WO2006122057A2 - Managing customer service staffing loads in the context of in-person transactions - Google Patents

Abstract

A customer service facility such as a bank branch is instrumented to measure customer loading as of the time of customer arrival, versus the time of delivery of services. Customer satisfaction is assessed from customer waiting time and queue length in a preliminary test environment using relatively extensive instrumentation to record data on customer arrival, departure, location between arrival and departure, queue wait time, etc. The results are analyzed and correlated to data collected from only a subset of the instrumentation, such as a limited set of customer proximity sensors at selected locations. An actual operation is managed based on inputs from the subset of instrumentation, relying on the analysis and correlation from the test, for inferring customer satisfaction at the same facility (using fewer sensors) or at a similarly configured operational facility.

Description

METHOD AND APPARATUS FOR

MANAGING CUSTOMER SERVICE STAFFING LOADS

IN THE CONTEXT OF IN-PERSON TRANSACTIONS

Cross Reference To Related Application

[0001] This application claims the priority of U.S. Provisional Patent

Application Serial No. 60/678,311 , filed May 5, 2005.

Field of the Invention

[0002] The invention relates to techniques for assessing, anticipating and responding to changing degrees of customer demand for services in situations where a relatively larger number of customers can be queued for delivery of services involving transactions executed by a smaller number of service representatives.

Background of the Invention

[0003] In various situations involving customer transactions, sales and other operations, for example delivered by businesses or governmental operations, a number of service representatives stand ready to interact with customers, one at a time. From time to time, the balance between the number of service representatives and the number of customers will vary. It is highly desirable to deliver services to customers promptly, for reasons of customer satisfaction. It is also desirable for reasons of efficiency to manage the size of the workforce.

[0004] A pertinent example is a banking branch facility. A banking branch typically has a demand determined by various factors that may include, without limitation, its location, the time of day, the day or date (for example as compared to customer wage and bill payment cycles) and even the current weather conditions. The branch has a finite capacity to execute customer transactions that is related to the number of teller stations and/or automatic teller machines (ATMs), the number of tellers currently engaged in actively servicing customers, the complexity of the transactions and perhaps whether management interaction is needed (in which case the availability of managers may be pertinent) and similar factors. Some of these constraints are not adaptable, but others such as the number of tellers currently at work, can be varied over time. The need is to manage the workforce and facilities supply with the customer demand.

[0005] There are many comparable examples of similar operations where it is important to balance the supply of customer interaction facilities and levels of current staffing with customer demand. A telephone based customer service operation may need a certain number of service representatives to handle varying customer call-in rates. A retailer may need to activate more or fewer point of sale cash register stations. Data processing helpdesks may need to vary the assignment of staff to answering calls. The field of telephone call-center service is distinct from in-person services, because telephone switching systems typically have complete switching data (true demand), whereas conventional in-person service operations typically rely only on transaction completion times.

[0006] A great deal of effort can be applied to design and plan a business facility for efficient delivery of services. This requires forecasting demand and variations in demand when setting up the operation and also when responding to periods of peak or slack demand. On a long term cycle of planning and forecasting, a service provider's physical limitations apply. A bank branch has a given available traffic area, a set number of available teller windows, a given number of computer network terminals. This applies as well to other operations, such as a barbershop or salon that has a given number of chairs.

[0007] On a short term cycle of planning and forecasting, the available facilities can be completely staffed or partially staffed, namely by having some teller windows vacant. On a mid-term cycle, a certain number of tellers can be moved by management from a back office job to serve as a teller, or perhaps called in to work from home. It is desirable to provide for sufficient staffing to serve customer demand but also to keep little or no excess staff. [0008] The foregoing constraints and considerations of current staffing levels determine the extent to which customers may be required to wait for their turn in some sort of queue. It is conventional periodically to note the number of customers waiting to be served. Management may establish rules as to whether staff should be added when a given count is reached, so as to shorten the average waiting time, and particularly to ensure that the waiting time does not exceed some time that will evoke customer dissatisfaction.

