Decision Frameworks & Models
Topic 7 of 11 · In Data Driven Decisions · September 9, 2026
Master structured decision-making tools: RAPID, decision trees, weighted scoring, cost of delay, expected value, scenario planning, sensitivity analysis, and MCDA.
Overview
Good decisions do not happen by accident. They happen when the right people use the right framework at the right time. Yet most organizations have no systematic approach to decision-making — each decision is improvised, politically influenced, or left to whoever speaks loudest.
This topic gives you a toolkit of proven decision frameworks that bring structure, rigor, and accountability to any choice. You will learn the RAPID framework for clarifying who decides what, decision trees for mapping options and probabilities, weighted scoring for comparing alternatives, cost of delay for prioritization, expected value for risk-adjusted choices, scenario planning for uncertain futures, sensitivity analysis for testing assumptions, and multi-criteria decision analysis (MCDA) for complex trade-offs. By the end, you will have a framework for every type of decision — from simple operational choices to complex strategic bets.
Before You Start
- Recall expected value and probability basics from Statistical Thinking — you will reuse them directly in the decision trees.
- Be quick at multiplying a probability by a payoff such as 0.30 times £500K without reaching for a calculator.
- Bring two of your own decisions to the lesson: one where you can estimate probabilities and one where you genuinely cannot.
- Rehearse the weighted-shortest-job-first formula (value times time criticality divided by duration) and know why each term matters.
- Name the five RAPID roles from memory and sketch who would hold each in your own team.
Learning Objectives
- Apply the RAPID framework to clarify roles and accountability in decision-making
- Build decision trees to map options, probabilities, and outcomes
- Use weighted scoring to evaluate alternatives against multiple criteria
- Calculate cost of delay to prioritize competing initiatives
- Apply expected value to make risk-adjusted decisions
- Design scenario plans for decisions under deep uncertainty
- Perform sensitivity analysis to test how robust a decision is to changes in assumptions
- Use Multi-Criteria Decision Analysis (MCDA) for complex trade-offs
Prerequisites
Start with DDDM Fundamentals for decision types and maturity. KPI Design & Metric Frameworks helps with metric-driven decisions. Statistical Thinking provides the probabilistic foundation for expected value and scenario planning.
The RAPID Framework: Who Decides What
The RAPID framework (developed by Bain & Company) clarifies who plays which role in every decision. Confusion about roles is the #1 cause of slow, bad decisions.
| Role | Responsibility | Example in a pricing decision |
|---|---|---|
| R — Recommend | Do the analysis and propose a course of action | Data analyst recommends a 10% price increase based on elasticity analysis |
| A — Agree | Must agree before the decision is final (compliance, legal, budget) | Legal confirms no contract violations; finance confirms margin impact |
| P — Perform | Executes the decision once made | Product team implements the price change in the billing system |
| I — Input | Provides input and expertise but does not decide | Customer success team shares feedback on price sensitivity |
| D — Decide | Has the final authority to say yes or no | VP of Product makes the final call on pricing |
Why RAPID works: It eliminates the two most common decision pathologies: (1) "everyone thinks someone else is deciding" (no D) and (2) "too many people think they have veto power" (too many A's).
Decision Trees: Mapping Options and Probabilities
A decision tree maps a decision as a series of branches: choices (square nodes), chance events (circular nodes), and outcomes (endpoints). Each branch has a probability and a payoff.
Example: Launch a new product?
- Decision: Launch vs Don't Launch
- If launch: Market response is Strong (60%, $5M profit) or Weak (40%, -$1M loss)
- If don't launch: Maintain current revenue ($2M)
Expected value calculation:
- Launch: (0.60 × $5M) + (0.40 × -$1M) = $3M - $0.4M = $2.6M
- Don't launch: $2M
Decision trees force you to quantify your beliefs (probabilities) and outcomes (payoffs) — replacing vague feelings with explicit calculations. Even if the numbers are rough, the discipline of writing them down improves decision quality.
