Unit 8.6
Prescriptive Analytics: How to Guide Business Decisions
Learning Objectives
By the end of this session, you will be able to:
- ✅ Define Prescriptive Analytics and the key question it answers.
- ✅ Explain how prescriptive analytics builds on predictive analytics.
- ✅ Describe the main techniques used, such as optimization and simulation.
- ✅ Provide business examples, including dynamic pricing and next best action.
The Analytics Journey 📊
🔍 Descriptive
"What happened?"
Past-focused. Summarizes historical data (e.g., dashboards, reports).
🔮 Predictive
"What will happen?"
Future-focused. Uses statistical models and forecasting (e.g., sales forecast).
🎯 Prescriptive
"What should we do?"
Action-focused. Recommends decisions to achieve goals (e.g., optimal price).
Interactive: Classify the Analytics Type
Click a scenario chip to select it, then click the correct analytics category bin.
🔍 Descriptive
🔮 Predictive
🎯 Prescriptive
What is Prescriptive Analytics?
Prescriptive Analytics is the most advanced form of analytics. It goes beyond predicting the future by recommending specific actions to achieve a desired outcome.
It's about moving from insight to automated, guided action.
It answers the fundamental question: "What should we do about it?"
🎬 Watch: What Prescriptive Analytics Is
How It Works: From Prediction to Action
Prescriptive analytics combines predictions with business logic to find the best path forward.
Key Technique: Optimization
Optimization: The core of prescriptive analytics. It is the process of finding the best possible solution from a set of alternatives, given a specific set of constraints.
- Goal: Maximize profit, minimize cost, increase efficiency.
- Constraints: Budget, time, resources, production capacity.
- Example: What is the optimal production schedule to meet demand while minimizing manufacturing costs?
🎬 Watch: Linear Programming
Interactive: Production Optimization Calculator
Adjust production quantities within the constraints to maximize profit.
Profit: 200 NPR/unit
Labor: 2 hrs/unit | Material: 3 kg/unit
Profit: 300 NPR/unit
Labor: 4 hrs/unit | Material: 2 kg/unit
Max Labor: 40 hrs
Max Material: 36 kg
🧮 Worked Example: How an Optimiser Finds the Best Plan
Illustrative: a Bhaktapur workshop makes chairs (profit NPR 500) and tables (NPR 800). Each week it has 40 carpentry hours and 38 units of wood. A chair needs 2 hrs + 3 wood; a table needs 5 hrs + 2 wood.
| Plan (chairs, tables) | Hours used | Wood used | Profit (NPR) |
|---|---|---|---|
| 0, 8 | 40 ✓ | 16 ✓ | 6,400 |
| 12, 0 | 24 ✓ | 36 ✓ | 6,000 |
| 10, 4 | 40 ✓ | 38 ✓ | 8,200 |
The best plan usually sits where the constraints meet (here, both hours and wood fully used). Software checks millions of such points in seconds.
Key Technique: Simulation
Simulation: Creating a computer model of a real-world system to test different scenarios and actions in a virtual, risk-free environment.
- Allows businesses to ask "What if...?" questions.
- Example: Simulating a supply chain to see how a factory closure in one country would impact delivery times worldwide.
- Helps evaluate the outcome of recommended actions before implementation.
Interactive: Supply Chain What-If Simulator
Toggle disruptions to simulate their impact on global delivery time.
🎲 Simulation: Run the Future 10,000 Times
Monte Carlo simulation runs a model many times with random but realistic inputs, then looks at the range of results.
- Illustrative: a trekking company plans next season. Visitor numbers, fuel prices and flight cancellations are uncertain.
- The model runs 10,000 seasons, each with different random values.
- Output: "In 90% of runs, profit is positive; in 5%, we lose money if we lease 3 extra jeeps."
Other Key Techniques
Rule-Based Systems
Uses a set of predefined "IF-THEN" rules to automate decisions.
Example: "IF a customer's cart value > 5000 NPR, THEN offer free shipping."
A/B Testing
A randomized experiment comparing two variants (A and B) to see which performs better.
Example: Showing two different website layouts to users to see which one leads to more sign-ups.
These techniques help translate analytical insights into concrete, testable business actions.
Interactive: IF-THEN Rule Builder
Build a prescriptive rule by choosing condition values, then see what action is recommended.
Example 1: Dynamic Pricing ⚡
Scenario: A ride-sharing app needs to set prices.
