HK1251383B - Techniques for hybrid behavioral pairing in a contact center system - Google Patents

Description

Technology for hybrid behavior pairing in contact center systems

CROSS-REFERENCE TO RELATED APPLICATIONS

This patent application claims priority to U.S. Application No. 14/956,074, filed December 1, 2015, which is a continuation-in-part of U.S. Application No. 14/871,658, filed September 30, 2015 (now U.S. Patent No. 9,300,802, issued March 29, 2016), which is a continuation-in-part of U.S. Application No. 12/021,251, filed January 28, 2008, and which is a continuation-in-part of U.S. Application No. 14/530,058, filed October 31, 2014 (now U.S. Patent No. 9,277,055, issued March 1, 2016). Continuation-in-Part application, which is a continuation-in-part of U.S. patent application No. 13/843,724, filed on March 15, 2013 (now U.S. Patent No. 8,879,715, issued on November 4, 2014), which claims priority to U.S. Provisional Patent Application No. 61/615,788, issued on March 26, 2012, U.S. Provisional Patent Application No. 61/615,779, filed on March 26, 2012, and U.S. Provisional Patent Application No. 61/615,772, filed on March 26, 2012, each of which is incorporated herein by reference in its entirety as if fully set forth herein.

Technical Field

The present disclosure relates generally to a contact center, and more particularly, to a technique for hybrid behavior pairing in a contact center system.

Background Art

A typical contact center algorithmically assigns contacts arriving at the contact center to agents available to handle those contacts. Sometimes, a contact center may have agents available and waiting to be assigned to an inbound contact or an outbound contact (e.g., a phone call, an internet chat session, an email) or an outbound contact. At other times, a contact center may have contacts in one or more queues waiting for an agent to become available for assignment.

In some typical contact centers, contacts are assigned to agents based on a time-of-arrival ranking. This strategy may be referred to as a "first-in, first-out," "FIFO," or "round-robin" strategy. In some contact centers, contacts or agents are assigned to different "skill groups" or "queues," after which a FIFO assignment strategy is applied within each such skill group or queue. These "skill groups" may also incorporate strategies for prioritizing individual contacts or agents within a baseline FIFO ranking. For example, high-priority contacts may be given queue positions ahead of other contacts that arrived earlier, or high-performing agents may be ranked ahead of other agents who have been waiting longer for their next call. Regardless of such changes in the queue or queues forming the caller or the ranking of available agents, contact centers typically apply FIFO to queues or other rankings. Once such a FIFO strategy has been established, assigning contacts to agents is automatic, with the contact center assigning the first contact in the ranking to the next available agent, or assigning the first agent in the ranking to the next arriving contact. In the contact center industry, the process of contact and agent distribution among skill queues, prioritization and sorting of skill queues, and subsequent FIFO assignment of contacts to agents is managed by a system known as an "Automatic Call Distributor" ("ACD").

Some contact centers may use a "priority queuing" or "PQ" approach to sorting the queues of waiting contacts. For example, contacts waiting to be assigned to an agent may be sorted by the highest-priority waiting contacts (e.g., those that contribute to the following: highest sales conversion rate, highest customer satisfaction score, shortest average handle time, highest performing agent for a particular contact profile, highest customer retention rate, lowest customer retention cost, highest first-call resolution rate). PQ sorting strategies attempt to maximize the expected outcome of each contact-agent interaction, but do so without generally considering evenly utilizing contacts in the contact center. Consequently, low-priority contacts may experience significantly longer wait times.

In view of the foregoing, it can be appreciated that there is a need for a system that attempts to utilize agents more evenly than PQ while improving contact center performance beyond what is presented by a FIFO strategy.

Summary of the Invention

A technique for hybrid behavior pairing in a contact center system is disclosed. In one particular embodiment, the technique can be implemented as a method for hybrid behavior pairing in a contact center system, comprising: ranking agents; ranking a plurality of contacts; applying, by at least one processor, a blending function to the ranking of the plurality of contacts to bias a first strategy for pairing toward a second strategy for pairing; comparing, by the at least one processor and based on the blending function, a first difference in ranking between an agent in a first pair and a first contact with a second difference in ranking between an agent in a second pair and a second contact different from the first contact; and selecting, by the at least one processor, the first pair or the second pair for connection based on the comparison.

According to other aspects of this particular embodiment, selecting the first pair or the second pair based on the comparison may further include applying, by at least one processor, a diagonal strategy to the sorting.

According to other aspects of this particular embodiment, the ranking of agents or the ranking of multiple contacts may be expressed as a percentage.

According to other aspects of this particular embodiment, the ranking of agents or the ranking of multiple contacts may be expressed as a percentage range.

According to other aspects of this particular embodiment, each contact in the plurality of contacts can be assigned a percentage within a corresponding percentage range for each contact.

According to other aspects of this particular embodiment, the allocated percentage may be the midpoint of a range of percentages.

According to other aspects of this particular embodiment, the allocated percentage may be any percentage within a range of percentages.

In accordance with other aspects of this particular embodiment, the method may further include determining, by the at least one processor, bandwidth for the agents in proportion to the relative performance of the agents.

In accordance with other aspects of this particular embodiment, the hybridization function can be enabled by at least one processor to controllably target unbalanced contact utilization.

