To
begin, let us clearly define workforce management historical
data integration. The automation of exporting the historical
data from an ACD or other data resource and importing to
WFM is commonly referenced as WFM historical data integration
for the use of reporting and forecasting staffing requirements.
Forecasting
activity workload is one of the core features of a WFM solution.
In order to accurately forecast activity, the patterns of
historical activity, also termed historical forecast models,
need to be built based upon the collecting and proportional
weighting of historical data that best represents the time
increment being forecasted. For example, the last four Mondays
are the best representation of next Monday. The source of
the historical data is typically an automatic call distributor
(ACD). WFM vendors have developed automated processes along
with ACD vendors to make the process efficient, easy to
implement and easy to administer.
However,
the point to be made is that the data elements requested
by the vendors should still be examined to see if they satisfy
the center's business needs. A center may have unique business
processes that require changes to the default data element
and report settings.
Before
setting up the data integration process, the data to be
collected must be identified. The base set of data elements
collected in WFM are offered events or calls, and average
handle time (AHT).
Workload
= events * AHT
In
addition to events and AHT, other data elements are also
collected primarily to enhance performance reporting, such
as agent performance. Allowing WFM to collect multiple data
elements promotes the workforce management system's use
as the contact center's centralized reporting performance
tool.
Close
attention should also be taken to determine which items
are going to be collected and reported as is, and which
items are going to be calculated within the WFM application.
For instance, is service level going to be collected or
is it going to be a calculation. If the WFM reporting calculates
the service level, it will likely not identically match
the service level reported by the ACD. This can make a substantial
difference in determining a centers performance and should
be identified prior to implementation.
A
CLOSE LOOK AT THE DATA INTEGRATION SETUP
Now
let us look at the setup of data integration process. The
data integration process itself is easy to implement and
usually requires little maintenance. However, not placing
the proper human resources to support an implementation
setup can cause a project to come to halt. Existing WFM
systems also need defined administration resources to assure
issues are efficiently addressed.
Time
increment data integration is the preferred means for reporting
the number of events offered over a period of time. This
also is the common process for collecting other variables
that are populated within WFM. This process is setup by
scheduling a designated set of reports in an ascii, text
file format containing all desired data elements to be sent
to a specific location on a WFM server or a workstation
dedicated to the process. The server then recognizes the
file and processes the information by importing the data
into the associated activity tables defined within WFM.
Most
WFM vendors have partnerships established with the major
ACD vendors to established generic reports specifically
for the data integration, usually one or two reports. WFM
vendors will provide the name(s) of the report(s); the reports
are purchased and programmed by the ACD vendor; the WFM
vendor then sets up the import process to the WFM data tables.
The ACD vendors charge $5 to $10k for each scheduled report
with no maintenance. Based upon the points made within this
article, each report should be reviewed for changes based
upon unique business requirements. Changes usually result
in a $1k additional charge to the changed generic report.
Charges for necessary changes are a small price to pay compared
to the total investment that would otherwise be producing
inaccurate center performance reports and staffing requirements.
A
CLOSE LOOK AT EVENTS
For
the setup of data integration, WFM vendors will ask for
events offered to use as the basis of calculating staffing
requirements. Some ACD will report offered while other ACD
will report handled and abandoned which comprises offered.
Offered
= Handled + Abandoned
Workload
= (Handled + Abandoned) * (AHT)
No
matter what ACD type, each data element needs to be validated
to assure the formulas correctly represent the data element
and business needs.
Handled
and abandoned values are usually easy to identify. However,
finding modified data element formulas within ACD reports
is not uncommon. This is especially true when the ACD has
been in place for several generations of IT personnel managing
the systems. Validation of the reporting consistency and
historical changes is always strongly recommended.
One
of the most common reasons for a changed formula is due
to the business decision to not include all abandon in the
service level performance goal. The majority of centers
will reference all events entering a queue to be potentially
worked upon as an offered event; however, some centers discount
abandoned events that occur during a defined length of time
within the queue. Special attention should be taken to assure
that the reported offered in WFM is the same offered that
is being used to grade the service level of an activity.
If reporting is not consistent, the data will not result
in an effective staffing requirement forecast. If discounted
abandons are factored in the calculation of service level
performance but not discounted in the integration of offered
events to the WFM system, the calculation of the forecasted
offered will be overstated resulting in overstated staffing
requirements.
An
example is a center with a goal of 80% of offered within
20 seconds but all abandoned events discounted within the
first 20 seconds. Most WFM vendors will only have algorithms
that reference the arrival patterns based upon the time
the event enters the queue.
Even
if the discounted abandons are factored out of the abandon,
the calculation of staffing requirements will be factoring
an algorithm-based percentage of the offered calls to still
be abandoned prior to the time limitation parameter of the
service level performance goal resulting in the understating
of staffing requirements. Seldom will a vendor customize
the forecasting algorithm to match a customized service
level calculation with a modified abandon formula, but the
trend in new releases and new vendors is to allow users
more options and the ability to manually customize the abandon
delay in the algorithm.
To
go to an extreme, event accounting is especially sensitive
when activities within the same center have different abandon
discount rates due to different service level performance
measurements. For example, activity A has a service level
of 80% in 20 seconds and abandons discounted up to 20 seconds,
while activity B has a service level of 80% in 30 seconds
and abandons discounted up to 30 seconds. In this case,
multiple customized integration processes need to be developed
for each scenario or a complex custom integration report
will need to be developed that compensates for the different
activity formulas prior to the integration process. This
is extremely complex and can be potentially quite expensive
when the environment changes resulting in the integration
process and reports to be scraped and re-implemented with
the new formulas.
Based
on the points made, three recommendations need emphasized.
The calculation of abandons in the formula for offered events
needs validated against each activities performance measurement.
Secondly, serious consideration should be taken to re-evaluate
the business positioning of discounting abandons compared
to the creation of a consistent performance measurement
across all activities for reporting as well as in the calculation
of forecasting staffing requirements. Lastly, develop and
communicate processes to support the integration process
to assure accurate data and to escalate issues for quick
resolution.