Weighted Scoring: Comparing Alternatives Fairly
Weighted scoring evaluates multiple alternatives against multiple criteria, each weighted by importance. It is the most practical framework for everyday business decisions.
Example: Choosing an analytics platform
| Criterion | Weight | Platform A | Platform B | Platform C |
|---|---|---|---|---|
| Ease of use | 30% | 8 | 6 | 9 |
| Cost | 25% | 7 | 9 | 5 |
| Integration | 25% | 9 | 7 | 6 |
| Scalability | 20% | 6 | 8 | 7 |
| Weighted total | 100% | 7.55 | 7.40 | 6.85 |
How to calculate: For each platform, multiply each score by the criterion weight, then sum. Platform A wins with 7.55.
Critical rules:
- Weights must sum to 100% — they represent relative importance
- Score on a consistent scale (1–10 or 1–5)
- Use at least 3 criteria — fewer criteria reduce discriminative power
- Validate weights: if you cannot justify why one criterion is 30% and another is 20%, the weights are arbitrary
Cost of Delay: Prioritizing with Economics
Cost of Delay (CoD) quantifies the economic impact of NOT doing something now. It turns prioritization from politics into economics.
Formula:
CoD = (Value × Time Criticality) / Duration
This is the WSJF (Weighted Shortest Job First) formula from SAFe (Scaled Agile Framework). It prioritizes items that deliver the most value in the shortest time.
Example:
| Initiative | Value ($/month) | Time criticality (1–10) | Duration (months) | WSJF score |
|---|---|---|---|---|
| Fix checkout bug | $50,000 | 10 | 1 | 500,000 |
| Build new dashboard | $30,000 | 5 | 3 | 50,000 |
| Redesign landing page | $40,000 | 3 | 2 | 60,000 |
The checkout bug has the highest WSJF — it delivers the most value per unit of time. This framework is especially powerful for breaking ties in prioritization meetings.
Expected Value: Making Risk-Adjusted Choices
Expected Value (EV) is the probability-weighted average of all possible outcomes. It is the foundation of rational decision-making under uncertainty.
EV = Σ (Probability × Payoff) for all outcomes
Example: Should we invest in a new market?
- Success (30% probability): +$10M profit
- Partial success (50%): +$2M profit
- Failure (20%): -$3M loss
EV = (0.30 × $10M) + (0.50 × $2M) + (0.20 × -$3M) = $3M + $1M - $0.6M = $3.4M
The expected value is positive ($3.4M), so the investment is rational — if you can absorb the $3M downside scenario.
EV is not everything: A $3.4M EV might not be worth it if the company goes bankrupt in the failure scenario. This is where risk aversion enters — and why EV needs to be combined with downside analysis.
Scenario Planning: Preparing for Multiple Futures
Scenario planning is for decisions under deep uncertainty — where you cannot assign reliable probabilities to outcomes. Instead of one forecast, you develop 3–4 plausible future scenarios and test your strategy against each.
Shell's approach (the originators of scenario planning):
- Identify the focal question: "How should we position our energy portfolio for 2030?"
- Identify key drivers: oil prices, renewable adoption rate, regulation, geopolitics
- Create a scenario matrix: two most uncertain drivers as axes (e.g., oil price × regulation stringency)
- Develop 4 scenarios: each quadrant gets a narrative, name, and set of assumptions
- Test your strategy against each scenario: does it survive? What would you need to change?
Example scenarios for a tech company:
- "Green Boom": high regulation, high sustainability demand → invest in green tech
- "Wild West": low regulation, price competition → optimize for cost
- "Fortress": high regulation, low demand → focus on compliance and retention
- "Disruption": low regulation, new entrants → innovate fast or acquire
Sensitivity Analysis: Testing Your Assumptions
Sensitivity analysis asks: "How much would this input need to change to change my decision?" It reveals which assumptions are load-bearing and which are not.