[0009] In some situations, customers have little tolerance for waiting. If a queue appears to be long, a customer may turn to an alternative source of services. At its simplest, the customer may choose the shortest line leading to different service points. Banking operations may have special service facilities such as drive up windows, "business" windows, unattended drop-off facilities, etc. Supermarkets may have self service lines, ten-items-or-less lines, etc. These options are intended to ensure customer satisfaction, to take advantage of available efficiencies, and in large part to control the supply of service facilities and operations staff so that customers are not required to wait unduly for service.

[0010] The subject of efficient staffing is highly studied. The considerations that apply vary a great deal among different facilities and situations. Customers may be more or less committed to service of a certain kind at a certain time or by a certain provider. Customers vary their behavior for their own convenience and preference. Thus different business facilities have different ways of monitoring service supply and demand issues.

[0011] In many instances, the execution of a transaction can be determined by associated operations of a computer terminal, e.g., a cash register, teller terminal or the like. A management program is provided to collect this information and to generate reports. The reports can determine the rate of transactions executed at a particular time or by a service representative having a known identity or capabilities. That information helps the service provider to plan to provide appropriate staffing levels to account for expected demand, but the information is limited to the time and number of completed transactions. Such data does not reflect the level of customer satisfaction or dissatisfaction associated with the manner in which the services were delivered.

[0012] In a telephone service center operation, the data available to management normally may include up to the moment information, including a log of exactly when each caller arrived, was serviced or disconnected, how many callers are waiting on hold, particularly if the computer system that records completed transactions is coupled to the system that switches callers on hold to available service representatives. A telephone switched operation thus can be rather completely instrumented and is readily programmed to report the time and number of calls, the maximum, minimum, mean and average wait times, the number of callers in the queue at any time, the occurrence of callers who abandon the wait, etc. Such data permits an assessment of the customer waiting experience so as to permit the operation to maintain a desired level of customer satisfaction.

[0013] Transaction information is helpful for information such as the number of completed transactions per unit time, the precise date and time, the service representative involved and other information. However, it does not assist in determining pertinent recent history involving that customer or other customers. The customer might have looked in one or more times before queuing up. The customer might have switched lines or services or previously abandoned a wait for service. The customer might have watched as other customers abandoned their wait, as staffing was added or changed.

[0014] Business managers are naturally interested in keeping their customers satisfied. It would be possible to employ an additional staff of customer satisfaction monitors to watch the ebb and flow of customer demand and service supply. Monitoring could be less sophisticated, for example, being limited to periodically counting and recording the number of customers waiting in line, or more sophisticated, including efforts to note and record queue abandonment incidents, investigative "look-in" customer visits that may be intended to assess waiting time and the like. However, extensive monitoring can be complex and expensive. [0015] The advisability of monitoring in a banking application is discussed in American Banker, January 2005, "Retail Delivery - How Videotaping Customers Can Improve Bank Branches." Automated devices are available to collect data, for example from Brickstream Corp., Atlanta, GA. Some of the associated techniques for monitoring the movement of customers can be found in the following patents, the teachings of which are hereby incorporated in their entireties:

6,295,367 System and method for tracking movement of objects in a scene using correspondence graphs

6,263,088 System and method for tracking movement of objects in a scene

6,236,736 Method and apparatus for detecting movement patterns at a self-service checkout terminal

6,185,314 System and method for matching image information to object model information

6,141 ,433 System and method for segmenting image regions from a scene likely to represent particular objects in the scene

[0016] These techniques are useful for studying and monitoring the movements of customers and to permit management to assess the level of customer satisfaction from various occurrences that the system can detect. A customer that approaches but does not enter a queue can be determined with some probability to have investigated, assessed the wait and declined to proceed. Customers who switch lines can be deemed to be impatient or perhaps to favor a faster service representative. Customers who leave a line and also quit the premises can be construed to be dissatisfied.

