Benchmarking methods, systems, and products for pairing strategies in contact center systems

Abstract

This disclosure provides methods, systems, and artifacts for benchmarking at least two pairing strategies in a contact center system. The method includes: assigning sequence numbers to a series of events; initiating a first pairing strategy based on the sequence numbers; terminating the first pairing strategy based on the sequence numbers; initiating a second pairing strategy based on the sequence numbers; terminating the second pairing strategy based on the sequence numbers; assigning a first set of values ​​to contacts assigned by the first pairing strategy; assigning a second set of values ​​to contacts assigned by the second pairing strategy; and determining a metric comparing the first and second pairing strategies based on the first and second sets of values, wherein at least one contact paired with the first pairing strategy and at least one contact paired with the second pairing strategy are paired with the same agent. The technology of this disclosure overcomes the drawbacks and / or defects of time-based benchmarking.

Description

Technical Field

[0001] This disclosure generally relates to contact centers, and more specifically, to techniques for benchmarking pairing strategies in contact center systems. Background Technology

[0002] A contact center is a system used to receive or transmit large volumes of contact information, such as voice phone calls, web text chat, emails, and video calls. Contact centers can include outbound contact centers, which generate a large volume of outgoing contacts. Such outbound contact centers are typically used for applications such as selling products, collecting outstanding credit balances, or surveying consumer sentiment. Contact centers may also include inbound contact centers, which receive a large volume of incoming contacts from customers. Such inbound contact centers are also used for selling products, as well as for customer service or technical support inquiries, customer retention, or other applications.

[0003] Contact centers may also include Interactive Voice Response (IVR) units that automatically respond to customer inquiries. IVR units can respond to customers pressing numbers on the telephone keypad or using voice recognition tools to respond to their verbal input. Typically, IVR units can handle relatively simple customer inquiries, while more complex customer interactions may require connecting the customer to a human agent. In terms of automating interactions that previously required a human agent, IVR units can reduce the labor costs associated with operating a contact center.

[0004] To distribute a large number of calls among a large number of agents, contact centers can employ algorithms that seek to balance the effort levels of agents across the network. For example, if a contact center has many agents available to receive calls, it might use a simple algorithm to assign incoming calls to the agent with the longest waiting time. Similarly, if all agents are occupied and many calls are waiting to be assigned, the contact center might use a simple algorithm to assign the longest-waiting call to the agent who becomes available first. Such algorithms are called First In First Out (FIFO) algorithms.

[0005] Contact centers may attempt to improve their performance by employing algorithms other than FIFO (First-In, First-Out). For example, if many agents are available to receive calls, the contact center may prioritize assigning a call to an agent with a better historical performance compared to other agents of similar caliber. Similarly, if all agents are occupied and many calls are waiting to be assigned, the contact center may use an algorithm that assigns the most valuable call to the agent who becomes available first. Such algorithms can be called performance-based routing (PBR) algorithms.

[0006] However, while PBR strategies can improve contact center performance, they also have some drawbacks. For example, if there are many contacts in the queue waiting to be assigned to an agent, selecting the highest-value contact may increase the time that the remaining contacts in the queue are waiting to be assigned to an agent. This increase may lead to a decline in the overall customer experience and reduce the overall performance of the contact center.

[0007] Therefore, an improved algorithm relative to the PBR strategy is needed. This algorithm would continue to balance agent workloads, minimize contact waiting time discrepancies, and also improve the performance of the baseline FIFO algorithm. Such algorithms might rely on predicting the likely behavior of customers contacting the contact center, predicting the likely behavior of agents working within the contact center, and assigning customer contacts to agents based on these behavioral predictions. This type of algorithm can be called a Behavioral Pairing ("BP") algorithm.

[0008] To determine which of many potential algorithms is best suited for a contact center, the center can establish test and control groups to attempt to determine the relative performance of two or more algorithms. For example, the contact center can divide its agents into two pools, one using a FIFO (First-In, First-Out) strategy to receive contacts, and the other using a PBR (Progressive Backwards) strategy. After a period of time, the performance of the two agent groups can be measured and compared to estimate the performance difference between the FIFO and PBR strategies.