In accordance with other aspects of this particular embodiment, applying the hybridization function may further include determining, by the at least one processor, a disproportionate bandwidth for each of the plurality of contact types.

According to other aspects of this particular embodiment, the selected contact or similar type of the selected pair may not be any of the following: a contact or contact type that lags in a fairness metric, a contact or contact type that is highest rated in a value, priority, or performance metric, a contact or contact type that is highest rated in a value, priority, or performance metric for a particular agent, a contact that was previously assigned to an agent of the selected pair, a sequentially labeled contact or contact type, or a randomly selected contact or contact type.

In other aspects of this particular embodiment, the selected one of the first pair and the second pair can include a worse expected immediate outcome than the other of the first pair and the second pair.

According to other aspects of this particular embodiment, higher ranked agents may remain available for subsequent assignment to similarly higher ranked contacts, or higher ranked contacts may remain available for subsequent assignment to similarly higher ranked agents.

According to other aspects of this particular embodiment, the first policy may include a behavior pairing policy, and the second policy may include a priority queuing policy.

According to other aspects of this particular embodiment, each successively higher-ranked contact of the plurality of contacts may be more likely to be selected than a corresponding lower-ranked contact.

According to other aspects of this particular embodiment, each successively higher-ranked contact in the plurality of contacts may include a lower average wait time than a corresponding lower-ranked contact.

In another particular embodiment, the technology can be implemented as a method for hybrid behavior pairing in a contact center system, including: ranking agents, by at least one processor; determining, by at least one processor, a first ranking of a plurality of contacts based on a first strategy for pairing; determining, by at least one processor, a second ranking of the plurality of contacts based on a second strategy for pairing; applying, by at least one processor, a blending function to combine the first ranking with the second ranking; comparing, by at least one processor and based on the blending function, a first difference in ranking between an agent in a first pair and a first contact with a second difference in ranking between an agent in a second pair and a second contact different from the first contact; and selecting, by at least one processor, the first pair or the second pair for connection based on the comparison.

In another particular embodiment, the technology may be implemented as a system for hybrid behavior pairing in a contact center system, comprising: at least one processor, wherein the at least one processor is configured to: rank agents; rank a plurality of contacts; apply a blending function to the rank of the plurality of contacts to bias a first strategy for pairing toward a second strategy for pairing; compare, based on the blending function, a first difference in rank between an agent in a first pair and a first contact with a second difference in rank between an agent in a second pair and a second contact different from the first contact; and, based on the comparison, select the first pair or the second pair for connection.

In accordance with other aspects of this particular embodiment, the at least one processor may be further configured to controllably target unbalanced agent utilization using a blending function.

In accordance with other aspects of this particular embodiment, the at least one processor may be further configured to determine a disproportionate bandwidth for each contact type of the plurality of contact types.

The present disclosure will now be described in more detail with reference to specific embodiments thereof as shown in the accompanying drawings. Although the present disclosure is described below with reference to specific embodiments, it should be understood that the present disclosure is not limited thereto. Those of ordinary skill in the art having access to the teachings herein will recognize additional implementations, modifications and embodiments, as well as other areas of use that are within the scope of the present disclosure as described herein and may have significant utility relative to the present disclosure.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to facilitate a more complete understanding of the present disclosure, reference is now made to the accompanying drawings, in which like elements are referenced using like reference numerals. These drawings should not be construed as limiting the present disclosure, but are intended to be illustrative only.

FIG1 shows a block diagram of a contact center according to an embodiment of the present disclosure.

FIG2 shows a schematic representation of a queue according to an embodiment of the present disclosure.

FIG3 shows a schematic representation of a queue according to an embodiment of the present disclosure.

FIG4 shows a schematic representation of a queue according to an embodiment of the present disclosure.

FIG5 shows a schematic representation of a queue according to an embodiment of the present disclosure.

FIG6 shows a schematic representation of a queue according to an embodiment of the present disclosure.

FIG7 shows a schematic representation of a queue according to an embodiment of the present disclosure.

FIG8 shows a flowchart of a hybrid behavior pairing method according to an embodiment of the present disclosure.

DETAILED DESCRIPTION

A typical contact center algorithmically assigns contacts arriving at the contact center to agents available to handle them. At times, a contact center may be in an "L1 state" and have agents available and waiting to be assigned to inbound or outbound contacts (e.g., phone calls, internet chat sessions, emails). At other times, a contact center may be in an "L2 state" and have contacts waiting in one or more queues for agents to become available for assignment. Such L2 queues may be inbound, outbound, or virtual queues. Contact center systems implement various strategies for assigning contacts to agents in both the L1 and L2 states.

The present disclosure generally relates to contact center systems, traditionally referred to as "Automated Call Distribution" ("ACD") systems. Typically, such ACD processes follow an initial "Skills-Based Routing" ("SBR") process, which serves to distribute contacts and agents among skill queues within the contact center. Such skill queues may differentiate contacts and agents based on language proficiency, customer needs, or agent proficiency at a particular task group.

The most common traditional method of assignment within a queue is "first-in, first-out" or "FIFO" assignment, where the longest-waiting contact is assigned to the longest-waiting agent. Some contact centers implement "priority queuing" ("PQ"), where the next available agent is assigned to the highest-priority contact. Two variations of this assignment method generally exist.

