High Definition WFM™ ROI Background
This document presents the return on investment business case for companies that wish to switch to High Definition WFM from either one of the following environments:
- Manual forecasting & scheduling using spreadsheets and Erlang calculations
- Automated forecasting and scheduling with any WFM solution that uses: Erlang based forecasting methods; random trials scheduling methods; binary adherence measurement.
The operational disadvantages of older planning methods are summarized below:
- Switching from a manual scheduling practice to a WFM solution that uses the same 100-year old forecasting method can at best perpetuate forecasting inefficiencies.
- Random trials scheduling methods typically generate schedules that are so inconsistent so as to be unusable. Manual or fixed scheduling perpetuates so no benefits accrue.
- Conventional WFM solutions often claim to provide an ROI because by way of reducing staffing levels. However, these are false savings. Older planning methods blindly reduce staffing levels based on the assumption that event driven activity is arriving very evenly. These older systems have no choice but to make such assumptions because they only make use of simple call counts and average durations. The use of such non-descript data leads to abhorrent staffing patterns that compromise revenue. If any cost savings result from understaffing, these savings are not sustainable. Frustrated customers generate less revenue, cost more to service and soon take their business elsewhere.
- Conventional shift trading processes are not well engineered. Agents still need to spend extensive time negotiating trades. Supervisors must still cope with all the complexities of overlapping trade requests.
- Conventional adherence measurement makes overly simple comparisons that are too cluttered to be meaningful.
High Definition Forecasting™ vs. Conventional Methods
High Definition WFM ROI flows largely from the improved accuracy of having an intricate, detailed and distortion free understanding of historical demand. These are real, sustainable cost savings. The substantial differences between conventional interval based forecasting methods and High Definition Forecasting are described below.
Interval Based Forecast:
Interval based forecasting was invented by Agner Erlang over 100-years ago. Erlang died 80-years ago, long before the invention of computers and detailed call tracking. His method was designed for the limitations of the day which only allowed him to obtain hourly or half hourly call counts by manually recording odometer readings from a mechanical switch. The inputs to Erlang’s methods included number of calls, average duration and target service level.
Erlang recommended that his method never be used for event driven activity. Event driven means that call volumes are either building to a peak or dropping off from a peak. Most if not all modern call centers face event driven activity throughout each day. With the Exception of WFM Wisdom, all known WFM solutions continue to use these 100-year old interval based forecasting methods. These older methods cause serious distortions that drive call center staffing patterns into dysfunctional states.
Any WFM system that uses the Erlang method is unable to increase staffing levels during periods where the customer base has been under serviced. This reality is often clouded by the fact that WFM solutions do collect past service levels for reporting purposes. As much as the data is collected, past service level is not considered in any Erlang calculation.
The onus is on the planner to recognize that older forecasting methods are limited to planning to a level at least as low as the capacity that was in place when statistics were gathered. The planner is suppose to make manual adjustments to the automated forecasting outcomes to correct for past under performance. These manual adjustments are rarely done and even more rarely are they done well.
High Definition Forecasts:
High Definition Forecast methods are the modern incarnation of what Erlang wanted to achieve but lacked the raw materials to perform. High Definition Forecasts are derived from the exact details of all calls. Call Details are gathered from all levels of the system including call router, IVR and ACD. The fully automated analysis of this data yields forecast that intricately recognizes the real skew of activity and all carry over effects. High Definition forecast also compensate for past periods of low service levels, abandons and service denial.
These unique capabilities produce distortion free capacity planning that automatically pulls resources into the evolving demand patterns of a growing call centre.
Logic Based
Uniform Scheduling™ vs. Lesser Methods
Another key principal of High
Definition WFM is the use of logic based scheduling algorithms instead of
random trials. The reader will benefit from an understanding of how these
two methods are different.
Random Trials
Scheduling:
Random trials (frequently called simulations), are an old and ineffective means
of creating a close fit between staff levels and a forecast. The process
of Random trials is to randomly assign agents to a schedule starting with the
left most interval of the first day of the week. Moving rightward additional
random assignments are made. Typically, agent shifts that cross over the
right hand boundary of the forecast are cropped to very short shifts. This
process is repeated approximately 20,000 times for each day of the
schedule. The most efficient schedule is chosen independently for each
day. Key deficits associated with this method are as follows:
· A proliferation of very short evening shifts that are all impossible to staff
· Extremely inconsistent and start timed for agents from day to day and week to week
· Schedule shrinkage from left to right unless available resources exceed the area under the forecast. The outcome is serious understaffing on evenings and towards the end of the week
· Exceptionally long schedule building timeframes in return for mediocre schedules
· Very poor allocation of a multi skilled labor force.
