Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q10/00—Administration; Management
  • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q10/00—Administration; Management
  • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
  • G06Q10/063—Operations research, analysis or management
  • G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
  • G06Q10/06375—Prediction of business process outcome or impact based on a proposed change
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q10/00—Administration; Management
  • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
  • G06Q10/067—Enterprise or organisation modelling

Definitions

• the present invention generally relates to telecommunications systems and methods, as well as contact center behavior. More particularly, an embodiment pertains to predicting the behavior of a contact center.
• a system and method are presented for predicting contact center behavior.
• closed form simulation modeling may be used to simulate behavior from input distributions. Models may be created through staging and analysis of historical Automatic Call Distribution data. Service level, average speed of answer, abandon rate, and other data may be predicted to generate forecasts and analysis of contact center behavior. Examples of behavior may include staffing levels, workload, and the Key Performance Index of metrics such as service level percentage, average speed of answer, and
• a method for calculating the predicted service performance of a contact center comprising the steps of: performing analysis and staging of the historical Automatic Call Distribution data for the contact center; building a model from the analysis and staging of historical Automatic Call Distribution data; validating the model; and using the validated model to predict the behavior of the contact
• a method for predicting contact center queuing and customer patience behavior in order to calculate service performance in a contact center comprising the steps of: performing analysis and staging of the historical Automatic Call Distribution data for the contact center, wherein the historical Automatic Call Distribution data comprises one or more of: contact volume, average handle time, full time equivalency (FTE), capture rate, contact handling data for contact types, and contact handling data for staffing types, for a specified interval; building a simulation model from the analysis and staging of the historical Automatic Call Distribution data, wherein from the staged Automatic Call Distribution data, a Key Performance Index is extracted at an interval level equivalent to the extracted historical Automatic Call Distribution data; validating the simulation model using the extracted historical Automatic Call distribution data; and using the validated simulation model to predict the contact center queuing and customer patience behavior.
• the historical Automatic Call Distribution data comprises one or more of: contact volume, average handle time, full time equivalency (FTE), capture rate, contact handling data for contact types, and contact handling data for staffing types, for a specified interval
• building a simulation model from the
• Figure 1 is a diagram illustrating the basic components of an embodiment of a system.
• Figure 2 is a flowchart illustrating an embodiment of a process for model creation.
• Figure 3 is a flowchart illustrating an embodiment of a process for prediction.
• Figure 4 is a diagram illustrating a sample abandon rate distribution profile.
• Figure 5 is a diagram illustrating a sample triangular distribution model.
• Models are traditionally used by analysts in the contact center industry to predict, plan and analyze contact center behavior. Examples may include time series forecasting-based models, regression- based models, queuing theory-based equations, and custom discrete event simulation modeling. These models are often computationally slow, difficult to use and limited in capability and applicability.
• Erlang equations are widely known and used in queuing theory in contact center queues. These equations may be used to estimate workforce demand given average handle time and volume offered. However, due to the simplifying assumptions, accuracy generally lacks in Erlang equations. Erlang equations assume that no communications are abandoned, which may mean, for example, that every customer stays on the line ad infinitum until their communication is handled. Erlang equations also assume that every communication is similar, which ignores the fact that contact center routings and/or strategies can change drastically over time.
• Discrete event simulation modeling is generally thought to be the most accurate and precise in its ability to predict contact center behavior due to its flexibility. It can be created/customized to any specific routings. A key factor that contributes to its accuracy is the ability to incorporate an accurate model of a customer patience profile which compares the abandon rate to the handling time.
• discrete event simulation modeling is very complex and also has a slow computational time. In order to decrease the computational time, simulations may be pre-run and the results stored in a database to be used in the workforce management/analytic application. Every time the underlying system behavior changes, the simulation models need to be rebuilt and pre-run again and again. Thus, there is a need for a way to easily, quickly and accurately predict contact center behavior.
• Figure 1 is a diagram illustrating the basic components of an embodiment of a system.
• the basic components of the system 100 may include: Inputs 105 which may include Contact Volume 1 10, Average Handle Time 1 15, and Available Resources 120; Continuous Simulation Engine 125; and Outputs 130 which may include Average Speed of Answer 135, Service Level 140, and Abandon Rate 145.
• the Inputs 105 may comprise information that is fed into the Continuous Simulation Engine 125. Inputs may be at an interval level, such that an interval in which the model is built on is specified.
• Intervals may comprise 30 minutes, 1 hour, daily, weekly, etc.
• Inputs 105 may be given for a contact type, such as a telephone call, for example, and may comprise Contact Volume 1 10, Average Handle Time 1 15, and Available Resources 120.
• the Contact Volume 1 10 may comprise the amount of interactions in the queue.
• a contact may comprise any kind of inbound communication, interaction, contact, or function requiring processing or response by the receiving entity.
