NICE WORKFORCE

MANAGEMENT SYSTEM
Understanding the Multi-skill
Efficiency Algorithm

WHITE PAPER
CONTACTS
The full list of NICE marks are the trademarks or registered trademarks of Nice Ltd. For the full list of NICE trademarks,
visit http://www.nice.com/nice-trademarks All other marks used are the property of their respective proprietors.

Global International HQ, Israel, EMEA, Europe & Middle East,

T +972 9 775 3777, F +972 9 743 4282 T +44 0 1489 771 200, F +44 0 1489 771 665

Americas, North America Asia Pacific, Singapore Office

T +1 551-256-5000, F +1 551-259-5252 T + 65 6222 5123, F +65 6222 5459

About NICE
With NICE (Nasdaq: NICE), it’s never been easier for organizations of all sizes around the globe to create
extraordinary customer experiences while meeting key business metrics. Featuring the world’s #1 cloud
native customer experience platform, CXone, NICE is a worldwide leader in AI-powered self-service and
agent-assisted CX software for the contact center – and beyond. Over 25,000 organizations in more
than 150 countries, including over 85 of the Fortune 100 companies, partner with NICE to transform - and
elevate - every customer interaction.

www.nice.com Copyright © 2021 NICE Ltd. All rights reserved

 Overview 3

The Challenges 6

TABLE OF The Solution 7

CONTENTS 1. Different Service Objectives 7

2. Complex Routing Scripts & Multi-Skilled Employees 8

3. Pooling Principal 10

4. Composite MSEff 11

Benefits 12

3 4
NICE WORKFORCE The Challenges
MANAGEMENT Multi-skill Efficiency is used to adjust ErlangC requirements
to account for efficiencies of multi-skilled employees and
SYSTEM simultaneously queuing a contact to multiple groups of employees.
Rather than require the user to derive these adjustment values (and
to eliminate the reliance on estimates based on assumptions), NICE’s
WFM solution automatically calculates the efficiencies for the user.

One method of determining the MSEff value would be to review

historical (or simulated) staffing and requirement information to

Overview identify periods where service objectives were met or exceeded

with fewer employees than ErlangC would have required. This is a
difficult analysis if attempted manually. Even when automated, the
This document provides a brief overview of how the analysis proves less than desirable since schedules are often not
closely matched to requirements to provide adequate data points
NICE Workforce Management System solves the
for analysis. This is usually due to constrained work rules, limited
challenge of understanding the adjustments needed headcount, surplus headcount, scheduling inflexibility. In other
to reduce ErlangC requirements when skill-based words, when service level or ASA (actual, forecasted, or simulated)
routing is used in a contact center. deviates from an expected service level or ASA based on ErlangC
requirements, it is difficult to determine how much of the deviation is
caused by multi-skill efficiency versus poor schedule fit.
Two of the basic assumptions of ErlangC are that
employees are singularly skilled and contacts are Therefore, an alternative approach needed to be developed. The
queued to only one group of employees. These algorithm needs to overcome the following challenges (which are
assumptions are made so that the ErlangC algorithm described in more detail below):
can account for the occupancy required to meet a • A Contact Type (CT) is often handled by more than skill group.
service objective. In a skill-based routing environment, • A skill group (SG) often handles more than one CT (this results in
it is possible to run employees at higher occupancy pairs of skill groups and CTs such as SG1/CT1, SG1/CT2, SG1/CT3,
and still achieve service objectives. Therefore, the SG2/CT1, SG2/CT3, etc.).
user needs fewer employees than ErlangC will predict • The CTs that a skill group handles may have different service
in a skill-based routing environment. An adjustment objectives.
is needed to reduce the ErlangC requirements to • Calculating requirements at the CT level produces a lower
account for the higher occupancy of the multiple skill requirement value than summing the calculated requirements
groups and multiskilled employees. NICE refers to this for each skill group / CT pair (this is known as the “pooling”,
adjustment as Multi-skill Efficiency. “economies of scale”, or “law of large numbers” principal).
• A single CT’s multi-skill efficiency is a composite of the efficiencies
from each skill group that handled the CT.

5 6
The Solution
To solve this problem of having more than one service objective for the contacts handled by
a skill group, the Multi-skill Efficiency algorithm uses the most restrictive service objective as
a common objective to temporarily recalculate each CT’s service objective percentage such
NICE’s unique approach to solving these challenges is to calculate partial efficiencies for that the requirements are the same as before the adjustment. For example, using the same
each skill group / CT pair and then combine the partial efficiencies for a composite CT CTs noted above, the service level percentages would be adjusted as follows:
efficiency. The challenge is to determine the values to use for the skill group's contact
volume, AHT and service objective, and how to partition the CT's contact volume and/or
AHT to each skill group (based on complex routing scripts or multiskilled employees) while
maintaining the positive effects of the “pooling” principal.