Logic Based Uniform
Scheduling:
High Definition WFM was designed from
the ground up to incorporate a scheduling algorithm that is capable of
translating a granular high definition forecast into an uncompromisingly
effective schedule. Agents receive consistent schedules and fair
rotations. Skills are effectively layered for optimal coverage across the
entire week. The scheduling method is also fault tolerant to forecast size.
Even if your forecast is many times larger than the available labor pool,
resources are still allocation rationally across time of day and work
week. This method also allows a global forecast to be automatically and
evenly allocated to multiple islands of productivity such as a plurality of
internal call centers, outsourcing partners and remote workers.
Uniform scheduling is a fully automated, single click process for building enterprise schedules. The process runs in about 10 – 20 minutes per 300 agents.
Uniform scheduling supports any blend of fixed, variable and rotational schedules.
Proactive Shift Trading™ vs. Lesser Methods
The interfaces and process for shift trading constitute another area where High Definition WFM delivers ROI that is unattainable from lesser WFM solutions.
Conventional Shift
Trades:
Conventional so called electronic shift trading requires agent to do a great
deal of legwork to propose one shift trade at a time. Typically, the
agent who needs to trade must graze the schedule to find a viable trading
partner. If the agent proposes more that one trade for the same shift, the
supervisor’s job becomes extremely complex. So while conventional shift trades
have an electronic component, most of the work is manual and little if any
stress and workload are removed.
Proactive
Shift Trading:
High Definition WFM incorporates a higher art of electronic shift trading
called Proactive Shift Trading. Proactive Trading automatically provides
an agent with all of the viable trade options:
· Trade one day for another
· Trade one time for another (same day)
· Offer to give away a shift
The agent may solicit as many shift trades as are desired in one fell swoop. Recipients of the trade request reply with a level of interest. Trades can be approves automatically, or by supervisors. In either case, as soon as the top ranked trade is approved, all overlapping trades and offers are automatically dispensed with by the system. The outcome is a significant reduction in absenteeism and additional savings from the reducing administrative effort to almost zero.
Intelligent Rich State Adherence™ vs. Lesser Methods
Vendors of conventional binary adherence promise big labor saving. These savings have little chance of materializing. High Definition WFM features a more highly-engineered form of adherence called Intelligent Rich State Adherence (i-Adherence™). The key differences are explained below.
Binary adherence:
Binary adherence makes consists of rudimentary same minute comparisons.
If an agent’s phone state is the same as scheduled then they are considered in
adherence. Unfortunately, calls seldom end exactly at the end of a
scheduled state change. The result is arbitrary non-adherence that is all
but meaningless to agents and supervisors alike. Typically there is so much
clutter that companies have to tune down the system to ignore the first three
minutes of non-adherence after each state change. This is a license for
underperforming agents to free load and an incentive for otherwise diligent
agents to follow suit.
i- Adherence:
i-Adherence analyses the flow of state
changes across the day to determine if agents are honoring the spirit of the
schedule. Are the agents taking their breaks as soon as they are able
to? Are agents taking the proper amount of break and lunch time
throughout the day. The results are presented in real time to agent and
supervisors in a manner that both find fair and meaningful. The remarkable
outcome is that agents are able to self-conform.
Rich State adherence also performs Payroll reconciliation in real time. The Schedule, punch clock, ACD state changes and supervisor approved exceptions are automatically processed into payable hours and performance measurement statistics.
Return on investment calculations
1) Improved Labor Utilization due to improved forecasting
More accurate forecasts are likely to reduce agent idle time, improve revenue, reduce outsourcer costs and improve customer satisfaction. Conventional WFM applications are limited to 30 minute planning intervals because these older technology forecasts are susceptible to call pegging errors. Call pegging errors occur because conventional forecasting methods can only peg calls to one interval even if the call crosses the boundary of one or more planning intervals. The call pegging errors tend to shift and squeeze forecasts in a manner that causes understaffing on both the leading edge and trailing edge of a forecast. Conventional 15-minute interval forecasts cause the call pegging errors effects to double. These limitations of conventional forecasting methods will impose states of dis-optimization that increase over time.
Staying with 30 minute conventional forecast intervals reduces the rate of decline (relative to 15 minute conventional forecasts). However, 30-minute planning intervals inevitably cause call centers to flat staff for 30 minutes at a time. Flat staffing across sloped demand curves is a key driver for both low service levels and low labor utilization.