• An entity may comprise an inbound contact center or a group of inbound contact centers that share incoming contact volume, where varying volumes of inbound contacts, such as telephone calls, e-mails, web 'chat', or traditional paper mail are processed with the goal of achieving desired service quality metrics.
• the Average Handle Time 1 15 may comprise the average amount of time a contact is handled by an agent across a pool of Available Resources 120.
• the Available Resources 120 may comprise the pool of some number of agents and/or servers that are available and able to service the contact from the queue.
• the Continuous Simulation Engine 125 comprises closed form simulation modeling.
• the Continuous Simulation Engine 125 may be an engine or a modular library (such as an API) that can be easily integrated with any existing workforce management and long term planning system to allow interaction-specific prediction of Outputs 130.
• the Continuous Simulation Engine 125 may function without the need to pre-run a simulation and store, or cache, the simulation results as discrete event simulation does.
• the Continuous Simulation Engine 125 does not require the assumption of no abandonments in forming the predictions, which is unlike Erlang-B or Erlang-C equations. By assuming that no customers who contact the contact center hang up if forced to wait an excessive amount of time for an agent, gross inaccuracies in forecast performance may result.
• the Continuous Simulation Engine 125 may work in two phases, as further described below in Figures 2 and 3.
• the Outputs 130 may be used for generating forecasts and analysis of contact processing center behavior for planning purposes over short, medium, and long-term planning horizons. Outputs may also be at an interval level, such that an interval in which the model is built on is specified. Intervals may comprise 30 minutes, 1 hour, daily, weekly, etc.
• the Outputs 130 from the Continuous Simulation Engine 125 may comprise an Average Speed of Answer 135, a Service Level 140, and an Abandon Rate 145. While the examples of Average Speed of Answer 135, Service Level 140, and Abandon Rate 145 are illustrated, it is within the scope of this disclosure to include other performance metrics of contact center behavior such as staffing levels, workloads, etc., to name a few.
• performance metrics may include staff occupancy/utilization, capture rate (how many of a contact is being handled by a staff type as compared to an other staff type), purity rate (how much of a staff type's time is allocating on handling a contact compared to an other contact), expected wait time until abandon, probability of waiting in a queue for some amount of time, and the queue length at any given time.
• Inputs 105 do not necessarily need to be at the same interval level as the desired output. For example, an input may be at a daily level, but the desired output is at the weekly level.
• Figure 2 is a flowchart illustrating an embodiment of a process 200 for model creation.
• the process 200 may be operative in the Continuous Simulation Engine 125 in the system 100 ( Figure 1).
• types are defined and classified.
• ACD Automatic Call Distribution
• Staff types may also be defined for planning purposes.
• Staff types may be defined automatically or by user instructions.
• a staffing type may comprise a skill set possessed by one or more agents who utilize the skill set to handle certain types of communications or interactions. Agents may belong to more than one groups determined by skill set, such as an agent who is bilingual, for example. Agents may also be classified based on their role within the workgroup, such as supervisor, representative, etc. Control is passed to operation 210 and process 200 continues.
• routing behavior is identified. For example, information may be obtained from the historical ACD data to determine how contact and/or interactions may be routed (i.e., which agents handle which interactions). Information about contact types, staff types, interaction priorities, agent utilizations, available staffing ratios and the expected range of interactions that each staff types will have may also be obtained from the historical ACD data. Control is passed to operation 215 and process 200 continues.
• models are built. For example, the process will iterate over the identified contact and/or interaction types and build models for each type.
• the customer patience profile of the contact is evaluated. This may be done, for example, by plotting how many interactions are abandoned after customers have been waiting for some amount of time, represented by 'x' below.
• a successful customer patience profiled has been identified, iteration over a set of model parameters is performed and then validated against historical ACD data for that type. Control is passed to operation 220 and process 200 continues.
• model parameters are estimated.
• the model parameters may be defined to contain the interval, a Customer Service Representative Factor (CSRF), a Service Level Threshold, a Delay, and the Patience curve profile.
• the interval width may indicate an amount of time, such as the number of minutes per planning interval. This may be 15 minutes, 30, or 60.
• the CSRF represents the efficiency factor of the agent groups, such as 0.9, or 90% efficiency.
• the Service Level Threshold may be a limit set, such as 20 seconds, for which a communication should be handled to meet the service level goal.
• Delay may comprise an artificial delay that is introduced to every communication before it can be serviced by an agent.
• control is passed to operation 225 and process 200 continues. [25] In operation 225, it is determined whether or not the model validates. If it is determined that the model validates, control is passed to operation 230 and process 200 continues. If it is determined that the model does not validate, control is passed back to operation 220 and process 200 continues.
• the determination in operation 225 may be made based on any suitable criteria.
• the Continuous Simulation Engine 125 may perform the validation through a mathematical process. If a good fit is not found, the process starts again from operation 220 until a suitable model that validates against historical ACD data is found.
• model and parameters are saved. For example, all the model parameters along with the customer patience profile may be saved in a central repository or a text based file. Control is passed to operation 235 and process 200 continues.