[0009] However, testing the relative performance of multiple algorithms by dividing agents into corresponding pools can introduce measurement errors. For example, multiple agent pools may differ in agent performance, so one algorithm may appear to be better than another, not because there is a real difference in the performance of the two algorithms, but because one agent pool may have more capable agents. Contact centers find it difficult to control for such errors.

[0010] Another approach to testing the relative performance of multiple algorithms is to avoid pooling agents and instead group contacts. For example, U.S. Patent No. 10,298,763 teaches that algorithms can be rotated over time, such that a first algorithm is used for contacts arriving in a first time period, and a second algorithm is used for contacts arriving in a subsequent second time period. This rotation is then repeated, such that the first algorithm is used again in a third period, and the second algorithm is used again in a fourth period, and so on. Periodically, all contacts assigned by the first algorithm can be grouped and compared with all contacts assigned by the second algorithm in similar groups. This time-based rotation strategy can help eliminate errors when calculating performance differences between contact assignment algorithms.

[0011] However, time-based rotation algorithms may introduce another source of error into measuring the relative performance of contact allocation algorithms. For example, there may be contact behavior patterns related to the time when contacts are received at the contact center. For instance, contacts received in the morning may have different average purchasing propensities than those received in the evening. Similarly, contacts received in the first half hour may have different average purchasing propensities compared to those received in the second half hour. Therefore, grouping contacts based on their arrival time at the contact center can introduce errors when measuring the relative performance of contact allocation algorithms.

[0012] To mitigate this potential bias, U.S. Patent No. 11,070,674 teaches another method for measuring the relative performance of contact allocation algorithms, which involves randomly or pseudo-randomly assigning contacts to different algorithms. In such a strategy, when any contact arrives, it can be randomly assigned to one algorithm. This random assignment of any contact cannot predict the assignment of subsequent contacts. This strategy may help reduce or eliminate time-based artifacts that can bias the measurement of the relative performance of contact allocation algorithms due to timely rotation of the algorithms.

[0013] However, while predetermined time windows can reduce bias, random assignment algorithms can introduce other sources of bias. For example, if the first contact assignment algorithm reflects a PBR (Progressive Business Response) strategy and the second reflects a FIFO (First-In, First-Out) strategy, then random rotation between these two strategies might result in a division of the agent population into high-performing and low-performing groups. The high-performing group might be assigned to the PBR strategy, while the low-performing group is assigned to the FIFO strategy. This could occur because the PBR strategy immediately assigns high-performing agents, while lower-performing agents wait longer. The FIFO strategy, targeting agents with longer wait times, then subsequently assigns these lower-performing agents. Such ability-based agent allocation can lead to inaccurate performance comparisons between multiple contact assignment algorithms.

[0014] Therefore, a fairer and more accurate technique may be needed to measure the matching strategy of contact centers compared to techniques based on seat division, fixed-time rotation, or random assignment. Summary of the Invention

[0015] The method disclosed herein includes: assigning sequence numbers to a series of events; initiating a first pairing strategy based on the sequence numbers; terminating the first pairing strategy based on the sequence numbers; initiating a second pairing strategy based on the sequence numbers; terminating the second pairing strategy based on the sequence numbers; assigning a first set of values ​​to contacts assigned by the first pairing strategy; assigning a second set of values ​​to contacts assigned by the second pairing strategy; and determining a metric comparing the first pairing strategy and the second pairing strategy based on the first set of values ​​and the second set of values, wherein at least one contact paired with the first pairing strategy and at least one contact paired with the second pairing strategy are paired with the same agent.

[0016] Optionally, in the above method, the series of events is a contact arriving at the contact center system.

[0017] Optionally, in the above method, the termination of the first pairing strategy is based on the target number of contacts.

[0018] Optionally, in the above method, ending the second pairing strategy is based on a target ratio of the number of contacts paired with the first pairing strategy to the number of contacts paired with the second pairing strategy.

[0019] Optionally, the above method further includes looping between the first pairing strategy and the second pairing strategy.

[0020] Optionally, in the above method, starting the first pairing strategy, ending the first pairing strategy, starting the second pairing strategy, and ending the second pairing strategy are based on the goal of reducing possible biases in the comparison.