FIFO variations generally aim for "fairness" in that they are designed to balance the distribution of contacts to agents ("utilization") over time. The PQ variation of FIFO takes a different approach, where the distribution of contacts to agents is purposefully skewed to increase utilization for higher-priority contacts and decrease utilization for lower-priority contacts. PQ does this despite the potential negative impact on lower-priority contacts.

This disclosure provides an optimization strategy for assigning contacts to agents that improves upon traditional assignment methods, such as "behavior pairing" or "BP" strategies. Behavior pairing targets balanced utilization of both agents and contacts within a queue (e.g., a skills queue), while potentially improving overall contact center performance beyond what FIFO or similar approaches would actually achieve. This is a remarkable achievement because BP operates on the same contacts and the same agents as FIFO, approximately balancing contact utilization as FIFO provides, while improving overall contact center performance beyond what FIFO actually provides.

BP improves performance by assigning agent and contact pairs in a way that takes into account potential subsequent agent and contact pair assignments, so that when the benefits of all assignments are aggregated, they can surpass the benefits of FIFO and PQ strategies. In some cases, BP results in instantaneous contacts and agent pairs that may be the opposite of what FIFO or PQ would dictate. For example, in an instantaneous situation, BP may select the shortest-waiting contact or the lowest-performing available agent. BP respects "posterity" because the system assigns contacts to agents in a way that inherently forgets what the highest-performing choice might be at the moment if such a decision increases the probability of better contact center performance over time.

As explained in detail below, embodiments of the present disclosure relate to a technique for "Hybrid Behavior Pairing" ("HBP") that combines the strategy of BP with the strategy of priority queuing in a way that a contact center administrator can adjust the balance between the two. For example, a contact center administrator can choose to make BP the dominant mechanism for allocating contacts from within a queue and bias it toward PQ. Instead of aiming to balance contact utilization, HBP can aim to skew contact utilization. In some configurations, this bias or skew can be slight; for example, the HBP policy can be calibrated to reduce or limit the number of occasions in which contacts of any one type in a queue (e.g., a skills queue) are assigned to more than one agent pairing before being assigned to contacts of other types in the queue.

FIG1 shows a block diagram of a contact center system 100 according to an embodiment of the present disclosure. The description herein describes network elements, computers, and/or components for simulating systems and methods for a contact center system that may include one or more modules. As used herein, the term "module" may be understood to refer to computing software, firmware, hardware, and/or various combinations thereof. However, a module is not to be interpreted as software that is not implemented in hardware, firmware, or recorded on a processor-readable, recordable storage medium (i.e., a module is not software itself). It should be noted that modules are exemplary. Modules may be combined, integrated, separated, and/or replicated to support various applications. Furthermore, instead of or in addition to functions performed at a particular module, functions described herein as being performed at a particular module may be performed at one or more other modules and/or by one or more other devices. Furthermore, modules may be implemented across multiple devices and/or other components that are local or remote to one another. Furthermore, modules may be moved from one device to another and added to another, and/or included in both devices.

As shown in FIG1 , a contact center system may include a central switch 110. The central switch 110 may receive incoming contacts (e.g., callers) or support outbound connections to contacts via a dialer, telecommunications network, or other module (not shown). The central switch 110 may include contact routing hardware and software to help route contacts between one or more contact centers or to one or more PBX/ACDs or other queuing or switching components within a contact center.

In the contact center system 100, if there is only one contact center or if there is only one PBX/ACD routing component, then the central switch 110 may not be necessary. If more than one contact center is part of the contact center system 100, each contact center may include at least one contact center switch (e.g., contact center switches 120A and 120B). Contact center switches 120A and 120B may be communicatively coupled to the central switch 110.

Each contact center switch for each contact center can be communicatively coupled to a plurality of agents (or "pools" of agents). Each contact center switch can support a certain number of agents (or "seats") logged in at one time. At any given time, a logged-in agent may be available and waiting to be connected to a contact, or a logged-in agent may be unavailable for any of a number of reasons, such as being connected to another contact, performing some post-call function (such as logging information about a call), or resting.

In the example of FIG1 , central switch 110 routes contacts to one of two contact centers via contact center switch 120A and contact center switch 120B, respectively. Each of contact center switches 120A and 120B is shown with two agents. Agents 130A and 130B may be logged into contact center switch 120A, and agents 130C and 130D may be logged into contact center switch 120B.

The contact center system 100 can also be communicatively coupled to integration services from, for example, third-party vendors. In the example of FIG1 , the hybrid behavior pairing module 140 can be communicatively coupled to one or more switches in the switch system of the contact center system 100, such as the central switch 110, the contact center switch 120A, or the contact center switch 120B. In some embodiments, the switches of the contact center system 100 can be communicatively coupled to multiple hybrid behavior pairing modules. In some embodiments, the hybrid behavior pairing module 140 can be embedded within a component of the contact center system (e.g., embedded within a switch or otherwise integrated therewith).

Hybrid behavior pairing module 140 can receive information about agents logged into the switch (e.g., agents 130A and 130B) and about incoming contacts via another switch (e.g., central switch 110) from a switch (e.g., contact center switch 120A) or, in some embodiments, from a network (e.g., the Internet or a communications network) (not shown).