• operation 235 it is determined whether or not all models have been built. If it is determined that all models have been built, control is passed to operation 240 and the process continues. If it is determined that not all models have been built, control is passed back to operation 215 and process 200 continues.
• the determination in operation 235 may be made based on any suitable criteria.
• each model is associated with a specific contact type.
• a model may be complete when all of its parameters (such as service level threshold, CSRF, delay, and patience curve profile, for example) have been estimated and, applying these parameters to the continuous simulation process, validated against historical ACD data. This may be repeated for as many contact types that are defined in the routing profile in operation 210.
• Figure 3 is a flowchart illustrating an embodiment of a process for prediction.
• the process 300 may be operative in the Continuous Simulation Engine 125 in the system 100 ( Figure 1).
• operation 305 information is input to the model for the different types. For example, contact center inputs in their original form (e.g., weekly or monthly staff, average handle times and volume) are distributed to the same intervals in which the simulation model was built for in process 200. Information is also considered pertaining to how contacts, or interactions, are routed or planned to be routed as well as what goals to achieve, agent prioritization and expected agent utilization. Control is passed to operation 310 and process 300 continues.
• contact center inputs in their original form e.g., weekly or monthly staff, average handle times and volume
• Information is also considered pertaining to how contacts, or interactions, are routed or planned to be routed as well as what goals to achieve, agent prioritization and expected agent utilization. Control is passed to operation 310 and process 300 continues.
• the behaviors are simulated for each contact/interaction type. For example, for each interaction type, contact center behavior is simulated and the marginal increase in Key Performance Indicators (KPI) when the input changes are evaluated.
• KPI Key Performance Indicators
• the simulation may be performed mathematically.
• the service level may be calculated mathematically as follows:
• ⁇ represents the Acceptable Wait Time, which may be defined with the Service Level Goal.
• SL represents the Service Level and is expressed as a percentage.
• ⁇ represents the mean processing rate of contacts, such as calls, for example, per unit of time or the call arrival rate
• s represents the number of agents or servers
• J represents the function of time variable and is described in more detail below
• ⁇ represents the call arrival rate
• ⁇ which represents the inverse of the probability of blocking
• the probability of blocking may refer to the probability of a new contact arriving and being rejected by a group of identical parallel resources that are all currently busy, such as telephone lines, circuits, traffic channels, etc., for example, u may be calculated as:
• u ⁇ ( ⁇ ) - ⁇ [42]
• ⁇ ( ⁇ ) H(x)
• ⁇ ⁇ hazard function for random variable X given ⁇ .
• Random variable X represents the average wait time (i.e., patience time, in seconds) before a caller abandons the queue. Solving for u is performed by:
• F(x) represents the cumulative distribution function of random variable X.
• the F(x) function is derived from the abandon distribution profile mined from real data.
• two probability density zones can be defined as illustrated in Figure 4.
• Figure 4 illustrates an embodiment of the abandon rate distribution profile with the Abandon Rate percentage, 401, compared with the Average Wait Time (expressed in seconds, in this example), 402.
• Zone 1, 405 is the step-function probability distribution, which comprises several uniformly distributed probability distribution functions.
• Zone 2, 410 is the triangular-distribution probability distribution, which comprises one triangular distribution with minimum, median, and maximum parameters. This triangular distribution governs the behavior of the customer who is waiting to wait in the queue for a certain amount of time before abandonment.
• H(x) represents the hazard function of random variables x through the passage of wait-time in queue. This may further be defined as the sum of the hazard function of Zone 1, 405, and Zone 2, 410.
• the generic definition of the Hazard function which is:
• F(x) is the cumulative distribution function of the random variables x
• Zone 1 405.
• a wait-time random variable x lies, bounded by the region of lower bound value of k w and upper bound value of k ub (i.e., the ad-hoc bucket of k w ⁇ x ⁇ k ub )
• its hazard function is derived as:
• Tzone l represents the total probability of the Zone 1, 405, area and the value may be between 0 and 1. Since k lb of the ad-hoc bucket is not necessarily equal to zero, and there can be many buckets preceding the ad-hoc bucket, the hazard function each of the i buckets is:
• H(x) H(ki), for all buckets that preceded the k lb ⁇ x ⁇ k ub bucket.
• H(ki) Jzone 1 — F(kw of i ) * (kub of i ⁇ k lb of i) + ( ⁇ ku/of i - fei6 of i ) ⁇ w °f l )i ⁇ ub °f i ⁇ °f _
• Hazard functions are determined sequentially for each bucket using the lower and upper bound of bucket, in the definite integral, and the values summed for all buckets until the bucket with the lower bound value equal to the lower bound value of the ad-hoc bucket is met.
• the final value of the summation may be represented by 'sumofHazzardPrecedmg'.
• the hazard function of the random variable x may be formulated and the final result will be a mathematical equation in terms of x.
• Zone 2 410
• Figure 5 represents an example of the solution determined from the following:
• H(x) H(x)of interval [0, a] + [(x - a) - ( 3 ⁇ b ⁇ ⁇ c _ a) )] * (1 - T Zone .
• H(x) H(x)of interval [0, a] + H(x)of interval [a, c] + ⁇ b ⁇ ) * (1 — T Zone -L)
• H(x) H(x)of interval [0, a] + H(x)of interval [a, c] + H(x)of interval [c, b]
• the abandon rate may be calculated mathematically as follows as a percentage:
• ASA Average Speed of Answer
• a new model is formed.
• the new model may combine all of the contact/interaction types using linear programming or heuristic where the KPI metrics are calculated, as well as how interactions flow and how agent are utilized. Control is passed to operation 320 and process 300 continues.
• KPI metrics such as service level, average speed of answer, abandon rate, and occupancy, to name a few, may be viewed. Staff and volume allocations may also be viewed.