[0021] Embodiments of this disclosure also provide a system for benchmarking at least two pairing strategies in a contact center system, comprising: at least one computer processor connected to a benchmarking module, wherein the at least one computer processor is configured to: assign sequence numbers to a series of events; initiate a first pairing strategy based on the sequence numbers; terminate the first pairing strategy based on the sequence numbers; initiate a second pairing strategy based on the sequence numbers; terminate the second pairing strategy based on the sequence numbers; assign a first set of values ​​to contacts assigned by the first pairing strategy; assign a second set of values ​​to contacts assigned by the second pairing strategy; and determine a metric comparing the first pairing strategy and the second pairing strategy based on the first set of values ​​and the second set of values, wherein at least one contact paired with the first pairing strategy and at least one contact paired with the second pairing strategy are paired with the same agent.

[0022] Optionally, in the above system, the series of events refers to a contact arriving at the contact center system.

[0023] Optionally, in the above system, the at least one computer processor is further configured to terminate the first pairing strategy based on the target number of contacts.

[0024] Optionally, in the above system, the at least one computer processor is further configured to terminate the second pairing strategy based on a target ratio of the number of contacts paired with the first pairing strategy to the number of contacts paired with the second pairing strategy.

[0025] Optionally, in the above system, the at least one computer processor is further configured to cycle between the first pairing strategy and the second pairing strategy.

[0026] Optionally, in the above system, the at least one computer processor is further configured to: initiate the first pairing strategy, terminate the first pairing strategy, initiate the second pairing strategy, and terminate the second pairing strategy based on the goal of reducing possible biases in the comparison.

[0027] Embodiments of this disclosure also provide an article of manufacture for benchmarking at least two pairing strategies in a contact center system, comprising: a non-transitory processor-readable medium; and instructions stored on the medium, wherein the instructions are configured to read from the medium at least one processor connected to a benchmarking module, thereby causing the at least one computer processor to operate to: assign sequence numbers to a series of events; initiate a first pairing strategy based on the sequence numbers; terminate the first pairing strategy based on the sequence numbers; initiate a second pairing strategy based on the sequence numbers; terminate the second pairing strategy based on the sequence numbers; assign a first set of values ​​to contacts assigned by the first pairing strategy; assign a second set of values ​​to contacts assigned by the second pairing strategy; and determine a metric comparing the first pairing strategy and the second pairing strategy based on the first set of values ​​and the second set of values, wherein at least one contact paired with the first pairing strategy and at least one contact paired with the second pairing strategy are paired with the same agent.

[0028] Optionally, in the above-mentioned products, a series of events is the contact reaching the contact center system.

[0029] Optionally, in the above-described article, the at least one computer processor is further operated to terminate the first pairing strategy based on the target number of contacts.

[0030] Optionally, in the above-described article of manufacture, the at least one computer processor is further configured to terminate the second pairing strategy based on a target ratio of the number of contacts paired with the first pairing strategy to the number of contacts paired with the second pairing strategy.

[0031] Optionally, in the above-described article of manufacture, the at least one computer processor is further operated to cycle between the first pairing strategy and the second pairing strategy.

[0032] Optionally, in the above-described article of manufacture, the at least one computer processor is further operated to: initiate the first pairing strategy based on a target reduction in possible deviations in the comparison, terminate the first pairing strategy, initiate the second pairing strategy, and terminate the second pairing strategy. Attached Figure Description

[0033] Figure 1 This is a schematic diagram illustrating a contact center system according to an embodiment of the present disclosure.

[0034] Figure 2 This is a schematic diagram illustrating a FIFO process for assigning contacts to agents according to an embodiment of the present disclosure.

[0035] Figure 3This is a schematic diagram illustrating the PBR process for assigning contacts to agents according to an embodiment of the present disclosure.

[0036] Figure 4 This is a schematic diagram illustrating a method for switching from one allocation algorithm to another in a time-based benchmark test according to an embodiment of the present disclosure.

[0037] Figure 5 This is a schematic diagram illustrating the disadvantages of the method for switching from one allocation algorithm to another in a time-based benchmark test according to embodiments of the present disclosure.

[0038] Figure 6 This is a schematic diagram illustrating a method for switching from one allocation algorithm to another in a random benchmark test according to an embodiment of the present disclosure.

[0039] Figure 7 This is a schematic diagram illustrating a method for switching from one allocation algorithm to another in a method for benchmarking at least two pairing strategies in a contact center system, according to another embodiment of the present disclosure.