[0017] By closely studying customer movements, and by correlating customer movements with other data collected, perhaps including information obtained from interviews with the particular customers, it is possible to set up an automated service status monitoring system that can be effective. But customer movement data needs to be interpreted, and the interpretation of movements may differ with changing circumstances. A change such as moving aisle-defining rope barriers to establish a queue to a different active teller location, for example, changes customer patterns. In order to properly assess and weight the importance of detectable customer movements, it is necessary to try a variety of scenarios. This is only practical and cost effective in relatively large scale operations.

[0018] Where studies have been undertaken, for example as to bank branches, the correlation of customer movements with satisfaction has been established. From 5 to 15% of visiting bank branch customers may be turned away by the perceived backlog. As many as 30% of bank customers will abandon an attempted transaction rather than enter a line of five or more customers. See, American Banker, supra.

[0019] What is needed is a way to assess in real time these and similar parameters that represent customer satisfaction as a function of service loading. However it would be advantageous if this could be done without the need to instrument and analyze all applicable facilities and/or all the scenarios that might apply at a given facility on a given date and time.

Summary of the Invention

[0020] It is an object of the invention to provide a technique whereby a plurality of customer movement parameters are detected in connection with at least one customer service operation, and the parameters are correlated to customer service levels, and by implication, customer satisfaction, and used as a basis to control operations.

[0021] It is another object to obtain by statistical studies a limited set of data representing customer movements, to infer from the limited data certain conclusions as to the probability of customer behavior, and to control service availability and/or staffing as a result.

[0022] A number of additional objects and aspects will be apparent from the following discussion of certain non-limiting examples and embodiments.

Brief Description of the Drawings

[0023] The drawings depict one or more examples and embodiments as presently preferred. Reference should be made to the appended claims as opposed to the drawings to determine the scope of the subject invention. In the drawings,

[0024] FIG. 1 is a schematic plan view showing a customer service facility.

[0025] FIG. 2 is a flowchart showing establishment of certain data value relationships.

[0026] FIG. 3 is a flowchart showing application of the established data value relationships to a facility that is partially instrumented.

[0027] FIG. 4 is a set of graphs showing an exemplary relationship of customer satisfaction to wait time, transaction time and queue length.

[0028] FIG. 5 is a sequence of views showing an embodiment of the system using a Brickstream Corp. customer movement monitor in connection with one or both of establishing data value relationships and applying such relationships as in FIGS. 1 and 3.

[0029] FIGS. 6a to 6c are time graphs illustrating exemplary customer transactions at three available queuing positions in a facility at different degrees of transaction loading.

Detailed Description

[0030] The invention as shown in FIGS. 1 to 5 concerns establishing a relationship between at least one measure of customer satisfaction versus parameters that are subject to monitoring and measuring. This can be accomplished by instrumenting a test facility that at least resembles a facility to be managed, or by temporarily instrumenting the actual facility to be managed. However, it is an aspect of the invention that it is not necessary to fully instrument and to fully and permanently monitor facilities to be managed according to the invention. Instead, a subset of the parameters subject to monitoring and testing first is collected. The subset can be sufficient of itself to make conclusions as to service level and customer satisfaction. Alternatively, additional data respecting customer satisfaction is collected, for example by interviewing a preferably random sampling of actual customers. In either case, a data set is collected of a statistically sufficient population to determine correlation between data parameter values and customer satisfaction.

[0031] FIG. 1 shows in plan view an exemplary facility 20 that in this case will be used preliminarily for data collection but resembles facilities that are to be managed as a result of conclusions made from data collected at the test facility 20.

[0032] Successive customers 22 enter the facility 20, at least some of the customers 22 intending to visit a customer servicing station 32, wherein service personnel (not shown) who may be human or automated, are to effect some form of transaction with the customers 22 involving a terminal 42. An example is a financial transaction effected by a human teller behind a counter in a bank branch office.

[0033] It is conventional that the terminals 42 employed for service to the customers 22 are coupled in data communications with some sort of network. According to the invention, at least one application appended to the network is a data collection monitor to which certain information is reported from the terminals, at least including a signal associated with the commencement or conclusion of a transaction, and preferably sufficient to report the time of both the commencement and conclusion of each transaction undertaken.

