Descriptive, Predictive & Prescriptive Analytics

Module: Module 1 — Management FoundationsTheme: Data & Analytics FundamentalsUse case: Managers commissioning, consuming, or challenging analytics work

Executive Readiness Lens

Analytics is only useful when it changes a decision. That sounds obvious, but most organisations spend most of their analytical effort on reporting what happened and very little on deciding what to do next. The practical ladder is simple: descriptive shows what happened, diagnostic explains why, predictive estimates what will happen, and prescriptive recommends what should happen. The business question is not which method is fashionable; it is which method is worth the cost for this decision.

Canonical Grounding

• Davenport & Harris: analytics maturity rises from descriptive to prescriptive, and value rises with it.
• Provost & Fawcett: data science and decision-making are separate disciplines that must be connected intentionally.
• McKinsey research: data-driven firms outperform peers because they turn information into action faster and more consistently.

Working Heuristic

Use this decision-first ladder before starting any analytics work:

1. Decision question first. What specific decision will change if the model is right?
2. Descriptive baseline. What happened, and is the data trustworthy enough to use?
3. Diagnostic layer. Why did it happen, and what variables are actually driving it?
4. Predictive layer. What is likely to happen next if current conditions persist?
5. Prescriptive layer. What action should we take, at what speed, and with what override logic?

If a project cannot answer step 1, it is probably a reporting exercise.

Corporate Reality Check

Analytics waste usually appears in familiar forms: dashboards without owners, models with no operating decision, and automation attempts before the descriptive layer is stable. The danger is not that the models are wrong; it is that they are irrelevant.

What to do instead:

• Tie each analytics project to one named decision owner.
• Define the cost of false positives and false negatives before modelling.
• Move to prescriptive analytics only when the process is fast enough or repetitive enough to justify automation.

Case Lens

Zara is a clean example of prescriptive analytics in action. Stores send daily sales data to headquarters. Descriptive analytics shows what sold. Diagnostic analytics shows why. Predictive models forecast demand. Prescriptive algorithms then generate replenishment orders and design signals within the operating cycle.

Lesson: analytics creates value when the operating model is designed to use the output quickly.

Full Case Walkthrough

Click here to read full case study

1. Case context

Fashion demand is volatile, markdown risk is high, and forecast error is expensive.

2. Decision trigger

Zara chose speed and responsiveness as the order winner, which required analytics at every layer.

3. Timeline

Phase	What happened	Management relevance
Descriptive	Store sales data flowed to HQ	Visibility improved
Diagnostic	Drivers of sell-through were analysed	Root causes surfaced
Predictive	SKU-level demand was forecast	Replenishment got faster
Prescriptive	Orders and design signals were generated	Human review focused on exceptions

4. Options considered

Zara could have stayed with long production runs, used a hybrid model, or adopted a highly responsive model. It chose responsiveness for core categories.

5. Execution moves

Key moves included real-time data flows, predictive demand models, prescriptive replenishment, and fast feedback loops to design teams.

6. Outcomes and evidence

Zara reduced markdowns and improved inventory turns by linking analytics directly to operating decisions.

7. Transfer to practice

• Start with decision questions.
• Match analytics level to decision value.
• Build data foundations before automation.

8. What we know vs what is inferred

Category	Statement type
What we know	Zara’s responsive operating model is well documented.
What is inferred	Each analytics layer improves the quality of the layer above it.

Monday Morning Playbook

30-minute prep

1. List your five most important decisions.
2. Label each as descriptive, diagnostic, predictive, or prescriptive.
3. Identify one dashboard that lacks a decision owner.

60-minute analytics review

1. Review the decision map for the top five decisions.
2. Inspect the most important predictive model for drift.
3. Approve or kill one analytics project using the decision-question test.

7-day follow-through

1. Require a decision question in every new analytics proposal.
2. Remove one dashboard that no one uses.
3. Ask one business unit to map its top decisions to the analytics ladder.

Role-Based Activation

• People Manager: ask what decision changes before any analysis starts.
• Functional Leader: cut analytics projects that lack decision linkage.
• BU Leader: review analytics maturity by decision category.
• Strategy/Founder Office: track analytics decision-quality improvement in the plan.

KPI and Evidence Block

Metric type	Suggested metric	Review cadence
Leading	Top decisions with a named predictive/prescriptive input	Monthly
Leading	Projects with a clear decision question and error cost	Monthly
Lagging	Outcome improvement from analytically informed decisions	Quarterly
Lagging	Accuracy trend for key predictive models	Monthly
Risk	Dashboards with no linked recurrent decision	Monthly

Tools Pack

Analytics Decision Audit

Decision	Current level	Target level	What changes	Investment required
Decision 1
Decision 2
Decision 3
Decision 4
Decision 5

Analytics Project Frame

1. Decision.
2. Owner.
3. Timeline.
4. Error cost.
5. Baseline.
6. Success metric.

Practice MCQ

A BI dashboard shows sales fell 18% in Q3 versus Q2. Which analytics type is this?

• A. Predictive
• B. Prescriptive
• C. Descriptive
• D. Diagnostic

Four analytics levels

Descriptive, diagnostic, predictive, prescriptive.

Click the card to flip

1 of 4

Continue Learning

📄HBR · ReadAnalytics 3.0Davenport’s case for embedding analytics into operations and products, not just reporting.📖BookCompeting on AnalyticsA foundational book on building analytics capability into the organisation.🌐CourseGoogle Data Analytics CertificateA practical upskilling path for managers who need to work with analysts more effectively.📊Investor RelationsInditex Investor RelationsUseful for understanding how a prescriptive operating model shows up in reported performance.

References

• Davenport, T. H., & Harris, J. G. (2007). Competing on analytics.
• Provost, F., & Fawcett, T. (2013). Data science for business.
• McKinsey Global Institute. Data-driven enterprise research.
• Inditex investor communications.

Analytics earns its keep only when it changes a decision — anything else is decision theatre.