[0040] Figure 8 This illustrates another embodiment of the present disclosure of a method for switching from one allocation algorithm to another in a method for benchmarking at least two pairing strategies in a contact center system.

[0041] Figure 9 This is a schematic diagram illustrating the performance comparison between two pairing strategies in a method for benchmarking at least two pairing strategies based on contact sequence numbers, according to an embodiment of the present disclosure. Detailed Implementation

[0042] To more clearly illustrate the technical solutions of this disclosure, the accompanying drawings required for the embodiments of the present invention will be briefly described below. Obviously, the following drawings only relate to some embodiments of the present invention. Based on these drawings, those skilled in the art can obtain all other embodiments without creative effort.

[0043] As used herein, the term "module" can be understood to refer to computing software, firmware, hardware, and / or various combinations thereof, which can be configured as components of network elements, computers, and / or systems. A module should not be construed as software not implemented in hardware or firmware or not recorded on a processor-readable storage medium. These modules can be combined, integrated, separated, and / or replicated to support a variety of applications. These modules can be implemented on multiple devices and / or other components, either locally or remotely. Furthermore, these modules can be removed from one device and added to another, and / or contained in two devices.

[0044] Figure 1This is a schematic diagram of a contact center system 100, which may include an interactive voice response (“IVR”) unit 110, an automated contact distribution (“ACD”) system 120, a benchmark module 130, a series of ten contacts A to J that have arrived at the contact center system 100, and a set of ten agents 141 to 150 available to the contact center system 100.

[0045] Figure 2 This is a schematic diagram illustrating a FIFO process for assigning contacts to agents according to an embodiment of this disclosure. For illustrative purposes, Figure 2 The public call center pool and the corresponding times when the agents can receive calls are shown: Agent 141 becomes available at 9:01, Agent 142 at 9:02, Agent 143 at 9:03, Agent 144 at 9:04, Agent 145 at 9:05, Agent 146 at 9:06, Agent 147 at 9:07, Agent 148 at 9:08, Agent 149 at 9:09, and Agent 150 at 9:10. Figure 2 A set of contacts and their corresponding arrival times at the contact center system 100 are also shown, namely, contact A arrives at 9:11, contact B arrives at 9:12, contact C arrives at 9:13, contact D arrives at 9:14, contact E arrives at 9:15, contact F arrives at 9:16, contact G arrives at 9:17, contact H arrives at 9:18, contact I arrives at 9:19, and contact J arrives at 9:20.

[0046] Contacts A through J arrive at the contact center system 100 in sequence and are processed by the IVR unit 110, which provides automated responses to customer inquiries. More complex customer interactions requiring connection to a live agent are routed to the ACD system 120. The ACD system 120 is responsible for routing contacts A through J to the most suitable agent based on available pairing strategies. By default, i.e., without any pairing strategy, the ACD system 120 may not communicate with the benchmark module 130 and may route contacts 141 through 150 based on FIFO, i.e., the agent with the longest waiting time in the agent queue occupies the first arriving contact.

[0047] for Figure 2The contact group and agent pool are shown. Contact A arrives promptly at 9:11. Agent 141 has been available since 9:01, making it the agent with the longest waiting time (10 minutes) in the agent queue. ACD system 120 routes contact A to agent 141. Next, contact B arrives at 9:12 and is routed to agent 142, who at that time is the agent with the longest waiting time in the agent queue because the previously longest-waiting agent, agent 141, has already been occupied. This strategy is repeated for contacts C to J and agents 143 to 150.

[0048] Figure 3 This is a schematic diagram illustrating the PBR process for assigning contacts to agents according to an embodiment of the present disclosure, wherein, Figure 3 With Figure 2 The same event occurs, the difference being that the allocation algorithm is PBR. Contact center system 100 can choose to allocate contacts to agents based on the PBR algorithm to improve performance relative to FIFO.

[0049] In this scenario, the contact center system 100 will not use a FIFO mechanism to allocate contacts, but will instead allocate seats based on the agents' performance scores, favoring available seats with the highest performance scores.