- Predicts: Future ride demand and driver supply in a specific area.
- Constraints: Maximize completed rides, ensure driver availability.
- Prescribes Action: Recommends an optimal price ("surge pricing") to balance supply and demand.
- Nepal Context: Think of Pathao or inDrive adjusting fares during peak traffic hours in Kathmandu or during a festival.
Interactive: Surge Pricing Simulator
Adjust ride demand and driver supply to see how prescriptive analytics sets the optimal surge fare.
Example 2: Next Best Action (NBA)
Scenario: A telecom company wants to retain a customer.
- Analyzes: Customer profile, call history, data usage, and recent complaints.
- Predicts: Likelihood of the customer switching to a competitor.
- Prescribes Action: Recommends the "next best action" for the customer service agent.
- Nepal Context: Ncell or Nepal Telecom could use this to proactively offer a specific data pack or a loyalty discount to a high-value customer at risk of leaving.
Example 3: Supply Chain Optimization
Scenario: An e-commerce company planning deliveries.
- Predicts: Traffic conditions, weather, delivery time windows.
- Constraints: Fuel costs, driver hours, vehicle capacity.
- Prescribes Action: Recommends the most efficient delivery route for each truck.
- Nepal Context: Daraz or Sastodeal optimizing delivery routes across the valley during the Dashain shopping season to ensure timely delivery and minimize costs.
🚚 Real Case: UPS Plans Every Driver's Route
Parcel company UPS built ORION (On-Road Integrated Optimization and Navigation), rolled out across the US by about 2016.
- Each morning it calculates the best stop order for each driver, using deliveries, time windows and road data.
- A driver with 120 stops has more possible routes than anyone could check by hand; ORION searches them.
- UPS reports large savings in miles driven and fuel each year.
- Famous detail: routes avoid left turns (UPS drives on the right) where possible, since waiting to turn wastes time and fuel.
A UPS Delivery Truck
🎬 Watch: Why UPS Trucks Never Turn Left
🧪 Quick Activity: Which Technique?
Optimization, simulation, rules or A/B testing?
- A college wants the exam timetable with the fewest student clashes. → Optimization
- A hospital tests "what if patient numbers double during an outbreak?" → Simulation
- "IF a card is used in two countries within an hour, THEN block it." → Rule-based
- A news site shows two headlines to different readers to see which gets more clicks. → A/B testing
Discussion & Ethical Considerations
- What is the relationship between predictive and prescriptive analytics? How do they depend on each other?
- Google Maps recommending the fastest route is a prescriptive action. What data does it use? What are the constraints?
- What are the potential ethical concerns of using prescriptive analytics for dynamic pricing? Could it lead to unfairness?
⚖️ Real Case: When the Optimal Price Is Wrong
On 15 December 2014, a gunman took hostages in a café in central Sydney, Australia.
- People rushed to leave the area. Uber's algorithm saw demand spike and raised prices automatically (surge pricing).
- There was public anger at "profiting from a crisis".
- Uber made rides free from the area and refunded surge charges.
Summary: Key Takeaways
- 🎯 Prescriptive analytics answers "What should we do?" by recommending specific actions.
- 🔗 It builds on predictive analytics by adding optimization, simulation, and business rules.
- 💡 Key applications include dynamic pricing, next best action, and supply chain optimization.
- 🤖 It represents a major shift from supporting human decisions to guiding and automating them.
Exam Practice 📝
Short answer
- What is prescriptive analytics?
- What question does prescriptive analytics answer?
- Name two techniques used in prescriptive analytics.
Exam Practice 📝 (continued)
Long answer
- Explain prescriptive analytics and how it differs from descriptive and predictive analytics, with examples.
- Discuss how optimization and simulation help managers make decisions, using an example.
Revise the answers in the full note.
Image & Video Credits
Images
- UPS Truck in Fresno, California: DoulosBen, CC BY 4.0, via Wikimedia Commons
- Monte Carlo histogram: course author (original), CC BY 4.0
Videos (YouTube)
- Decisive Data, "What is Prescriptive Analytics? – Data Science Wednesday" (51–140 s)
- Don't Memorise, "Linear Programming – Introduction" (60–150 s)
- CNN Business, "Why UPS trucks never turn left" (8–98 s)
Thank You!
Questions?
Next Topic: Unit 8.7: Data Presentation Tools: Dashboards and Reports | IT 233