The simplified illustration below shows the effect of flat staffing across demand that is either building or tapering off. In most call centers, over 95% of planning intervals face these types of skewed demand. The orange shaded areas depict the periods of overstaffing (aka low labor utilization). The red shaded areas depict periods of under staffing (aka poor service levels, frustrated customers, longer talk time and lost revenue)
The grey horizontal lines represent the flat staffing patterns that arise from assuming that demand is not skewed (aka random). These mediocre staffing levels give rise to a toggling effect. Toggling causes call centers to alternate between periods of overstaffing and understaffing. Toggling effects are an unavoidable consequence of conventional interval based forecasting methods. Planning systems that use these methods cannot be a source of improved efficiency, reduced labor costs and improved revenue.
High Definition Forecasts are immune to the interval effects described above. This modern method eliminates all call pegging errors. Even calls that span many intervals are accurately translated into granular demand for agents. High Definition Forecasts intricately understand skewed demand patterns. A fully automated, logic based scheduling algorithm translates these forecasts into shift patterns and break and lunch timings. The outcome is a plan that allows on phone resources to track intricately to historical demand.
These technological strengths eliminate the toggling effects described above.
Understaffing (red) represents periods where allocating additional labor would improve customer access, service levels and revenue.
Lost labor utilization (orange) is easily measured and attributed directly to toggling effects and call pegging errors.
The following forecast analysis compares conventional forecasts to High Definition Forecasts.
Above, the yellow line shows a distortion free measurement of demand. The pink line is a demand measurement that is subject to call pegging errors and toggling effects.
The gaps between these two curves allow us to easily measure the hours of overstaffing and understaffing. The green line shows the cumulative understaffing hours (total = 24 hrs). The blue line shows cumulative overstaffing hours (total = 27 hrs)
The measured distortions included call pegging errors, over-allocation to periods of flat demand, under-estimating peak demand and misinterpretation of carry-over effects. Differences between the interval based forecasts and High Definition Forecasts indicate that 27 hours of idle labor can be reallocated to periods where the labor can be fully utilized. The 27 hours of labor savings break out as follows:
24 hours of new labor productivity
3 hours per day of reduced payroll cost (agents not needed)
27 hours per day of total savings
The total number of labor hours required to staff to the conventional forecast is 793 hours.
As a percentage of scheduled hours, the implied improvement in labor productivity is 3.4% (calculation: 27/793 = 0.03405)
The above analysis compared 30-minute conventional forecasts to High Definition Forecasts prepared to 15-minute granularity. The conventional 30 minute interval forecasts featured call pegging errors of 7%. In other words, 7% of the work was pegged to an incorrect interval.
Performing the same comparison as above, using conventional 15 minute interval forecasts, causes the call pegging errors to double (15%). This is such a large phase shift that the call centre would be in a constant state of understaffing. Labor utilization would be nearly 100% but at the expense of very high redial, high abandon, frustrated customers, long talk time, fatigued agents and lost revenue.
Switching from conventional 30-minute interval forecasts to conventional 15-minute interval forecasts would be very damaging to this call centre’s business. This is a key point because most WFM solutions promote their interval based 15-minute forecasts as being more accurate than 30 minute forecasts. The reverse is true.
Unlike interval based forecasts, High Definition Forecasts are distortion free across any forecast granularity (15, 10, 5, etc.). Hence the granularity brings improved labor utilization, better service levels and more revenue with no downside.
2) Increased Revenue from better timing of Agents
Calculating the value of increased revenue requires calculations that are tuned to the individual business. For example, some businesses have a monthly billing relationship with their customers (ie,Cable, phone, internet). Other businesses like take-out food chains have customers that return frequently with new transactions. The degree to which customers are denied service by the business in question is a key factor in determining how much revenue can be gained from improved planning. The following examples illustrate improved revenue that was measured under various scenarios.
Case 1) Food ordering industry, Regional Chain, Low service denial.
This study found that 5% of callers who encountered excessive wait times elected to either not order or to order from an alternate company. Lost revenue from inappropriate staffing patterns was measured at $200,000 annually.
Case 2) Food ordering industry, National Chain, high service denial.
This Study involved a national food ordering chain with an exceptionally poor fit between demand for service and available resources at various sites. The problems were compounded by an ineffective call routing logic that sent floods of network wide traffic to one outsourcer at a time. Large volumes of calls were therefore transferred to busy lines. As is typical of older planning methods, staff planning was done on the basis of calls successful call counts. These planning methods perpetuated the poor timing of staff and allowed the organization to ignore the impact of denying service to customer.
Over the course of 2 weeks the following statistics were tabulated from network routing details.
Transfer errors are specific instances of customers encountering busy signals. Lost orders are specific customers who gave-up on ordering after one or more unsuccessful attempts. The estimated annual lost revenue from lost orders was $3.7M based on the assumption that no customers made permanent brand changes. Assuming some portion of the customer base did make permanent brand changes, the revenue losses would be much higher.