The hybrid behavior pairing module 140 can process this information and determine which contacts should be paired (e.g., matched, assigned, distributed, routed) with which agents. For example, multiple agents are available and waiting for a connection to a contact (L1 state), and the contact arrives at the contact center via the network or a central switch. As explained below, without the hybrid behavior pairing module 140 or a similar behavior pairing module, under a "fair" FIFO policy, the contact center switch would typically automatically distribute the new contact to any available agent who has been waiting for the longest amount of time for an agent, or under another policy such as a performance-based routing ("PBR") policy, to any available agent who has been determined to be the highest-performing agent.

In the case of a hybrid behavioral pairing module 140 or similar behavioral pairing module, contacts and agents can be assigned scores (e.g., percentages or percentage ranges/bandwidths) based on a pairing model or other artificial intelligence data model so that contacts can be matched, paired, or otherwise connected to preferred agents. In some embodiments, the hybrid behavioral pairing module 140 can be configured with an HBP strategy that is a hybrid of BP and PBR strategies, targeting biased rather than balanced agent utilization.

In the L2 state, multiple contacts are available and waiting to be connected to an agent, and the agent becomes available. These contacts may be queued in a contact center switch, such as a PBX or ACD device ("PBX/ACD"). Without the hybrid behavior pairing module 140 or a similar behavior pairing module, when an agent chooses to be unavailable, the contact center switch will typically connect the newly available agent to any contact that has been waiting in the queue for the longest amount of time, as in a "fair" FIFO policy or a PBR policy. In some contact centers, priority queuing may also be incorporated.

In the case of a hybrid behavior pairing module 140 or similar behavior pairing module in an L2 scenario, as in the L1 state described above, contacts and agents can be given percentages (or percentage ranges/widths, etc.) based on, for example, a model (such as other artificial intelligence models) so that the soon-to-be-available agents can be matched, paired, or otherwise connected to preferred contacts.

Under the HBP policy, a blending factor or function may be applied to one or more rankings of agents to achieve a desired balance between the BP policy, which targets balanced utilization, and the PQ policy, which targets highly skewed utilization during periods when the contact center is in the L2 state (i.e., multiple contacts awaiting assignment).

In some embodiments, the hybridization function can combine two (or more) sorting or other types of ranking systems. For example, a contact center may have four contacts of different types: Contact A, Contact B, Contact C, and Contact D ("A," "B," "C," and "D") available for pairing with agents. Contacts can be ranked according to multiple sorting systems. For example, under a typical FIFO policy, agents can be ranked based on how long each contact has been waiting for an assignment relative to the other contacts. Under a typical priority queuing policy, contacts can be ranked based on how well each contact performs on a metric relative to the other contacts. Under a BP policy, agents can be ranked based on the quality of each agent's "behavioral fit" relative to the other agents.

One technique for combining two rankings is to determine the sum. For example, if the PQ policy ranks four contacts as A=1, B=2, C=3, and D=4, the PQ policy will prefer to pair the highest "performing" contact A with the next agent. And if the BP policy ranks the contacts as A=4, B=2, C=3, and D=1, the BP policy will prefer to pair the best-fitting contact D with the next agent. In this example of an HBP policy, the sum of the two rankings would be A=5, B=4, C=6, and D=5. The HBP policy would prefer to pair contact B with the next agent, which is the second-highest-performing and second-best-fit agent according to the original ranking.

Other embodiments may use other techniques for combining multiple rankings of agents. For example, the HBP ranking may be a product obtained by multiplying two or more rankings together. For another example, the HBP ranking may be a weighted sum or product obtained by scaling one or more of the rankings by a scaling factor. In this way, the HBP may be configured to weight the relative performance of the agents more or less than their relative behavioral fit.

FIG2 illustrates a queue 200 according to an embodiment of the present disclosure operating under a BP strategy 210. The queue 200 represents a simplified hypothetical scenario in which four types of contacts can be assigned to any of four agents in an environment where a contact center seeks to maximize a desired metric (e.g., sales). The four evenly distributed types of contacts are assigned percentage ranges (or "bandwidths"): 0.00 to 0.25 ("0-25% of contacts"), 0.25 to 0.50 ("25-50% of contacts"), 0.50 to 0.75 ("50-75% of contacts"), and 0.75 to 1.00 (75-100% of contacts). The four agents occupy equally spaced percentage bandwidths and are assigned percentages at the midpoints of their respective ranges: 0.00 to 0.25 ("0.125 agent"), 0.25 to 0.50 ("0.375 agent"), 0.50 to 0.75 ("0.625 agent"), and 0.75 to 1.00 ("0.875 agent"). The four agents can also be sorted by performance according to a desired metric (e.g., sales), such that the lowest performing agent is assigned the lowest percentage (0.125 agent) and the highest performing agent is assigned the highest percentage (0.875 agent).

By applying the diagonal strategy, 0-25% of contacts can be preferentially assigned to 0.125 agents, 25-50% of contacts can be preferentially assigned to 0.375 agents, 50-75% of contacts can be preferentially assigned to 0.625 agents, and 75-100% of contacts can be preferentially assigned to 0.875 agents. BP strategy 210 aims for balanced utilization, where each agent receives approximately the same proportion of contacts over time. Therefore, there is no bias towards the PQ strategy, where contact utilization would be skewed towards more heavily utilizing the highest-performing 75-100% of contacts.