[0050] like Figure 3 As shown, the performance scores of agents 141 to 150 are 70, 50, 20, 100, 30, 80, 10, 90, 60, and 40, respectively, based on their historical performance. Contact A arrives first at 9:11, and agent 144 has the highest performance score of 100, so the ACD system 120 routes contact A to agent 144. Next, contact B arrives at 9:12 and is routed to agent 148, who at that time is the available agent with the highest performance score (90) in the agent queue, since the previously highest-scoring agent 144 has already been occupied. This strategy is repeated for contacts C to J and the remaining agents. As a result, agent 146 is assigned to contact C, agent 141 to contact D, agent 149 to contact E, agent 142 to contact F, agent 150 to contact G, agent 145 to contact H, agent 143 to contact I, and agent 147 to contact J.

[0051] Figure 4 This is a schematic diagram illustrating a method for switching from one allocation algorithm to another in time-based benchmarking according to an embodiment of the present disclosure. The time-based benchmarking process establishes a rotation time window for the benchmarking of the pairing algorithm. For example, as Figure 4As shown, using a 30-minute time window, all contacts arriving in the first time window (9:01 to 9:30) (i.e., contacts A and B) are routed using pairing algorithm A (i.e., FIFO); all contacts arriving in the second time window (9:31 to 10:00) (i.e., contact C) are routed using pairing algorithm B (i.e., PBR); all contacts arriving in the third time window (10:01 to 10:30) (i.e., contacts D, E, and F) are routed using pairing algorithm A (i.e., FIFO); all contacts arriving in the fourth time window (10:31 to 11:00) (i.e., contacts H, I, and J) are routed using pairing algorithm B (i.e., PBR); and so on.

[0052] like Figure 4 As shown, contact A arrives at 9:12, and contact B arrives at 9:20. Therefore, contacts A and B fall within the first time window and require FIFO routing. Contact C arrives at 9:45, therefore it falls within the second time window and requires PBR routing. Contact D arrives at 10:06, contact E arrives at 10:10, and contact F arrives at 10:29. Therefore, contacts D, E, and F fall within the third time window and should be routed using FIFO. Contacts G through J fall within the fourth time window and should be routed using PBR.

[0053] The time-rotation algorithm may introduce another source of error into the relative performance of the contact allocation algorithm. For example, if a group of high-performing agents tends to be idle during times when they are highly correlated with pairing algorithm A but lowly correlated with pairing algorithm B, then pairing algorithm A may inaccurately demonstrate superiority over pairing algorithm B. Similarly, if a high-performing contact group tends to arrive at contact center system 100 during times when they are highly correlated with pairing algorithm B but lowly correlated with pairing algorithm A, then pairing algorithm B may inaccurately demonstrate superiority over pairing algorithm A.

[0054] Figure 5 This is a schematic diagram illustrating the average contact quality of a method for switching from one allocation algorithm to another in a time-based benchmark test according to embodiments of the present disclosure. The average purchase preference may differ between contacts received in the first half hour and contacts received in the second half hour. For example, as... Figure 5As shown, the average contact quality from 9:01 to 9:30 (60) is higher than the average contact quality from 9:31 to 10:00 (57), the average contact quality from 10:01 to 10:30 (59) is higher than the average contact quality from 10:31 to 11:00 (58), and so on. In this example, the average contact quality of contacts received in the first half hour is always higher than the average contact quality of contacts received in the second half hour. Therefore, grouping contacts based on the time they arrive at the contact center system 100 may introduce errors when measuring the relative performance of the contact allocation algorithm.

[0055] Time-based benchmarking processes can also be influenced by the fact that they are obvious to contact center agents and therefore subject to manipulation. For example, if an agent is aware that a matching algorithm always runs at a specific time of day, the agent might intentionally suppress the algorithm's apparent performance by choosing to underperform at that particular time of day.

[0056] Figure 6 This is a schematic diagram illustrating a method for switching from one allocation algorithm to another in a random benchmark test according to an embodiment of the present disclosure. The random benchmark test process sets the contact pairing algorithm based on a random number or pseudo-random number generated for each contact.

[0057] like Figure 6 As shown, a pseudo-random integer of 1 or 2 is generated for each connection, where pseudo-random integer 1 represents the PBR algorithm and pseudo-random integer 2 represents the FIFO algorithm. Connection A is assigned pseudo-random integer 2, so routing is based on the PBR algorithm, while connection B is assigned pseudo-random integer 1, so routing is based on the FIFO algorithm, and so on.