[0034] The commencement or conclusion of transactions is sufficient to determine the successful throughput of customer transactions at the facility. If commencement and conclusion are reported, the time spent during the transaction can also be monitored by managers. However, the throughput and time spent in the process of transacting business is a less than optimal measure of customer satisfaction.

[0035] An exemplary and important parameter that correlates inversely with customer satisfaction is customer waiting time in a queue for service. Another important parameter is the customer's perception of likely waiting time.

[0036] Customers 22 who enter the facility 20 are encountered by a scene of other customers 22, some of whom are in queues 33 waiting for access to a service representative (e.g., a teller). Some customers who enter the facility will conclude that their transaction is not sufficiently urgent to warrant standing in the queue 33, or that the queue is likely to be shorter if the customer visits again after a time. Such customers may depart the facility or move about in the facility before taking a place in the queue.

[0037] The customer's perception of waiting time, their movements in the area, the occurrence of abandonment when a customer abandons the wait, the frequency at which customers who visit choose to leave rather than to wait, are important measures of customer satisfaction. The Brickstream Corporation has developed an analysis system wherein a video camera 35 is employed to collect customer images and a processor is programmed to infer facts from the movement of customers in a facility. This analysis system is capable of determining a number of facts that correlate with customer satisfaction, for example by timing waiting intervals, and counting customer who abandon queues.

[0038] It is an aspect of the invention to use a fact based image analyzer such as the Brickstream Behavior IQ^tm system to develop data as described, but moreover to also collect parallel data that is perhaps less sophisticated or less expensive to obtain. Thus, for example, the data collection monitor also is coupled to one or more simple sensors 37 that are placed at locations to sense the passage of customers into and out of the facility, and their movements passing fixed sensor locations. The sensor locations can be regularly spaced or strategically placed with respect to the teller queues, as well as entrances, exits, alternative service locations such as drop boxes, ATMs that service limited functions and the like.

[0039] It may not be possible to positively conclude that any specific visiting customer turned around and exited the facility after viewing the backlog of earlier arrivals awaiting service and being dissatisfied as a result. Some other interest may have intervened. Nevertheless, if sufficient information is sensed via sensors 37, preferably after also using an automated image analyzer in a test facility to correlate sensor data with other data, it is possible according to the invention to infer customer waiting time statistically, as well as other parameters that are closely associated with customer satisfaction.

[0040] Thus, the occurrence of a customer passing a doorway or other portal into the facility is detected. Sensor around the premises can detect movement or proximity of customers in the area of the queues. This data, and in addition the data respecting transaction times, are collected at data collection monitor 44. It may also be appropriate to monitor for customer movement in areas away from the queues, which might represent customers taking advantage of secondary service offerings while hoping that the queues shortly will become shorter.

[0041] After a statistically significant sample is collected, a correlation is determined between sensor values or combinations of sensor values, and various customer satisfaction parameters. The matter of customer satisfaction may involve two or more sensed variables in addition to the timing data from the transaction terminals 44. Thus, for example, if a doorway sensor 37 detects a larger number of passages than the data collection monitor counts completed transactions, then either the queues are building or the customers are abandoning the wait. Poor customer satisfaction conditions can be inferred and presented in management reports. Of course, the information is also useful in assessing the overall demand and the extent to which the demand was serviced and serviced promptly at a given point in a workday or other time cycle.

[0042] In order to assess and manage service levels accurately, it is also advantageous to monitor the extent to which customers typically are interested in teller services. Some percentage of customers may be interested in other specific services and queues, such as account opening operations, ATM, platform and other queued services. These other services and queues can be monitored and managed in a similar manner, independent of teller services.

[0043] FIG. 2 is a flowchart showing the data collection and correlation phases. FIG. 3 shows the application of the data to reach conclusions that otherwise would require either human assessment or a sophisticated image analyzer to interpret and analyze changes in customer locations as positive or negative developments (i.e., satisfaction or dissatisfaction indicators).