One such technique for generating performance-biased contact type percentages according to embodiments of the present disclosure is to adjust the "initial" midpoint percentage ("CP _initial ") of each contact type via a blending function or factor such that relatively higher-ranked (e.g., higher-performing) contacts occupy relatively more bandwidth and, therefore, receive relatively more contacts than lower-ranked (e.g., lower-performing) contacts. For example, the blending function may raise the percentage of each contact by a power, as shown in Equation 1 below:

CP _adjusted = CP _initial ^ρ (Equation 1)

The power parameter (e.g., "ρ" or "Rho parameter" as in Equation 1) can determine the amount of bias toward the PQ, with higher values of Rho generating a greater amount of bias. A Rho parameter of 1.0 would generate no bias (CP _adjusted = CP _initial ). Thus, this "neutral" value for Rho results in a goal of balancing contact utilization. In effect, the BP policy 210 is equivalent to a Rho-based HBP policy with Rho equal to 1.0. As Rho increases, the degree of contact utilization skew increases as the bias toward the PQ increases.

FIG3 illustrates a queue 300 to which this technique is applied using a Rho value of 2.0. Queue 300 represents the same four types of contacts and the same four agents as in queue 200. However, in queue 300, the midpoint of the percentage of the contact type has been squared (CP _adjusted = CP _initial ^2.0 ). Applying a diagonal strategy with the HBP strategy 310, the lowest ranked contact type (CP _adjusted ≈ 0.016) will occupy the least bandwidth and be selected least frequently, and so on, up to the highest ranked contact type (CP _adjusted ≈ 0.766), which will occupy the most bandwidth and be selected most frequently.

In some embodiments, the bandwidth for each contact type can be determined such that the midpoint of the adjusted percentage for each contact type is the midpoint of the newly adjusted bandwidth for each contact type. For example, the bandwidth for the lowest ranked 0.016 contact type can be approximately 0.000 to 0.031. In other embodiments, the bandwidth for each agent can be determined by equally distributing the "distance" between adjacent adjusted percentage midpoints. For example, the bandwidth for the lowest ranked 0.016 contact type can be approximately 0.000 to 0.079.

Another version of the HBP technique applied to the queue 300 of FIG. 3 is to adjust the initial percentage range for each contact type rather than the initial midpoint percentage for each contact type, as shown in Equation 2 below:

CP _{adjusted_range} =CP _{initial_range} ^ρ (Equation 2)

The effect will be the same: relatively higher ranked (e.g., higher valued) contact types occupy relatively more bandwidth and are therefore selected relatively more frequently than lower ranked (e.g., lower valued) contact types.

FIG4 illustrates a queue 400 to which this technique is applied using a Rho value of 2.0. Queue 400 represents the same four contact types and agents as in queue 300. However, in queue 400, the initial percentage ranges of the contact types have been squared (CP _{adjusted_range} = CP _{initial_range} ^2.0 ) rather than their initial midpoint percentages. Applying the diagonal strategy under HBP strategy 410, the lowest-ranked contact type (occupying an adjusted percentage range from 0.00 to approximately 0.06, with a midpoint of approximately 0.03) will be selected least frequently, and so on until the highest-ranked contact type (occupying an adjusted percentage range from approximately 0.56 to 1.00, with a midpoint of approximately 0.82) is selected most frequently.

Conceptually, exploitation of target skew will result in the highest ranked contact type being selected slightly less than half the time typically when one of the top half of agents becomes available, and lower ranked contact types being selected slightly more than half the time typically when one of the bottom half of agents becomes available. Other techniques for visualizing or implementing these blending functions or factors include adjusting a "fit function" for the diagonal strategy.

FIG5 shows a queue 500 having the same percentage and range of contact types as queue 100 ( FIG1 ), and which has not yet been adjusted. Unlike queue 110, in which BP strategy 110 can be visualized by a 45-degree diagonal line (CP=AP), HBP strategy 510 can be visualized by a different, hybrid fitting function (e.g., "bending" or "curving" the diagonal line). In the example of FIG5 , the fitting function is an exponential function, as shown in the following equation 3:

AP = CP ^ρ (Equation 3)

Conceptually, instead of determining the preferred pairing by selecting the pair closest to the diagonal AP=CP as in BP strategy 110, the preferred pairing in HBP strategy 510 can be determined by selecting the pair closest to the exponent AP=CP ^2.0 , as in cohort 500, where Rho equals 2.0. Notably, the effect of fitting to AP=CP ^2.0 is a continuous mathematical simulation of a discontinuous process of widening or shrinking the percentage range (e.g., squaring the percentage range and then fitting to AP=CP, as in cohort 400 and HBP strategy 410 ( FIG. 4 )).

Many variations of blending functions can be used to vary the target utilization of an agent based on its performance or other ranking or metric. For example, the blending function can be a piecewise function.