[0058] However, these benchmarking processes are susceptible to contamination from transformation effects. For example, if a randomized benchmarking process compares the PBR algorithm with the FIFO algorithm, the performance of the PBR algorithm relative to the FIFO algorithm will appear artificially exaggerated because the PBR algorithm will continue to utilize higher-performing seats, while the FIFO algorithm will leave lower-performing seats.

[0059] Figure 6 This also illustrates an example of potential bias in agent selection within the random assignment algorithm. For a series of ten contacts (Contacts A to J) arriving sequentially at Contact Center System 100, a random assignment algorithm was implemented, resulting in the following assignment sequence: PBR, FIFO, FIFO, PBR, PBR, PBR, FIFO, PBR, PBR, and FIFO. For Contact A, agents 141 to 150 were available, each with a corresponding agent score based on their historical performance, where agent 144 had the highest score of 100 and agent 147 had the lowest score of 10.

[0060] For Contact A, the random assignment algorithm produces a PBR (Pick-Up-Break) strategy, which selects the highest-performing agent from the available agent queue. In this case, all 10 agents are available, and Agent 144 is paired because Agent 144 has the highest score. Next, for Contact B, the random assignment algorithm produces a FIFO (First-In, First-Out) strategy, which looks at the agent with the longest waiting time in the queue, which is Agent 141 in this example. The same operation is repeated for Contact C, which results in the pairing of Agent 142, the next agent with the longest waiting time in the queue. However, for Contact D, the random assignment algorithm produces a PBR strategy, which selects the highest-performing agent from the remaining available agent queue, which is Agent 148 with a score of 90 in this example.

[0061] Because the random assignment algorithm works with ten contacts and ten agents in a queue, a pattern emerges: each time the PBR strategy is used to assign agents, the agent with the highest score is selected, while each time the FIFO strategy is used, the agent with the longest waiting time is selected. In the process of assigning ten contacts and ten agents, the random assignment algorithm reveals this bias by creating two agent groups based on performance. Figure 6 This is further illustrated by comparing the average agent performance score of 66.7 for six agents assigned using the PBR strategy with the average agent performance score of 37.5 for four agents assigned using the FIFO strategy. However, without the switching effect, the average agent performance score should be 55, which is the average score for ten agents (i.e., agents 141 to 150). Random algorithms are more susceptible to the contamination of switching effects than other algorithms because random assignment algorithms always have more switching. The term "switching" as used in this paper can be understood as a switch from one assignment algorithm to another.

[0062] In a contact center system 100, the preferred benchmarking process should minimize the impact of agent and contact selection biases among multiple matching algorithms. Such a preferred process should also be imperceptible to agents and contacts to avoid conscious or subconscious biases and to prevent the placebo effect. This process should also accurately determine which matching algorithm was used when assigning each contact and should provide a simple mechanism for subsequently calculating performance differences between matching algorithms.

[0063] Therefore, a fairer and more accurate technology is needed to benchmark the pairing strategy of the contact center system 100.

[0064] Figure 7 This is a schematic diagram illustrating a method for switching from one allocation algorithm to another in a method for benchmarking at least two pairing strategies in a contact center system 100 according to an embodiment of the present disclosure.

[0065] like Figure 7 As shown, the method of this disclosure alternates between at least two pairing strategies based on the length of the contact sequence. For example, the method can alternate between the two pairing strategies in a specific pattern. In this case, the first two contacts (A and B) received by the contact center system 100 can be assigned based on a first algorithm (FIFO). Subsequently, the next three contacts (C, D, and E) received by the contact center system 100 can be assigned based on a second algorithm (BP). Subsequently, the next two contacts (F and G) received by the contact center system 100 can be assigned based on the first algorithm (FIFO). Subsequently, the next three contacts (H, I, and J) can be reassigned based on the second algorithm (BP), and so on. Calculations can be performed periodically to measure the performance difference between contacts paired using the first algorithm and contacts paired using the second algorithm.

[0066] The benchmarking solution disclosed herein can define the number of contacts for each pairing strategy and cycle through the pairing strategies in a round-robin manner. This reduces contamination caused by biases in random benchmarking processes that prioritize assigning higher-performing agents to the PBR algorithm. It also reduces and potentially eliminates contamination caused by contact center event relevance, such as daily shift changes at fixed times, in which case time-based benchmarking processes overexpress specific pairing algorithms.