[0044] In this way, management of staffing, hours, facilities and the like are then based on a forecast or inference of customer satisfaction parameters, as detected from actual customer arrivals and movements, and not only on the incidence of commenced or completed customer transactions. The loading as measured by the time distribution of transactions does not provide a true picture of customer demand, and in fact is blind to the matter of customer service level and satisfaction assessment. This is because transactions data lags or misrepresents actual demand because the transactions can only reflect the situation after the customer has waited in the queue and reached the teller, if the customer decides to undertake the wait and does not abandon the attempt after doing so.

[0045] Transaction logs are only useful to assess customer loads to the extent to which the tellers have capacity. Tellers (or other service representatives) may attempt to complete transactions more quickly when they see an increased number of customers awaiting service. But when a given number of tellers reach their capacity, additional loading produces longer lines but does not appreciably produce additional transactions. Therefore, transaction loading does not accurately represent demand or in particular the time at which the customers demand service.

[0046] By providing a correlation as disclosed, the invention provides a more accurate indication of demand than either the transaction time distribution data or the door passage data alone. Transaction times understate demand. Door passage incidence overstates demand if customers abandon or decline to enter the queue. Thus according to the invention, a subset comprising at least two of a larger set of available variables preferably is used to assess demand in conjunction with preliminary testing to develop statistically reliable correlations.

[0047] In an exemplary application of the invention, a Brickstream customer Behavior IQ customer movement assessment and a Global Management Technologies workforce management and scheduling system are operated in conjunction with one another at least at a test facility having attributes in common with one or more target facilities, and optionally during a temporary test phase of operations of the target facility itself. Then partially instrumented target facilities are operated in reliance on the data thereby obtained. Namely, the target facility is operated using only a subset of the instrumentation used in the test facility to establish a correlation between data that can be collected by the subset of instrumentation, and customer satisfaction levels that are determined with confidence using the highly instrumented full test facility instrumentation. The target facility and other target facilities or branches that resemble the target facility are operated in reliance on customer satisfaction and service level parameters that can be inferred without collecting the full panoply of data that was used initially to develop variable correlations that enable inference of customer satisfaction parameters.

[0048] Projections of staffing requirements that are otherwise generated by the workforce management system at the partially instrumented facility can be modified to account for and to improve customer satisfaction levels where possible. This composite approach provides statistically defensible improvements in customer service level and satisfaction at controlled expense.

[0049] Although it is an aspect of the invention to control expenses by partially instrumenting the target facilities, an alternative embodiment is fully to instrument all the facilities at least temporarily or periodically or after a period of changing customer loading conditions. In this way, the inference can be customized to the individual target facilities and updated when conditions change or simply to prove to the managers that the assumptions and inferences used for staffing, facilities planning and the like, are still operating as intended to maximize customer satisfaction.

[0050] Operation of the invention allows planning with due regard not only to transactions but also to how many customers pass a detector (e.g., walk through the door) and how many customers are actually waiting in line. The invention thereby detects and responds to current loading conditions. In comparison, a transaction sensitive load sensor is not responsive to waiting time, making it appear that customers arrive later, namely the time of their arrival at the teller rather than their.arrival at the facility. The invention thus more precisely and accurately reflects real load-time relationships, for example the precise time when "the lunch crowd" hits from one nearby employer, or when the 4PM shift brings their week-end paychecks from anther employer.

[0051] In addition to improved accuracy, the invention provides a way at reasonable expense to test previous projections, i.e., to determine for management purposes what actually happened in the branch, such as whether the schedule was wrong, whether the schedule was not followed, and/or whether projected wait time standards were actually achieved. The invention measures the important parameters rather than projecting them, and does so using limited input data.