FIG6 illustrates a queue 600 and HBP policy 610 that impacts utilization of the bottom half of contact types differently than utilization of the top half of contact types. For example, a contact center may determine that half of the contacts should be distributed to below-average agents in a balanced manner (e.g., Rho = 1.0), but the other half should be distributed to above-average agents (e.g., Rho > 1.0) based on the relative ranking of each contact type. Thus, contacts ranging from 0% to 50% can be evenly distributed to lower-performing agents (0.125 agents and 0.375 agents), which can be visualized as a fit along the 45-degree line AP = CP, where 0.00 ≤ CP < 0.50 (or, for example, 0.00 < CP ≤ 0.50, etc.). Contacts ranging from 50% to 100% can be distributed to higher-performing agents (0.625 agents and 0.875 agents) as a function of their relative ranking (e.g., value) of the contact type (e.g., an exponential function scaled to this portion of the contacts and agents). The HBP strategy 610 can be visualized as a fit along an exponential curve AP = 2(CP - 0.5) ^2.0 + 0.5, where Rho = 2.0 and 0.50 ≤ CP < 1.00.

Occasionally, such a strategy will result in some higher-ranked contact types (here, the 0.50 to 0.75 contact types) being selected less frequently over time than their lower-ranked counterparts. FIG7 shows a queue 700 and HBP strategy 710 that also uses a piecewise blending function to affect the utilization of the lower half of contacts differently from the utilization of the upper half of contacts. For example, a contact center may determine, based on their relative ranking, that a larger portion of contacts should be distributed to agents above average (e.g., Rho > 1.0), and that the remainder of contacts should be distributed in a balanced manner to agents below average (e.g., Rho = 1.0). Thus, for Rho = 2.0 and CP ≥ 0.50 (or CP > 0.50), the pairing can be fitted along an exponential curve: AP = CP ^2.0 . For Rho = 1.0 and CP < 0.50, the pairing can be fitted along a linear function scaled to this portion of contacts and agents: AP = 0.5·CP.

In a real-world contact center, there may be more or fewer agents, as well as more or fewer contact types in the queue. In these examples, each contact type is evenly distributed across the full range of percentile rankings; however, in some contact centers, the distribution of the range may be based on, for example, the frequency with which a particular type of contact arrives at the contact center relative to the frequency with which other types of contacts arrive. The simplified example described above with four agents and four contact types is used to illustrate the effects of implicit forms of HBP, such as those based on the Rho parameter and exponential scaling or other blending functions. However, HBP—including Rho-based techniques—can also be applied to larger, more complex real-world contact centers.

In some embodiments, Rho can be selected or adjusted to vary the bias toward PQ (or the skew in contact utilization). For example, a Rho less than 2.0 (e.g., 1.0, 1.01, 1.1, 1.2, 1.5, etc.) will result in a relatively lower bias toward PQ compared to the above example of Rho equal to 2.0. For example, if a contact center administrator wants to avoid the situation where higher-ranked contact types are selected multiple times while lower-ranked contact types remain unselected, a significantly lower value of Rho than 2.0 may be more appropriate. Conversely, a Rho greater than 2.0 (e.g., 2.01, 2.1, 2.5, 200.0, etc.) will result in a relatively greater bias toward PQ.

Importantly, the impact on contact utilization is subtle under Rho-based HBP policies because they controllably influence the extent to which contacts of differently ranked contact types wait to connect to an agent. By increasing the power by which the contact percentage is boosted, the present invention controllably decreases the average time between selections for higher-ranked contact types and increases the average time between selections for relatively lower-ranked contact types. Similarly, decreasing the power by which the contact percentage is boosted has the opposite effect. For a neutral BP policy (e.g., Rho = 1.0), each agent has approximately the same expected average wait time between contacts. As Rho increases, the relative expected average wait time decreases gradually (e.g., exponentially) as the relative contact type ranking (e.g., contact type value) increases.

In some embodiments, the HBP policy can use a potentially more gradual technique to target relative contact utilization. For example, contact types can be assigned relative "utilization adjustments" based on the contact type ranking. In one example, the highest-ranked contact type can be assigned a relative utilization adjustment of 100%, the second-highest-ranked contact type can be assigned a relative utilization adjustment of 99%, the third-highest-ranked contact type can be assigned a relative utilization adjustment of 98%, and so on. In this example, the target utilization of the second-highest-ranked contact type would be 99% of the target utilization of the highest-ranked contact type. Relative utilization adjustments can be more aggressive in other configurations. For example, the highest-ranked contact type can be assigned a relative utilization adjustment of 100%, the second-highest-ranked contact type can be assigned a relative utilization adjustment of 90%, the third-highest-ranked contact type can be assigned a relative utilization adjustment of 80%, and so on. In this example, the target utilization of the second-highest-ranked contact type would be 90% of the target utilization of the highest-ranked contact type.

8 illustrates a hybrid behavior pairing method 800 according to an embodiment of the present disclosure. At block 810, the hybrid behavior pairing method 800 may begin.

At block 810, a percentage (or n-slice, quantile, percentage range, bandwidth, or other type of "score" or range of scores, etc.) can be determined for each available contact. For contacts currently waiting in a queue, a percentage can be determined for each of the contacts currently waiting in the queue. For contacts not currently waiting in a queue, a percentage can be assigned to the next contact to reach the contact center. Based on information about the contacts, the percentage can be bounded by a range of percentages defined for a particular type or group of contacts. The percentage bounds or ranges can be based on a frequency distribution or other metric for the contact type. Percentages can be randomly assigned within the range of percentages for the type.

