Prescriptive Analytics: The "What Should We Do?

" Approach
Prescriptive analytics is the branch of analytics that goes beyond describing what happened
(descriptive) and predicting what might happen (predictive) to recommend actions that will lead to a
desired outcome. It essentially answers the question: "What should we do?"

Prescriptive analytics utilizes optimization, simulation, and other advanced techniques to identify the
best course of action among various possibilities, considering constraints and objectives. It's the
most sophisticated level of analytics, building upon the insights gained from descriptive and
predictive analytics.

Key characteristics of prescriptive analytics:

 Focus on optimization: Finding the best solution given a set of constraints and objectives.

 Action-oriented: Providing specific recommendations for decision-making.

 Forward-looking: Concerned with future outcomes and how to influence them.

 Data-intensive: Requires large amounts of data for accurate modeling and analysis.

 Technologically advanced: Employs sophisticated algorithms and computational methods.

Comparative Analysis of Prescriptive Analytics Techniques

Technique Description Best Use Cases Advantages Limitations

- Simple and
A mathematical fast for large-
- Resource allocation
Linear technique for optimizing scale
- Production planning - Limited to linear
Programming a linear objective problems
- Transportation relationships
(LP) function, subject to - Well-
problems
linear constraints. developed
solvers

- Suitable for
A variant of linear
discrete
Integer programming where - Scheduling
decision - More complex and
Programming some or all decision - Facility location
problems slower than LP
(IP) variables must be - Capital budgeting
- Finds exact
integers.
solutions

- Can model
- Portfolio
Optimization where the real-world
Nonlinear optimization - Difficult to solve
objective function or complexity
Programming - Engineering design - Prone to local
constraints are - Supports
(NLP) - Supply chain optima
nonlinear. multiple
management
objectives

Mixed Integer Combines linear and - Supply chain - More flexible - Computationally
Technique Description Best Use Cases Advantages Limitations

than LP or IP
integer programming, optimization
- Models real-
Programming allowing both - Workforce planning
world intensive
(MIP) continuous and integer - Logistics
scenarios
variables. management
effectively

- Shortest path
- Efficient for
problems
large-scale - Limited to
Optimizes problems - Distribution
Network problems problems that can
modeled as networks network design
Optimization - Provides be modeled as
(nodes and edges). -
clear visual networks
Telecommunications
solutions
routing

- Suitable for
Solves problems by - Requires large
- Inventory multi-stage
Dynamic breaking them down memory for
management problems
Programming into simpler complex problems
- Pricing strategies - Finds
(DP) subproblems and solving - Time-consuming
- Project scheduling optimal
each once. for large datasets
solutions

Approximate methods - Finds good

for solving complex solutions
- Vehicle routing - No guarantee of
optimization problems, quickly
- Job scheduling optimality
Metaheuristics including genetic - Effective for
- Complex - Results can vary
algorithms, simulated large and
engineering design between runs
annealing, and particle complex
swarm optimization. problems