[0052] Existing software associated with banking, customer service and similar transactions is capable and sophisticated in collecting data from the inside of the teller station, namely from switched operations of the tellers' terminals. That software provides for workload management, staff scheduling and the like. The invention as facilitated by sensors disposed on the customer side of the teller window allows a substantial improvement in the precision of time and loading data, thus improving the quality of the staffing demand forecast. The scheduling situation in facilities having wide swings in loading at different times is vastly improved. The customer side data can be analyzed in conjunction with teller terminal data to assess the effects of customer load variations on the accuracy and efficiency of particular tellers or other service representatives.

[0053] The service attributes most valued by consumers of bank branch and similar services are fast teller access, convenient locations, long hours of operation, freedom from errors, availability of multiple optional banking channels, and layouts that facilitate speedy completion of transactions. As shown in Fig. 4, time and wait queue considerations are quite clear. A wait over five minutes is too long. A queue over five persons is too long. These levels can be construed as an acceptable level of services. The responsiveness of the customer loading and staffing system should be quick and accurate enough to maintain these benchmarks. [0054] FIG. 5 demonstrates a Brickstream video monitoring arrangement in an exemplary bank branch. In this arrangement, the video monitor comprises two overhead cameras. In this case the customer path involves a single queue leading to a number of tellers. As an alternative, the posts defining the queue can provide a technique for proximity sensing of customer locations, which would be comparable in a fixed customer queue arrangement, but these arrangements may be altered from time to time, in which case the video sensing arrangement is most effective. After a sufficient period of test data collection, however, various sensors (even a few sensors on selected queue posts) can be statistically correlated to the condition of the queue. If operated together with passage monitoring such as doorway entry/exit events, the collected data provides the necessary input to operate as discussed.

[0055] A first generation of disciplined staffing management ensued when time and efficiency studies employed manual data collection techniques in an effort to determine staffing needs in bank branches and similar facilities. A second generation of staffing management was enabled when powerful computer technology associated with the customer service transactions was improved to the extent that it collected data to enable automated staffing schedule creation based on customer loading. In a third generation, bank branches and central operations are web connected and the software for forecasting loading and staffing needs is increasingly sophisticated. What has been missing heretofore is a technique to respond not only to the loading as perceived at the point of the transactions, but also the loading perceived by the customers as wait time, annoyance and frustration that from time to time has led customers to switch bankers.

[0056] In conjunction true loading data, the managed facilities can establish a service level and sales goals that are appealing to customers. True demand is sensed by measuring demand from the standpoint of customers arriving to be served as opposed to customers who have successfully run the teller's gauntlet. Although full real time video monitoring is advantageous for continuous ongoing monitoring, by sensing a subset of the data available from full monitoring the invention permits continuous monitoring of arrival, waiting and service times in a statistical manner. In addition to forecasting and scheduling to meet expected demand according to management goals, the invention also enables ongoing performance monitoring in the same statistical way.

[0057] The particular sensors employed to sense movement of customers comprise distributed, input sensor that may be operated by people at least during preliminary data collection phases, or may be more or less automated up to the extent of automated image analysis video systems. The sensors can be fixed or placed according to a plan, such as along an aisle or path defined by theater-like rope barriers defining queues. The data collection and input devices can comprise one or more of cameras, stopwatch spotters or timers (human operated or timed passage photodetector pairs), door/passage counters that can be simple or paired to distinguish entering from exiting, proximity detectors especially placed to sense queue length and access to alternative services, etc.

[0058] An object of the invention is to determine loading in an up to the minute sensing manner. As a practical matter, intermediate output and collected data summaries can be generated according to a schedule convenient for management to take corrective action when needed. For example, reporting periods or report offsets between events and transactions can be scheduled to occur regularly, e.g., in 5 minute, 15 minute or 30 minute reporting periods (perhaps reporting different parameters at different incremental scheduled times) throughout the day. The scheduling can be the same each day or modified over a cycle such as a week or month to accommodate customer pay periods and other special days such as three day weekends, holidays, etc.

