Prescriptive Analytics

Most companies have plenty of data.

They know:

• what happened

• where performance changed

• which customers bought

• how much inventory moved

• how many deliveries were made

But there is a more difficult question:

What should we do differently because of what the data tells us?

That is where many data initiatives stop short.

A dashboard can tell a manager that delivery costs are rising.

A forecast can estimate tomorrow’s package volume.

But neither necessarily tells the organization which route each driver should take tomorrow morning.

That requires a different form of analytics.

A useful way to think about analytics is as a progression:

What happened?

Why did it happen?

What is likely to happen?

Given what we know, what should we do?

Prescriptive analytics goes beyond describing or predicting an outcome. It combines data, models, constraints and objectives to recommend a course of action.

This distinction is important.

A predictive model might tell you:

“Traffic will be heavy on Route A.”

Prescriptive analytics asks:

“Given traffic, delivery commitments, vehicle capacity and hundreds of other constraints, what route should the driver take?”

That is the difference between using data to understand a business and using data to make a business decision

Prescriptive analytics is essentially an optimization problem.

You have:

What are we trying to improve?

For example:

Minimize delivery miles

What cannot be violated?

For example:

• delivery commitments

• pickup requirements

• vehicle capacity

• driver schedules

• road restrictions

What can actually be changed?

For example:

The sequence in which deliveries are made.

The analytical system searches through possible decisions and identifies the combination that best satisfies the objective without violating the constraints.

The important managerial shift is:

Not “What does the data say?” but “Given the data and constraints, what action produces the best feasible outcome?”

When faced with a data-rich business problem, work through four questions:

Don’t start with the data.

Start with the decision.

What needs to be different because of this analysis?

Cost?

Revenue?

Speed?

Quality?

Customer experience?

Usually, there will be more than one objective.

This is where real-world complexity enters.

A mathematically attractive answer is useless if it violates operational reality.

A model can produce an optimal solution that humans cannot realistically implement.

The final test is therefore:

Can the organization convert the analytical recommendation into action?

Consider United Parcel Service.

Every day, UPS faces a massive operational decision:

How should thousands of drivers move through their delivery routes to serve customers efficiently while meeting service commitments?

This is not simply a routing problem.

UPS has to account for enormous numbers of delivery and pickup locations, package characteristics, traffic and operational constraints.

Objective

Reduce miles driven while maintaining UPS’s service commitments.

↓

Data

Package deliveries and pickups, locations, route history, telematics and mapping information.

↓

Constraints

Customer commitments, delivery requirements, operational rules and the practical realities of driving the routes.

↓

Optimization

Calculate the sequence of stops that provides an efficient feasible route.

↓

Decision

The driver receives a recommended route for that day’s workload.

This is prescriptive analytics in action:

Data → Optimization → Recommendation → Operational decision

And the results were substantial.

UPS reported that its prescriptive analytics system (ORION) generated more than $400 million in annual cost savings and avoidance after the first full year of deployment, alongside reductions of roughly 100 million miles driven and 100,000 metric tonnes of CO₂ emissions annually.

The goal of data-driven management isn’t to produce more reports.

It’s to make better decisions.

The highest-value analytics doesn’t just tell you what happened. It helps determine what you should do next.