High Definition Forecasting uses network level data to intricately measure unsatisfied demand that may result when customers encounter busy lines and/or long wait time messages that cause them to hang-up prior to being offered to the ACD queue. This technological advantage not only sets staffing levels correctly but also identified when infrastructure and or routing logic are not properly engineered to ensure that callers reach agents.
Case 3) Telecom/ ISP Industry, National Carrier
A national telecom carrier was using the conventional forecasting methods of a market leading WFM solution. Despite diligent forecasting and scheduling efforts, the call centre falls into a recurring pattern of extremely long wait times, high abandons and high redials. The contact centre then adopts High Definition Forecasting and Uniform Scheduling. The scheduling patterns are shifted dramatically. The immediate effect is high service levels, reduced talk times and significantly reduced account cancelations. The following graphic illustrated the magnitude of the poorly timed labor problems that were able to persist prior to High Definition Forecasting.
The original resource levels are presented in grey. These staffing levels were set according to forecasts that were based on calls offered to the queue and assumed that calls arrived randomly. The grey line is a good example of the degree to which older capacity based forecasting methods will pull staffing levels into dis-optimized states. High Definition Forecasts produced a dramatically different forecast, schedule and resource levels. The High Definition Staffing levels are shown in pink. The total correction is shaded green. The total distortion is shaded red.
The impact of the change was that managers and supervisors suddenly found themselves slightly ahead of the demand curve. Agents, who had previously faced the continuous pressure of long queues, found the revised schedule allowed them to easily meet service levels--with no significant reduction in labor utilization. Talk time decreased by 5%. A high cancelation rate was substantially eliminated.
The effect shown above is due in part to the super capacity forecasting capabilities of High Definition Forecasting. Conventional methods are unable to move resources into periods of excess demand. High definition methods accurately distinguish the pent-up demand that older methods either ignore or misinterpret.
The speed and accuracy with which high definition methods can turn around a distressed call centre is a compelling benefit. High Definition methods can right set the staffing patterns in a single cycle of forecasting and scheduling. High Definition methods also ensure that a call centre with a growing customer base is able to keep pace with change.
3) Improved Operational Efficiency
· High Definition Forecasts allows forecasters to create weeks of distortion free forecasts in less than a minute. Forecasts are automatically adjusted for seasonality, absenteeism, multi-skill efficiency, holiday effects and other factors.
· Logic based Uniform Scheduling produces weeks of schedules for hundreds of agents at the push of a button. Unattended processing time is about 6 minutes per 100 agents per 4 weeks of schedules. Agents get highly consistent schedules that respect preferences, rotations and individual needs for exceptions.
· Web based schedules and schedule change notifications eliminate the administrative effort of publishing and maintaining paper schedules.
· Proactive shift trading eliminates the costly inefficiencies that result when agents graze the agent pool to find trading partners. The administrative effort is reduced to just seconds per trade. Customers report significant reductions in absenteeism rates. This is especially true for unpopular work times and/or remote workers. Customers have seen 20% absenteeism rates on weekends drop to less than a percent as a direct result of Proactive Shift Trading.
· Better Schedules plus effective shift trading interfaces reduce agent attrition rates. This in turn reduces training costs and helps to retain the most productive agents.
· I-Adherence reduces excess breaks and keeps resources focused on generating revenue. Typical results include a 4% improvement in labor productivity. The administrative effort of payroll reconciliation can be reduced from days to about an hour per 100 agents.
ROI Summary:
|
ROI Source |
Calculation |
Total Annual ROI |
|
Improved Labor Utilization |
3.4% of Labor Cost |
|
|
Improved Revenue |
Specific to Customer, industry and current level of service delay or service denial. However, the substantial elimination of the current cancelation or brand switching rate is the best estimate of ROI. |
|
|
Reduced Talk Time |
Compare the average talk time of callers who waited longer than 5 minutes with those who wait less than 5 minutes. Take the percentage difference and multiply this by the annual labor budget. |
|
|
Reduced Absenteeism |
Compare the absenteeism rate on weekends and evenings to early and mid-day shifts. Multiply the percentage difference by annual cost of sick pay. |
|
|
Improved Adherence |
4% times the annual payroll cost
|
|
|
Payroll Reconciliation Time Savings |
Cost savings associated with reducing the time it takes to reconcile payroll down to 1 hour per 100 agents per pay period. |
|
|
Forecasting Time Savings |
Cost savings associated with reducing the time it takes to create forecasts down to 1 hour per 100 agents per forecasting cycle. |
|
|
Scheduling Time Savings |
Cost savings associated with reducing the time it takes to create forecasts down to 1 hour per 100 agents per scheduling cycle. |
|
|
Shift Trading Time Savings |
Cost savings associated with reducing the time it takes to manage the current shift trading practice down to 1 hour per 100 agents per week. |
|
|