[0059] Instead of monitoring and reporting by time increment, the system can log data regularly but generated alarms based on loading conditions or other detected quality of service parameters. For example, an alarm can be generated if the ratio of customer abandoning the wait for tellers is considered to exceed a preset percentage, which is optionally subject to variation as a management choice. An alarm can be generated if activity falls below a threshold related to the number of tellers on duty or the number of transactions reported over a period.

[0060] Processing of the data from the customer passage or proximity detectors and/or timers is intended to at least statistically relate to the whether an arriving customer who required service enters a queue or not, and if so, the time interval between the point at which the service is requested, namely arrival of a customer, and the point at which the customer obtains the service from the teller. The associated variables include facility entry and exit counts and times representing the customers who balk or abandon, the progress of customers over time in their passage from the entry to point at which their transaction is processed by the teller, the associated queue length (mean, average, minimum, maximum), and similar counts and measurements.

[0061] As stated above, full data collection is advantageous. When image analysis, human monitoring for counts and timing are not practical, passage counters of customers to the point of the transaction, and calculation of associated timing and percentage figures, can be employed.

[0062] Preferably, the data is processed as an application associated with a staffing forecasting and planning software system as otherwise known in the trade. However the invention renders the calculations more timely and accurate than known arrangements. The processing of the data to affect forecasts, planning, reports and the like can be selectable or built into the forecasts otherwise made. Insofar as the data is based on a subset of sensor inputs as compared to a video image analyzer, the available calculations may be more limited that might be available from an image analyzer.

[0063] The techniques according to the invention can rely to a greater or lesser extent on the respective data values used as inputs. For example, as active transaction time reaches a greater percentage of the available time, conclusions can be made regarding the service level. Although extensive data input is advantageous, the invention can be useful with limited data, including transaction start and/or end data alone. [0064] This aspect of the invention is illustrated in FIGS. 6a through 6c.

Assuming, for example, that a teller or other service representative may have a nominal capacity to complete three average transactions 122 within a given incremental time span 125, a service pattern for a group of tellers as in FIG. 6a could represent a situation in which there is no queue or a very short queue. The customers generally distribute themselves among the available queues. The tellers keep up with the demand and in this instance might have sufficient capacity to handle the demand with fewer tellers.

[0065] In FIG. 6b, the transaction density has increased to the point that more than a nominal number of transactions is occurring during the nominal time span 125. It may not be possible in a conventional system to determine the state of the queues or to assess waiting time. However, one can apply certain knowledge based rules to discern, for example, that there are time gaps occurring between the transactions even at this transaction density. Given that and also the fact that customers generally move to the shortest queue, one can deduce from the transaction timing alone that there is still sufficient capacity to meet the customer load. There may be some queuing, but the occurrence of uneven inter-transaction times, some of which could equal the time of a transaction, shows that there is minimal queuing and the customer transactions are moving nicely.

[0066] In FIG. 6c, there is no substantial inter-transaction time. One can infer from the transaction time data that there are standing queues. The wait times are considerably longer than in the previous case and more labor is absolutely required. According to the invention, the transaction time data can be integrated and preferably also correlated with measurements concerning the rate at which customers may be balking or abandoning the queues. In the preferred arrangement, the additional sensed information also permits an assessment of these factors, enabling inferences to be made as to the service level being maintained and the need to adjust staffing levels.

[0067] Over time, information obtained according to the invention allows the operator of an establishment to refine and improve projections based on the available data and the service level as detected in one or more of these ways. The available data can be extensive, as with an image analyzer, or at least partial and preferably correlated to data from careful testing. Or according to the invention, the available data can be minimal.

[0068] For example, an image analyzer might distinguish adults from children, staff personnel from customers, etc. If only passage or proximity detectors are used, the software process can infer the true demand from detector inputs plus historical cross correlations between detector counts and entry/exit counts as well as teller transaction completion counts, applying offsets as a guide. In any case the point is to calculate and if necessary approximate the time increment between the time that service is first demanded and might be possible, versus the time the service is actually delivered, and to factor that increment into the loading assessment of a customer service facility that otherwise would have no basis other than the completion of transactions to automatically assess the current customer load.