Different Service Objectives

The basic problem is skill groups do not have contact forecasts, AHT forecasts or service
objectives (these are CT variables). To solve this dilemma, a “temporary” (for MSEff
calculation purposes) skill group contact forecast and AHT forecast can be derived from the
CTs that “use” the skill group. But, all the CTs that “use” a skill group may not have the same
service objective.
Complex Routing Scripts & Multi-Skilled Employees
For example, suppose “Skill Group 1” takes contacts from three different CTs with the
following contact volumes, AHTs and service level objectives: Before proceeding in the discussion of how MSEff is derived, assume the CTs handled
by Skill Group 1 (above) are also handled by other skill groups (due to cross utilization of
employees either through routing rules or skill assignments.) Instead of Skill Group 1 handling
all 100 contacts from CT1, other skill groups will also handle some of the contacts from CT1.
Furthermore, the other skill groups may or may not also handle contacts from CT2 and CT3.
This further complicates the process of calculating MSEff values at a CT level.

When there are multiple skill groups that handled contacts for a CT, simulation must be
used. The MSEff algorithm uses the skill group scheduled open values from the simulation to
determine how much of a CT’s workload was handled by each skill group, and use only that
portion when determining each skill group's contribution toward the CT's overall efficiency.
In other words, skill group simulated scheduled open values are used to “portion” out a CT’s
contacts to the skill groups that service the CT. For example, assume the following:

If the service objectives were identical, the Multi-skill Efficiency algorithm could simply
use the total contacts of 425 and AHT of 306 seconds to calculate total Skill Group 1
requirements. This composite skill group requirement value could then be compared to
the sum of the individual CT requirements to derive the MSEff value. Instead, the service
objectives are different (80%/20secs, 50%/10secs, or 90%/30secs), so requirements for Skill
Group 1 cannot be calculated because only one of the service level objective can be used to
calculate the aggregated skill group requirements.

7 8
The following skill groups handle CT1 Contacts:

The following skill groups handle CT2 Contacts:

Pooling Principal
Efficiency is much greater for small workloads than large workloads. As CTs are added to
The following skill groups handle CT3 Contacts: a skill group, the efficiency gets better. There is a point of diminishing returns, though. For
example, an 80% efficiency might be achieved with two equivalent CTs in a skill group;
70% efficiency may occur with a third CT; and with each additional CT, efficiency will start
leveling off with about 65% efficiency with CT4, 62% with CT5, etc.

The effect described above poses a problem for calculating multi-skill efficiency. Notice
that the process of “portioning” out a CT’s contacts to various skill groups “reduces” the
contact volume of the CT on a “per skill group” basis. This reduces the multi-skill efficiency
that should be seen at the CT level.

To deal with this effect, the NICE MSEff algorithm rescales the number of contacts per skill
group so that the total contacts for the skill group is equal to the total of contacts for the
If Skill Group 1 provides 30% of the scheduled open for CT1, then it should handle 30% of the CTs. In the example above, the total contacts for CT1 and CT2 is 350 contacts. If Skill Group
contacts. Applying this logic to all the skill groups and CTs, the skill groups' “forecasts” would 1’s total contacts of 96.67 is rescaled to a total of 350, the following results occur:
be as follows, which now provides the basis for calculating skill group requirements:
Rescaled Split for SG1

9 10
Rescaled Split for SG2

Benefits

There are several benefits to the NICE

approach to determining efficiency gains:

• During the first pass of schedule

generation, the system creates schedules
to match requirements that have been
adjusted with an initial estimate of Multi-
Rescaled Split for SG3
skill Efficiency adjusted requirements
(rather than creating schedules against
unadjusted ErlangC requirements).
This process will improve schedule “fit”
with fewer passes by making an initial
assessment of how each CT’s contacts will
be portioned to each skill group.

• The resulting Multi-skill Efficiency values

per interval not artificially impacted by
constrained work rules, limited headcount
The AHTs and service objectives for the skill groups can now be recalculated and a partial
and scheduling inflexibility. Multi-skill
multi-skill efficiency for each SG / CT pair can be calculated. (The details of this phase of the
Efficiency is calculated consistently and is
algorithm are omitted for proprietary reasons.)
independent of these factors.

• If there is a single CT handled by one or

more skill groups and those skill groups
Composite MSEff do not handle any other CT, the Multi-skill
One more problem manifests itself in the final calculation of MSEff. If efficiencies for each Efficiency value will be 1.0.
skill group / CT pair are combined into a single CT efficiency by assuming every CT gains
the same amount of efficiency as the other CTs handled by a skill group, the resulting
composite efficiency often leads to an occupancy projection of greater than 100%. Although
a mathematical possibility, occupancy greater than 100% is not achievable in reality.

To account for reality, the NICE MSEff algorithm assumes that the occupancy of a skill group
is the same for each CT the skill group handles. Therefore, a composite efficiency value can
be calculated that by necessity will cause less than 100% occupancy. The MSEff algorithm
calculates a composite occupancy and resulting requirements for each CT. This “composite”
requirement value is compared to the sum of the individual CT requirements to determine the
MSEff value seen on the Adjust Forecast screen. (The details of this phase of the algorithm are
omitted for proprietary reasons.)

11 12