goal) is to keep the cost of employed agents as low as possible, the second objective is to deliver as good a service to customers as possible. The cost of agents is kept low by using fewer agents while the quality of service improves with an increase in the number of agents employed. Now we can clearly see that there is a conflict between these two goals.
The service provided is measured by something known as a Quality of Service (QoS) measure. Average Speed of Answer (ASA), Service Level (SL, which is a type of Value-at-Risk measure for the ratio of calls, or other inbound contact types, answered within a certain amount of time) and probability of delay are different types of such QoS measures. In the presentation I focused on two QoS measures: One that is closely related to the Service Level measure, called the Conditional Value-at-Risk (CVaR), and the other which is based on the fraction of customers abandoning the queue before receiving service.
The CVaR measure is in many ways superior to the VaR type measure of Service Level. It has some nice mathematical features as well as providing a means to control the outcome, not just for the customers that receive service in time, but also lends itself to control the outcomes for the customers not serviced within the acceptable time.
The optimization method used is known as the Marginal Allocation Algorithm. It is an iterative algorithm that step by step adds the agent that provides the greatest marginal benefit and provides the corresponding efficient point. The main advantages of using this algorithm is that it is easy to implement and that it can solve large systems at a low computational cost. The main disadvantage is that it requires strong assumptions on the goal functions. One such requirement is that the functions should be what is known as convex (read more about convex functions here). Convexity is a characteristic that simplifies optimization procedures in general, which is in addition to the existence of many reliable software solvers for convex problems.
One important contribution of this work is to show that the CVaR (Conditional Value-at-Risk) measure is convex in the number of agents employed, thus it may be used in conjunction with the Marginal Allocation Algorithm.
In my case, a problem with N different queues, representing different skills, is considered when there is a budget constraint on the system, as well as a limit on the available agents with certain skills. In the work, I compare the two different choices of the two Quality of Service (QoS) measures, CVaR and abandonment based. I highlight the similarities as well as the differences. A large-scale system is also provided and then solved (quickly) to show the power of the Marginal Allocation Algorithm.
The paper, on which I based my presentation, had been peer reviewed by three external professionals and has been published in the proceedings booklet. I have also been invited to extend my paper and have it published in a book by the Springer publishing company, as a chapter. The audience was attentive and interested, which led to a fruitful discussion on measures.
This means that my purposed method can be used to procure the staffing needs for large queueing systems quickly. The CVaR measure is suited for situations where all customer service times are of importance, like in a healthcare situation.
Publish Date: August 10, 2018 5:00 AM
Combines knowledge management with call support as continue improvement.
eGain Knowledge + AI
eGain's knowledge management software, powered with Case-Based Reasoning, an Artificial Intelligence technology, lets you provide intranet or extranet access to the common knowledge base, enabling contact center agents to provide distinctive, productive, and brand-aligned experiences in every interaction, across all channels and touchpoints.
FuzeDigital offers an affordable yet comprehensive knowledge base to answer your consumers' and staff questions. When assistance is needed, our email management system ensures your timely and accurate delivery of responses. Used by companies large and small that seek to deflect common questions while providing exceptional support.
Knowmax is a knowledge management software for enterprises. Customer service of any organization can leverage this tool to create, curate & distribute the knowledge at assisted as well as digital channels promoting self service.
Knowledge Management software
livepro are experts in Customer Experience Knowledge Management and are passionate about improving customer experience. livepro is feature-rich yet easy to use, delivering answers to agents – not long complex documents to dig through. This makes customer service quicker, easier and more efficient. Staff require next to no training on complex procedures thanks to livepro’s intuitive design, which brings confidence up and training costs down.
Increase customer loyalty, drive revenue and boost customer lifetime value with the Synthetix suite of AI-powered integrated customer contact tools.
Synthetix was founded in 2001 and since then has become a leading innovator in creating multi-channel online customer service software. We are also known as one of the pioneers for AI-driven Virtual Agent technology, creating Virtual Agents for companies like the BBC, Vauxhall, and Ford, comparable to Siri, Cortana and Google Now, before they became the norm.
We believe that customer service across all customer touch-points should be simple, efficient and, above all, consistent. Our technology has assisted in reducing our clients’ inbound...
PH: +441279 5555 80
|Data is everything in the new era of customer service||May 20, 2019 5:00 AM|
|Real-Time Adherence – the art of smarter scheduling||May 8, 2019 5:00 AM|
|Evolving employee engagement with Workforce Management (WFM)||April 30, 2019 5:00 AM|
|Guest Blog: There’s a brighter future for contact centre agents||April 12, 2019 5:00 AM|
|Centralized or de-centralized WFM: 5 ways to strike the perfect balance||April 8, 2019 5:00 AM|
|How patient are your customers? Erlang A may have the answer||March 25, 2019 5:00 AM|
|Evaluating if the integration between your WFM & CX platform will be a big “Yes”||February 25, 2019 5:00 AM|
|The whole organization is a customer service center – not everyone knows it yet!||February 4, 2019 5:00 AM|
|Resistance to change – the arch nemesis of customer experience||January 22, 2019 5:00 AM|
|5 steps to supercharge your customer service with contextual intelligence||January 7, 2019 5:00 AM|