✦ScienexySCIENCE KNOWLEDGE LIBRARY
topic · SCI-CS-004-012

Data-Driven Decision Making

Data-Driven Decision Making is an approach to making decisions by systematically collecting, analyzing, interpreting, and applying relevant data alongside domain knowledge, experience, and contextual judgment. It uses descriptive, diagnostic, predictive, and prescriptive analytics to understand situations, evaluate alternatives, estimate possible outcomes, and support informed actions. Its effectiveness depends on data quality, appropriate analytical methods, clear objectives, and responsible interpretation.

NameEquationMeaningVariablesConditionsLimitation

Expected value of an uncertain outcome

E(X) = Σ(i=1 to n) p_i x_i

Calculates the probability-weighted average of possible outcomes. It is useful for comparing alternatives when their possible outcomes and probabilities can be reasonably estimated.

E(X)

Expected value of the outcome, expressed in the same unit as x_i.

P i

Probability of outcome i; each probability is between 0 and 1 and all probabilities sum to 1.

X i

Value of outcome i.

N

Total number of possible outcomes.

The outcomes must be defined and the probabilities must form a valid probability distribution.

Expected value does not describe the full distribution of risk. Two alternatives with the same expected value may have very different variability and worst-case consequences.

Net benefit

NB = B - C

Represents the difference between the estimated benefits and costs of a decision over a specified scope and time period.

NB

Net benefit, expressed in a consistent unit such as currency.

B

Estimated total benefit.

C

Estimated total cost.

Benefits and costs must be defined using a consistent valuation approach and time horizon.

Nonfinancial effects, uncertainty, distributional consequences, and differences in the timing of costs and benefits may require additional analysis.

Explanation

Overview

Data-driven decision making uses evidence from data to improve the quality and consistency of decisions. Rather than relying exclusively on intuition or assumptions, decision-makers examine relevant observations, statistical findings, analytical models, and measurable outcomes. Data does not make decisions independently; people and organizations must interpret evidence in context, consider uncertainty, and evaluate the consequences of available choices.

Decision Making Process

StageExplanationExample

Define the decision

Clarify what decision must be made, who is responsible, what outcome is desired, and which constraints apply.

A school must decide how to allocate additional academic support across subjects.

Establish objectives and criteria

Define what a successful decision means and identify the criteria that will be used to compare alternatives.

Criteria may include learning improvement, cost, accessibility, and available teaching resources.

Identify and collect relevant data

Determine what information is required and obtain it from reliable, appropriate sources.

Collect assessment results, attendance summaries, course information, and relevant feedback under suitable privacy safeguards.

Assess data quality and context

Check whether the data is accurate, sufficiently complete, representative, current, and appropriate for the intended decision.

Determine whether assessment results are comparable across subjects and whether missing records affect particular groups.

Analyze the evidence

Apply descriptive, statistical, predictive, or causal methods as appropriate to understand the situation.

Compare learning outcomes across subjects and investigate whether attendance patterns are associated with performance.

Develop and compare alternatives

Use analytical findings and domain knowledge to identify feasible actions and compare their likely benefits, costs, and risks.

Compare additional tutoring, revised teaching schedules, and supplementary learning materials.

Make and implement the decision

Select an appropriate option, document the reasoning, and put the decision into practice.

Introduce a support program in selected subjects.

Monitor outcomes

Measure the actual results and compare them with the objectives and expected outcomes.

Review subsequent assessment results, participation, and resource use.

Review and improve

Use new evidence to determine whether the decision should be continued, modified, expanded, or discontinued.

Adjust the support program if its observed outcomes do not justify its cost.

Analytics Used In Decision Making

TypeQuestionExplanationExampleCaution

Descriptive analytics

What happened?

Summarizes historical and current data using counts, averages, distributions, dashboards, and reports.

Calculating monthly sales and identifying which products generated the highest revenue.

Diagnostic analytics

Why did it happen?

Investigates possible explanations for observed patterns through comparisons, segmentation, statistical analysis, and investigation of contributing factors.

Examining whether reduced sales are associated with seasonal demand, product availability, or changes in customer behavior.

Observational relationships can suggest explanations but do not automatically establish causality.

Predictive analytics

What is likely to happen?

Uses statistical or machine-learning models to estimate future or unknown outcomes.

Forecasting next month's product demand using historical transactions and relevant external variables.

Prescriptive analytics

What action should be taken?

Uses optimization, simulation, decision analysis, and constraints to compare possible actions and identify suitable recommendations.

Determining inventory quantities that balance expected demand, storage capacity, and procurement costs.

Data And Decision Types

TypeExplanationExample

Strategic decisions

Long-term decisions that influence the direction and major objectives of an organization.

Selecting a new market or investing in a major technology platform.

Tactical decisions

Medium-term decisions about how resources and activities should support strategic objectives.

Planning regional marketing campaigns or allocating departmental budgets.

Operational decisions

Routine or short-term decisions concerning daily activities and processes.

Adjusting daily staffing or identifying products requiring replenishment.

Quantitative Decision Concepts

ConceptExplanationFormulaVariables

Decision criteria

The measurable or qualitative standards used to evaluate alternatives, such as cost, accuracy, time, safety, and expected benefit.

Expected value

A probability-weighted average of possible outcomes. It can help compare uncertain alternatives when probabilities and outcome values are reasonably specified.

E(X) = Σ p_i x_i

E(X)

Expected value of the uncertain outcome.