In some embodiments, the percentages may be ranked according to a particular metric or combination of metrics to be optimized in the contact center, and contacts determined to have relatively high percentages may be considered "higher value" contacts for the contact center because these contacts are more likely to contribute to higher overall performance in the contact center. For example, a relatively high percentage of contacts may have a relatively high likelihood of making a purchase.

In some embodiments, the percentage can be determined for a contact at the time the contact arrives at the contact center. In other embodiments, the percentage can be determined for a contact at a later point in time, such as when the contact arrives at a specific skill queue or ACD system, or when a pairing request is made.

After the percentage has been determined for each contact available for pairing, the behavioral pairing method 800 may proceed to block 820. In some embodiments, block 820 may be performed before or concurrently with block 810.

At block 820, a percentage can be determined for each available agent. For situations where agents are idle, waiting for a contact to arrive, a percentage can be determined for each of the idle agents. For situations where all agents in a queue are busy, the percentage can be determined for the next agent to become available. The percentage can be defined by a range of percentages (e.g., "bandwidth") defined based on all agents assigned to a queue (e.g., a skill queue) or only the available agents assigned to a particular queue. In some embodiments, the bounds or ranges of the percentages can be based on desired agent utilization (e.g., for fairness, efficiency, or performance).

In some embodiments, agent percentages can be ranked according to a particular metric or combination of metrics to be optimized in the contact center, and agents determined to have a relatively high percentage can be considered higher performing agents of the contact center. For example, agents with a relatively high percentage can have a relatively high likelihood of making a sale.

In some embodiments, the percentage of agents may be determined at the time that the agents become available within the contact center. In other embodiments, the percentage may be determined at a later point in time, such as when a request for pairing is made.

After the percentages have been determined for each available agent and contact, the behavior pairing method 800 may continue to block 830 .

At block 830, a blending function may be applied to the contact type percentages (or contact type percentage ranges or bandwidths). For example, a Rho value may be determined for an exponential blending function or a fitted curve or line. In some embodiments, the blending function may operate on a single ranking that implicitly incorporates both behavioral fit and performance information. In other embodiments, the blending function may combine (e.g., add, multiply, weight) multiple rankings of contact types. After the blending function has been applied or otherwise determined or configured, the blended-behavior pairing method 800 may proceed to block 840.

At block 840, based on the blending function, a pair of available contacts and available agents can be determined based on the percentages (or percentage ranges) determined for each available contact at block 810 and for each available agent at block 820. In some embodiments, the selection can be determined based on the percentages or percentage ranges adjusted for each waiting contact or contact type at block 830. In some embodiments, the pair can be determined based on a diagonal strategy, where contacts and agents with more similar percentages (or the most similar percentages) are selected for pairing. For example, the blended behavior pairing module can select the contact-agent pairing with the smallest absolute difference between the contact's score and the agent's score. In some embodiments, the diagonal strategy can be visualized as a 45-degree diagonal line. In other embodiments, the diagonal strategy can be visualized as a blending function (e.g., an exponential function or a piecewise function).

In some cases, multiple agents may be idle when a contact arrives (L1 state). In the case of HBP, the newly available contact may be paired with a selected one of the available agents that has a percentage or range of percentages that is more similar to the contact's adjusted percentage than the other available agents. In other cases, multiple contacts may be waiting in the queue when an agent becomes available (L2 state). In the case of HBP, the newly available agent may be paired with a selected one of the contacts waiting in the queue that has a percentage or range of percentages that is more similar to the agent's adjusted percentage than the other contacts waiting in the queue.

In some cases, selecting pairings based on similarity of scores may result in selecting an instantaneous pairing that may not be the highest performing instantaneous pairing, but instead increases the likelihood of better future pairings.

After a pairing has been determined at block 840, the hybrid behavior pairing method 800 can proceed to block 850. At block 850, a module within the contact center system can connect the contacts and agents of the contact-agent pair to each other. For example, the behavior pairing module can instruct an ACD system or other routing device to distribute a particular contact to a particular agent.

After the contacts and agents are connected at block 850, the behavior pairing method 800 may end. In some embodiments, the behavior pairing method 800 may return to block 840 to determine one or more additional pairs (not shown). In other embodiments, the behavior pairing method 800 may return to block 810 or block 820 to determine (or redetermine) a percentage or range of percentages for available contacts or agents (not shown), and then apply (or reapply) the blending function at block 840.

At this point, it should be noted that hybrid behavior pairing in a contact center system according to the present disclosure, as described above, may involve, to some extent, the processing of input data and the generation of output data. This input data processing and output data generation may be implemented in hardware or software. For example, specific electronic components may be employed in a behavior pairing module or similar or related circuitry as described above for implementing the functionality associated with behavior pairing in a contact center system according to the present disclosure. Alternatively, one or more processors operating according to instructions may implement the functionality associated with behavior pairing in a contact center system according to the present disclosure, as described above. If this is the case, it is within the scope of the present disclosure that such instructions may be stored on one or more non-transitory processor-readable storage media (e.g., disks or other storage media) or transmitted to one or more processors via one or more signals implemented in one or more carrier waves.

