SKIPPER ASchool and Advancement Optimization by Pay Grade

1
SKIPPER A-School and Advancement Optimization by
Pay Grade
Eighth Annual Navy Workforce Conference May 5 -
7, 2008

Dr. Chariya Punyanitya (CSC), Sanjay Nayar (CSC),
Dr. Colin Osterman (NPRST), Angela Cho (CSC)
2
Agenda

• Interest in optimizing Accessions by Pay Grade
  (PG)
• SKIPPER Brief Overview
• Approaches Accession Optimization by PG
• Formal Linear Optimization
• Iterative - based on historic Advancement by PG
  profile
• Iterative - based on optimum Advancement by PG
• Conclusions

3
Interest in Accession Optimization by PG

• Current emphasis on Accession Optimization by LOS
  ignores PG
• Official EPA is only by PG
• Provides an alternative to managers as a
  comparison against LOS approach

4
Model Overview

• SKIPPER (Skilled Personnel Projection for
  Enlisted Retention)
• Web-based Open-Box Model
• Powerful Scenario and What-If Analysis
  Capability to support Strategic Planning
• Generalized and Expandable Modeling Framework
• Multi-year Inventory Projection
• Personnel Master File-based historical data with
  override capability
• School Input Optimization and Skill Conversion
  Planning
• Advancement, Selective Reenlistment Bonus and
  Rotation Modeling
• Suite of Reporting Tools
• Career Reenlistment Objectives (CREO)
• Scenario Comparison
• Skill Rollup to the All Navy Level

5
Model Overview

• The components can be used in any order
• The system is holistic
• All parts are executed together
• Scenarios become the key to what-if analysis
• Divides areas of responsibility among components
  to minimize inconsistency

6
Overview of the Integrated SKIPPER System
A-School Planning and RG Overall Inventory
management, Targeting EPA
SRB Planning Impact of a change in SRB on
continuation behavior and cost analysis
C-School Planning Strength and
Inventory management for a NEC targeting BA
7
Integrated SKIPPER Scenario Management
Historic Plans and Overrides
8
Current Capabilitiesthat can be leveraged for PG
Optimization

• SKIPPER models and forecasts Inventory
  (multi-dimension including LOS, PG)
• SKIPPER also models and forecasts Advancements
• Users can explore what-if Scenarios Input
  based on over-ridden historic data,
• SKIPPER has Accession planning functionality
• Total School Input Optimization
• Optimize to hit EPA by LOS
• SKIPPER has EPA by PG or LOS

9
Approach 1 - Linear Optimization by PG

• Default will be to target Total EPA
• The PG Weight is assigned a value of 1 for all PG
• When targeting EPA by PG
• PG Weight is assigned the user input values for
  associated PG

10
BackgroundLinear Optimization by LOS

• Uses Solver optimization engine
• Default is to target Total EPA
• The LOS Weight is assigned a value of 1 for all
  LOS
• When targeting EPA by LOS
• LOS Weight is assigned the user input values for
  associated LOS.
• Targets EPA LOS peak where user inputs a LOS peak
  value peak and discount rate DiscountRate
• The LOSWeight is then derived from the following
  formular
• LOSWeight(LOS)

11
BackgroundLinear Optimization by LOS

• Minimize Z
• Subject to Constraints
• EPA
• A-School Limit
• Freeze School Inputs, Minimum School Inputs and
  Maximum Deviations

12
BackgroundA-School Optimization by LOS
Provides users with long range Accession numbers
needed to meet requirements
13
Full EPA OptimizationA-School Plan
14
Full EPA OptimizationPG Manning
15
LOS Zone A OptimizationA-School Plan
16
LOS Zone A OptimizationPG Manning
17
Approach 1 Formal Linear Optimization by PG

• Minimize Z
• Subject to Constraints (same as before)
• EPA
• A-School Limit
• Freeze School Inputs, Minimum School Inputs and
  Maximum Deviations

18
So far so good, but
19
what about Advancements Distribution?

• The Advancements Distribution
• is the multi year advancement distribution
  starting from FY i to FY j
• Future Advancements are not known so a fixed
  historic distribution must be used

20
what about Advancements Distribution?

• Problems with using a fixed Advancements
  Distribution
• Advancements are based on the underlying
  inventory
• Inventory changes as a result of
• Continuation
• Previous Advancements under various constraints
• Gains
• During a full SKIPPER projection the actual
  Advancements will be different from historic

21
Approach 2 Iterative based on Historic
Advancement PG distribution

• Current Functionality
• A-School optimization by LOS produces total
  A-School gains that optimize total manning gap
  for every projected FY
• PG dimension of the Gains are dictated by
  historic Advancement by LOS and PG throughout the
  projected FY
• Iterative Approach
• SKIPPER determines Total A-School Gains that will
  bring projected Inventory closer to EPA by LOS
  and PG
• Uses historic Gains Distribution to get all
  dimensions for the gains
• The PG dimension of Projected Advancements is
  still based on historic values

22
Approach 2 Full ManningA-School Plan
23
Approach 2 Increased GainsA-School Plan
24
Approach 2 Full Manning vs. Increased GainsPG
Manning
25
Approach 3 Iterative PG Optimization
independent of Historic Advancement

• Second iterative PG Optimization
• SKIPPER determines Total A-School Gains by PG
  that will bring projected Inventory closer to EPA
  by LOS and PG
• School Gains Distribution override
• SKIPPER also computes the best fit Predicted
  Advancement by PG
• Starting with a 5 year average historic
  advancements distribution

26
Approach 3 Override School Gains and
Advancement Exam DistributionsA-School Plan
27
Approach 3 Override School Gains and 5 year
Average Advancement Exam DistributionsPG Manning
28
Approach 3 Override School Gains and
Advancement Exam DistributionsPG Manning
29
School Gains Distribution Overrides
30
Advancements Examined DistributionHistoric vs. 5
Yr Average Override
31
Advancements Examined Distribution5 Year Average
vs. Modified 5 Year Average
32
ATISA By PGFull EPA Optimization vs. Approach 3
33
Conclusions Next Steps

• PG Optimization is worth pursuing focusing on
• Accessions Distribution
• Advancements Distribution
• An iterative approach versus Linear Optimization
• Utilizes the power of the existing full model
• Takes into account the changing nature of
  advancements under various constraints
• Can be combined with impending High-Year Tenure
  functionality
• Work with N104 to ensure that modified
  advancement zones are acceptable

34
Contact Information

• CSC
• Navy Personnel Planning and Policy Analysis
  Group
• Federal Sector Defense Group
• Dr. Chariya Punyanitya (301.538.2935), Sanjay
  Nayar (703.461.2075), Angela Cho (858.450.1494)
• cpunyanitya, snayar, acho3_at_csc.com
• NPRST
• Dr. Colin Osterman (901.874.4643)
• colin.j.osterman_at_navy.mil