[0067] Figure 8 This illustrates a method for switching from one allocation algorithm to another in a method for benchmarking at least two pairing strategies in a contact center system 100, according to another embodiment of this disclosure.

[0068] like Figure 8As shown, sequence numbers 1 through 75 are assigned to contact arrivals in contact center system 100. The target number of FIFO strategy contacts in the sequence is set to 5, and the target ratio of FIFO-paired contacts to BP strategy-paired contacts is set to 1:4. Then, the FIFO strategy is determined to start at contact 1 and end at contact 5. The BP strategy is determined to start at contact 6 and end at contact 25, with 20 contacts paired with the BP strategy (5 x 4). The BP and FIFO strategies are then cycled in a round-robin manner. The FIFO strategy is determined to restart at contact 26 and end at contact 30, while the BP strategy is determined to restart at contact 31 and end at contact 50. Then, the first set of performance data can be assigned to contacts assigned by the FIFO strategy, and the second set of performance data can be assigned to contacts assigned by the BP strategy. The performance of the total contacts assigned by the FIFO strategy is compared with the performance of the total contacts assigned by the BP strategy, and the performance difference can be used as a metric for comparing the two strategies. In this example, both strategies can use a common agent pool, and at least one contact paired with the FIFO strategy and at least one contact paired with the BP strategy can be paired with the same agent.

[0069] Figure 9 It shows the basis Figure 8 A schematic diagram comparing the performance of two pairing strategies in a method for benchmarking at least two pairing strategies based on contact sequence numbers.

[0070] like Figure 9 As shown, the average performance of all communications assigned the FIFO strategy in communications 1 to 5 was calculated and plotted; the average performance of all communications assigned the BP strategy in communications 6 to 25 was calculated and plotted; the average performance of all communications assigned the FIFO strategy in communications 26 to 30 was calculated and plotted, and so on. Figure 9 As shown, the average performance of aggregate contacts assigned using the BP strategy is higher than that of aggregate contacts assigned using the FIFO strategy. While the average performance may fluctuate among several contact sequences assigned using the same strategy, overall, the average performance of the BP strategy is expected to be higher than that of the FIFO strategy. For example, the average performance of aggregate contacts assigned using the FIFO strategy is lower for sequences 1-5 than for sequences 25-30, but higher for sequences 51-55. Similar patterns may exist for aggregate contacts assigned using the BP strategy.

[0071] The technology disclosed herein is more resistant to malicious actors who may intentionally cause benchmark bias because, since the choice of strategy is irrelevant to any time period, it is more difficult for agents to determine the strategy being used to handle a particular contact.

[0072] The scope of this disclosure is not limited to the specific embodiments described herein. In fact, various other embodiments and modifications of this disclosure, besides those described herein, will be apparent to those skilled in the art from the foregoing description and drawings. Therefore, such other embodiments and modifications are intended to fall within the scope of this disclosure. Furthermore, although this disclosure has been described herein in the context of at least one specific implementation for at least one specific purpose in at least one specific environment, those skilled in the art will recognize that its usefulness is not limited thereto and that this disclosure can be advantageously implemented in any number of environments for any number of purposes.

Claims

1. A method for benchmarking at least two pairing strategies in a contact center system, comprising: assigning sequence numbers to a series of events, wherein the sequence numbers are assigned to the events in order of the order in which the contact events corresponding to the sequence numbers arrive; starting a first pairing strategy based on the sequence numbers; ending the first pairing strategy based on the sequence numbers; starting a second pairing strategy based on the sequence numbers; ending the second pairing strategy based on the sequence numbers; assigning a first set of values to contacts assigned by the first pairing strategy; assigning a second set of values to contacts assigned by the second pairing strategy; and determining a metric comparing the first pairing strategy and the second pairing strategy based on the first set of values and the second set of values, wherein at least one contact paired with the first pairing strategy and at least one contact paired with the second pairing strategy are paired with the same agent, and the number of sequence numbers started to ended for the first pairing strategy is proportional to the number of sequence numbers started to ended for the second pairing strategy.