Landscapes

• Business, Economics & Management (AREA)
• Human Resources & Organizations (AREA)
• Engineering & Computer Science (AREA)
• Strategic Management (AREA)
• Economics (AREA)
• Entrepreneurship & Innovation (AREA)
• Operations Research (AREA)
• Physics & Mathematics (AREA)
• Development Economics (AREA)
• Marketing (AREA)
• Educational Administration (AREA)
• Quality & Reliability (AREA)
• Tourism & Hospitality (AREA)
• Game Theory and Decision Science (AREA)
• General Business, Economics & Management (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Telephonic Communication Services (AREA)
• Management, Administration, Business Operations System, And Electronic Commerce (AREA)
• Testing And Monitoring For Control Systems (AREA)

Priority Applications (1)

Applications Claiming Priority (1)

Related Child Applications (1)

Publications (2)

Family

ID=54196120

Family Applications (2)

Family Applications After (1)

Country Status (7)

Families Citing this family (2)

Family Cites Families (13)

• 2014
  • 2014-03-25 EP EP14887258.3A patent/EP3123413A4/de not_active Withdrawn
  • 2014-03-25 BR BR112016021995A patent/BR112016021995A2/pt not_active Application Discontinuation
  • 2014-03-25 CA CA2943160A patent/CA2943160C/en active Active
  • 2014-03-25 WO PCT/US2014/031685 patent/WO2015147798A1/en active Application Filing
  • 2014-03-25 AU AU2014388386A patent/AU2014388386A1/en not_active Abandoned
  • 2014-03-25 JP JP2016558661A patent/JP6495938B2/ja active Active
  • 2014-03-25 EP EP19210030.3A patent/EP3629260A3/de not_active Withdrawn
• 2016
  • 2016-09-19 ZA ZA2016/06452A patent/ZA201606452B/en unknown

Also Published As

Similar Documents

Legal Events