The present disclosure will not be limited in scope by the specific embodiments described herein. In fact, various other embodiments and modifications of the present disclosure, in addition to those described herein, will be apparent to those of ordinary skill in the art from the foregoing description and accompanying drawings. Therefore, such other embodiments and modifications are intended to fall within the scope of the present disclosure. Further, although the present disclosure has been described herein in the context of at least one specific embodiment in at least one specific environment for at least one specific purpose, it will be appreciated by those of ordinary skill in the art that its usefulness is not limited thereto and that the present disclosure can be beneficially implemented in any number of environments for any number of purposes. Therefore, the claims set forth below should be interpreted in view of the full breadth and spirit of the present disclosure as described herein.

Claims (15)

1. A method for pairing in a contact center system, the contact center system comprising at least one switch communicatively coupled to a plurality of agents, the method comprising: determining, by at least one processor, a percentage of agents in the plurality of agents; determining, by the at least one processor, a percentage for each of a plurality of contacts; applying, by at least one processor, a blending function to each of the plurality of percentages of contacts to adjust the percentage of each of the plurality of contacts based on a behavioral pairing policy for pairing and a priority queuing policy for pairing, and wherein the blending function biases the behavioral pairing policy for pairing toward the priority queuing policy for pairing and controllably enables, by the at least one processor, targeting of unbalanced contact utilization; comparing, by the at least one processor, a first difference in percentages between the percentage of the agent in a first pair and the adjusted percentage of a first contact with a second difference in percentages between the percentage of the agent in a second pair and the adjusted percentage of a second contact different from the first contact; and The first pair or the second pair is selected, by the at least one processor, for connection based on the comparison. 2 . The method of claim 1 , wherein selecting the first pair or the second pair based on the comparison further comprises applying, by the at least one processor, a diagonal strategy to the percentages such that a pair with a more similar percentage is selected. The method of claim 1 , wherein determining the percentage of the agent or the percentage of each of the plurality of contacts is expressed as a percentage range. The method of claim 3 , wherein each of the plurality of contacts is assigned a percentage within a corresponding percentage range for each contact. The method of claim 4 , wherein the allocated percentage is the midpoint of a range of percentages. The method of claim 4 , wherein the assigned percentage is a random percentage of a range of percentages. 7. The method of claim 1, further comprising determining, by the at least one processor, a percentage range for the agent to be proportional to the relative performance of the agent. 8. The method of claim 1 , wherein the blending function comprises determining, by the at least one processor, a percentage range for each successively higher percentage of a plurality of contact types, the percentage range being greater than the percentage range for a corresponding lower percentage of the plurality of contact types. 9. The method of claim 1 , wherein the selected contact or contact type of the selected pair is not any of the following: The contact or contact type that is lagged in the fairness metric, The contacts or contact types that are rated highest in value, priority, or performance metrics, the contacts or contact types that are rated highest in value, priority, or performance metric for a particular agent, or Contacts of the agent previously assigned to the selected pair. 10. The method of claim 1, wherein a higher percentage of agents remain available for subsequent assignment to a similarly higher percentage of contacts, or a higher percentage of contacts remain available for subsequent assignment to a similarly higher percentage of agents. The method of claim 1 , wherein each successively higher percentage of the plurality of contacts is more likely to be selected than a correspondingly lower percentage of the contacts. 12. The method of claim 1, wherein each successively higher percentage of the plurality of contacts comprises a lower average wait time than a corresponding lower percentage of the contacts. 13. A method for pairing in a contact center system, the contact center system comprising at least one switch communicatively coupled to a plurality of agents, the method comprising: determining, by at least one processor, a percentage of agents in the plurality of agents; determining, by the at least one processor, a first percentage of the plurality of contacts according to a behavioral pairing strategy for pairing; determining, by the at least one processor, a second percentage of the plurality of contacts according to a priority queuing strategy for pairing; applying, by the at least one processor, a blending function to each of the plurality of percentages of contacts, adjusting the first percentage according to the second percentage, and wherein the blending function controllably enables, by the at least one processor, targeting unbalanced contact utilization; comparing, by the at least one processor, a first difference in percentages between the percentage of the agent in a first pair and the adjusted percentage of a first contact with a second difference in percentages between the percentage of the agent in a second pair and the adjusted percentage of a second contact different from the first contact; and The first pair or the second pair is selected, by the at least one processor, for connection based on the comparison. 14. A system for pairing in a contact center system, comprising: at least one switch communicatively coupled to the plurality of agents; at least one processor communicatively coupled to the contact center system and configured to operate in the contact center system, wherein the at least one processor is further configured to: determining a percentage of agents among the plurality of agents; determining a percentage for each of a plurality of contacts; applying a blending function to each of the plurality of percentages of contacts to adjust the percentage of each of the plurality of contacts based on a behavioral pairing policy for pairing and a priority queuing policy for pairing, and wherein the blending function biases the behavioral pairing policy for pairing toward the priority queuing policy for pairing and is controllably enabled, by the at least one computer processor, to target unbalanced contact utilization; comparing a first difference in percentages between the percentage of the agent in a first pair and the adjusted percentage of a first contact to a second difference in percentages between the percentage of the agent in a second pair and the adjusted percentage of a second contact different from the first contact; and The first pair or the second pair is selected for connection based on the comparison. 15. The system of claim 14, wherein the at least one processor is further configured to determine a percentage range for each successively higher percentage of the plurality of contact types, the percentage range being greater than the percentage range for a corresponding lower percentage of the plurality of contact types.