2. The method of claim 1, wherein, the series of events are contacts arriving at a contact center system.

3. The method of claim 1, wherein, ending the first pairing strategy is based on a target number of contacts.

4. The method of claim 1, wherein, ending the second pairing strategy is based on a target ratio of a number of contacts paired with the first pairing strategy to a number of contacts paired with the second pairing strategy.

5. The method of claim 1, further comprising cycling between the first pairing strategy and the second pairing strategy.

6. The method of claim 1, wherein, starting the first pairing strategy, ending the first pairing strategy, starting the second pairing strategy, and ending the second pairing strategy are based on targetingly reducing potential bias in the comparison.

7. A system for benchmarking at least two pairing strategies in a contact center system, comprising: at least one computer processor connected to a benchmarking module, wherein the at least one computer processor is configured to: assign sequence numbers to a series of events, wherein the sequence numbers are assigned to the events in order of the order in which the contact events corresponding to the sequence numbers arrive; start a first pairing strategy based on the sequence numbers; end the first pairing strategy based on the sequence numbers; start a second pairing strategy based on the sequence numbers; end the second pairing strategy based on the sequence numbers; assign a first set of values to contacts assigned by the first pairing strategy; assign a second set of values to contacts assigned by the second pairing strategy; and determine a metric comparing the first pairing strategy and the second pairing strategy based on the first set of values and the second set of values, wherein at least one contact paired with the first pairing strategy and at least one contact paired with the second pairing strategy are paired with the same agent, and the number of sequence numbers started to ended for the first pairing strategy is proportional to the number of sequence numbers started to ended for the second pairing strategy.

8. The system of claim 7, wherein, the series of events are contacts arriving at a contact center system.

9. The system of claim 7, wherein, the at least one computer processor is further configured to end the first pairing strategy based on a target number of contacts.

10. The system of claim 7, wherein, the second pairing strategy based on a target ratio of a number of contacts paired with the first pairing strategy to a number of contacts paired with the second pairing strategy.

11. The system of claim 7, wherein, the at least one computer processor is further configured to cycle between the first pairing strategy and the second pairing strategy.

12. The system of claim 7, wherein, the at least one computer processor is further configured to initiate the first pairing strategy, end the first pairing strategy, initiate the second pairing strategy, and end the second pairing strategy based on a targeted reduction in potential bias in the comparison.

13. An article of manufacture for benchmarking at least two pairing strategies in a contact center system, comprising: a non-transitory processor-readable medium; and instructions stored on the medium, wherein the instructions are configured to be readable from the medium by at least one processor in connection with a benchmarking module to cause the at least one computer processor to operate so as to: assign sequence numbers to a series of events, wherein the sequence numbers are assigned to the events in an order corresponding to an order in which the contact events corresponding to the sequence numbers arrive; initiate a first pairing strategy based on the sequence numbers; end the first pairing strategy based on the sequence numbers; initiate a second pairing strategy based on the sequence numbers; end the second pairing strategy based on the sequence numbers; assign a first set of values to contacts assigned by the first pairing strategy; assign a second set of values to contacts assigned by the second pairing strategy; and determine a metric comparing the first pairing strategy and the second pairing strategy based on the first set of values and the second set of values, wherein at least one contact paired with the first pairing strategy and at least one contact paired with the second pairing strategy are paired with a same agent, and a number of sequence numbers initiated to ended for the first pairing strategy is proportional to a number of sequence numbers initiated to ended for the second pairing strategy.

14. The article of claim 13, wherein, the series of events are contacts arriving at a contact center system.

15. The article of claim 13, wherein, the at least one computer processor is further caused to operate so as to end the first pairing strategy based on a target number of contacts.

16. The article of claim 13, wherein, the at least one computer processor is further caused to operate so as to end the second pairing strategy based on a target ratio of a number of contacts paired with the first pairing strategy to a number of contacts paired with the second pairing strategy.

17. The article of claim 13, wherein, the at least one computer processor is further caused to operate so as to cycle between the first pairing strategy and the second pairing strategy.

18. The article of claim 13, wherein, the at least one computer processor is further caused to operate so as to initiate the first pairing strategy, end the first pairing strategy, initiate the second pairing strategy, and end the second pairing strategy based on a targeted reduction in potential bias in the comparison.

Images