P i

Probability of outcome i, with probabilities summing to 1.

X i

Value associated with outcome i.

Cost-benefit analysis

Compares expected benefits with the financial and nonfinancial costs of an action. The assessment should include relevant assumptions, time horizons, and uncertainty.

Risk analysis

Examines possible adverse outcomes, their likelihood, their consequences, and the available ways to reduce or manage them.

Sensitivity analysis

Investigates how a decision or model output changes when important assumptions, parameters, or input values are varied.

Optimization

Finds a solution that maximizes or minimizes a defined objective while satisfying specified constraints.

Role Of Human Judgment

Explanation

Data provides evidence but cannot fully represent every ethical, social, contextual, or practical consideration. Decision-makers must interpret results using domain expertise, understand uncertainty, and take responsibility for the consequences of decisions.

Important Considerations

  • A dataset may not represent all affected people or situations.

  • A model may reproduce historical inequalities or measurement biases.

  • An apparently optimal numerical solution may violate ethical or practical requirements.

  • A prediction does not automatically determine the best action.

  • Stakeholder knowledge may reveal important information absent from the dataset.

Common Challenges

ChallengeExplanation

Poor data quality

Incomplete, inaccurate, outdated, or inconsistent data can lead to misleading conclusions.

Selection and sampling bias

The collected observations may not represent the population or situations to which the decision will be applied.

Correlation and causation confusion

Two variables may move together without one causing the other, so causal claims require suitable evidence and study design.

Overreliance on models

Decision-makers may treat predictions as unquestionable facts instead of estimates affected by assumptions and uncertainty.

Misleading metrics

A single performance measure may hide important errors, subgroup differences, or operational consequences.

Privacy and ethical concerns

Using personal or sensitive information without appropriate safeguards can harm individuals and undermine trust.

Organizational resistance

People may be reluctant to adopt evidence-based procedures because of existing practices, limited analytical literacy, or concerns about changes in responsibility.

Worked Example

Problem

A retailer must decide how much stock of a particular product to order for the next month.

Data

  • Historical sales by date.

  • Current inventory.

  • Supplier delivery time.

  • Product price and procurement cost.

  • Seasonal and promotional information.

  • Storage capacity and budget constraints.

Analysis

  • Use descriptive analysis to summarize previous sales.

  • Use forecasting methods to estimate future demand.

  • Estimate uncertainty in the forecast rather than relying only on a single predicted value.

  • Compare possible order quantities under budget, storage, and supplier constraints.

  • Evaluate the consequences of excess inventory and stock shortages.

Decision

Select an order quantity that balances expected demand, financial constraints, delivery time, and the consequences of uncertainty.

Evaluation

After the month ends, compare actual sales and inventory outcomes with the forecast and decision assumptions.

Interpretation

The final decision combines data analysis with practical constraints. The forecast informs the decision, but it does not independently determine the correct order quantity.

Example or application

A worked example has not yet been added.

Definition

Formal

Data-Driven Decision Making is a systematic decision-making approach in which relevant data is collected, analyzed, interpreted, and applied alongside domain knowledge and contextual judgment to evaluate alternatives and guide actions.

Conceptual

It means using evidence from data to understand a situation, compare possible choices, and evaluate the results of decisions.

Key Distinction

Data-driven decision making is the application of analytical evidence to actual choices, whereas data analysis is the process of examining data and may not itself result in a decision.

Worked Example

Problem

A project has two possible outcomes. A successful outcome provides a benefit of 100 units with probability 0.7, while an unsuccessful outcome provides a benefit of 20 units with probability 0.3. Calculate the expected benefit.

Calculation

  • E(X) = (0.7 × 100) + (0.3 × 20).

  • E(X) = 70 + 6.

  • E(X) = 76 units.

Interpretation

The expected benefit is 76 units. This is a probability-weighted average, not a guarantee that the project will produce exactly 76 units.

Purpose

  • To improve decisions by using relevant observations and analytical evidence.

  • To identify patterns and potential causes behind important outcomes.

  • To compare alternative actions using measurable criteria.

  • To estimate future conditions and associated uncertainty.

  • To allocate resources more effectively.

  • To evaluate whether implemented decisions produce their intended outcomes.

Importance

  • It can reduce dependence on unsupported assumptions and intuition alone.

  • It improves transparency by connecting decisions with documented evidence.

  • It helps organizations identify opportunities, risks, and inefficiencies.

  • It supports measurable evaluation of policies, products, and operational changes.

  • It connects descriptive analysis, predictive modelling, optimization, and practical action.

  • It encourages continuous improvement through feedback from actual outcomes.

Prerequisites

  • Basic statistics and probability.

  • Understanding of data collection and data quality.

  • Knowledge of descriptive and predictive analytics.

  • Basic understanding of decision criteria and optimization.

  • Awareness of privacy, fairness, and responsible data use.

Related Topics

  • Introduction to Data Science

  • Data Science Applications

  • Data Analysis

  • Statistical Inference

  • Predictive Analytics

  • Prescriptive Analytics

  • Decision Theory

  • Optimization

  • Data Visualization

  • Data Governance

Sources and verification

Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.

  • Provost and Fawcett, Data Science for Business
  • Thomas H. Davenport and Jeanne G. Harris, Competing on Analytics
  • NIST, Artificial Intelligence Risk Management Framework
  • OECD, Data-Driven Innovation

Suggest a correction or source · More in Data Science