[0069] The foregoing examples and preferred arrangements are intended to be representative and not limiting. It will be appreciated that the invention is subject to a variety of specific implementations that benefit from one or more of the techniques and instrumentalities discussed herein. Furthermore, the invention has been described with reference to bank tellers, and advantageously could be applied to other environments such as retail product and service sales, restaurants (especially fast food), government services, etc.

Claims

What is claimed is:

1. A method of managing at least one facility with respect to staffing loads that change with variations in customer demand, comprising the steps of: providing a data collection monitor; equipping a test facility with a plurality of sensors in a test mode, and tracking movements of a plurality of customers about the facility using the sensors, wherein data corresponding to the movements are transmitted to the data collection monitor from the sensor for a plurality of customers; analyzing the data transmitted to the data collection monitor in the test mode, including correlating at least one customer satisfaction level with data corresponding to said movements for a subset of the plurality of sensors; equipping an operational facility with said subset of the plurality of sensors, wherein the operational facility can be said facility used in the test mode or another facility that resembles said facility used in the test mode; running the operational facility while collecting data from the subset of sensors and correlating collected data from the subset of sensors, and inferring said customer satisfaction level exclusively from the subset of sensors.

2. The method of claim 1 , further comprising adjusting staffing used in the operational facility based on the inferred customer satisfaction level.

3. The method of claim 1 , the subset of sensors includes customer proximity detectors at least at an entrance and an exit of the facility.

4. The method of claim 3, wherein the facility is further equipped with a sensor at a customer service station.

5. The method of claim 4, wherein the subset of sensors includes keystrokes from terminals at the customer service station operated by one of the customers and staff members serving the customers using the terminals.

6. The method of claim 4, wherein the subset of sensors includes at least one customer proximity detector placed to sense a predetermined customer queue length.

7. The method of claim 6, further comprising inferring a length of plural customer queues from the detector placed to sense the predetermined customer queue length.

8. The method of claim 1 , further comprising adjusting a configuration of an area where customers for access to a customer service station.

9. The method of claim 1 , comprising generating at least one of an alarm and a report when inferred customer satisfaction levels meet a predetermined criterion.

10. The method of claim 5, wherein said analyzing comprises inferring statistics from the movements of the customers and operation of the terminals, the statistics representing a time duration that customers remain in the facility, and wherein said customer satisfaction levels are at least partly correlated to said time duration.

11. A system for managing staffing loads that change with variations in customer demand, comprising: a test facility equipped with a plurality of data collection monitoring sensors operable to collect information representing customer activities in the test facility, said sensors including at least one customer proximity sensor and at least one of a customer queue length monitor and a process for determining timing of keystrokes executed on terminals operated to process customer transactions; at least one data processor coupled in communication with data collected by the data collection monitoring sensors, wherein the processor is operable to collect data representing movements of a plurality of customers using the test facility and data representing at least one customer satisfaction parameter, wherein the processor is programmed to correlate the customer satisfaction parameter with data from a subset of said plurality of data collection monitoring sensors; an operational facility that is at least partly configured to resemble the test facility and is equipped with the subset of data collection monitoring sensors, wherein the operational facility is equipped actively to collect data from said subset of sensors; and, at least one operational data processor coupled in communication with the subset of sensors and programmed to infer and report a customer satisfaction level.

12. The system of claim 11 , wherein a single facility functions as the test facility with the plurality of sensor and as the operational facility using said subset.

13. The system of claim 11 , wherein the operational facility is configured similarly to the test facility such that correlation of the customer satisfaction parameter with data from a subset of said plurality of data collection monitoring sensors at the test facility at least partly is true of the operational facility.

14. The system of claim 11 , wherein the plurality of data collection monitoring sensors include a video sensor operable to collect relatively more extensive customer location data in the test facility, and at least one customer proximity sensor operable to collect relatively less extensive customer location data, wherein the less extensive customer location data correlates to the more extensive customer location data, and the less extensive data is employed in the operational facility.