Example: Should we build Feature X?
- Assumption: Feature X will increase conversion by 10%
- Cost: $200K to build
- Revenue impact at 10% lift: $500K/year
- Break-even: the lift must be at least 4% for the investment to pay off
If you are confident the lift will be at least 4%, the decision is robust. If you are unsure whether the lift will be 2% or 15%, the decision is sensitive to this assumption and you should gather more data before committing.
How to perform it:
- Identify the 3–5 key assumptions in your decision
- For each, define a plausible range (optimistic, base, pessimistic)
- Recalculate the outcome for each combination
- If the decision flips within the plausible range, the assumption is critical
Multi-Criteria Decision Analysis (MCDA)
MCDA is the formal name for what weighted scoring does — but extended to handle complex trade-offs with mathematical rigor.
Common MCDA methods:
| Method | How it works | Best for |
|---|---|---|
| AHP (Analytic Hierarchy Process) | Pairwise comparison of criteria, then alternatives | Complex decisions with many criteria where weights are hard to assign directly |
| TOPSIS | Rank alternatives by distance from the ideal and anti-ideal solution | Large numbers of alternatives with quantitative criteria |
| PROMETHEE | Outranking method — compares each pair of alternatives on each criterion | Decisions where some criteria are qualitative |
| Weighted Sum Model | Simple weighted average (what you already know) | Everyday business decisions with clear criteria |
When to use MCDA: When the decision involves 3+ criteria, 3+ alternatives, multiple stakeholders with different priorities, and the trade-offs are not obvious. Common in vendor selection, project portfolio management, and policy decisions.
Hands-On Practice
Apply these frameworks to real or realistic decisions.
- Map a recent work decision using the RAPID framework. Who played each role? Was it clear?
- Build a decision tree for a personal or professional decision with at least two uncertain outcomes. Calculate expected value.
- Use weighted scoring to compare three options for a real choice (tool, job, project, vendor). Justify your weights.
- Calculate the Cost of Delay for three competing projects. Which should be prioritized?
- Perform a sensitivity analysis on a business case: identify the three most critical assumptions and test how much each would need to change to flip the decision.
- Design a simple scenario plan for your company or industry: identify two key uncertainties and create four scenario narratives.
For each scenario below, select the most appropriate decision framework and explain why:
- A startup must choose between three product roadmaps, each with different risk profiles and uncertain market response.
- An engineering team has 15 features requested and must decide which 5 to build next quarter.
- A CEO must decide whether to acquire a competitor — the outcome depends on regulatory approval and market conditions.
- A product manager must choose between two vendors, each excelling on different criteria.
Answers: (1) Decision tree or scenario planning (high uncertainty, multiple outcomes). (2) Cost of Delay / WSJF (prioritization with economic reasoning). (3) Scenario planning (deep uncertainty, multiple plausible futures). (4) Weighted scoring / MCDA (multiple criteria comparison).
Deeper Worked Examples
A decision tree for cutting the free trial from 14 days to 7
Step 1
Step 2
Step 3
Step 4
Step 5
Step 6
Weighted scoring to select a customer-support outsourcing partner
Step 1
Step 2
Step 3
Step 4
Step 5
Step 6
Common Mistakes
- Making complex decisions without any framework — relying on gut feel or HiPPO
- Using the wrong framework for the uncertainty level — simple decisions do not need scenario planning
- Assigning arbitrary weights in weighted scoring without justification
- Ignoring the downside in expected value calculations — EV is not everything; survival matters
- Building one scenario plan and treating it as a forecast — scenarios are for preparing, not predicting
- Performing sensitivity analysis on too many variables — focus on the 3–5 that matter most
- Not clarifying decision roles (RAPID) before meetings — leads to confusion and delays
Real-World Usage
Amazon uses a structured decision-making approach they call "disagree and commit" — but only after the RAPID roles are clear and data has been presented. Netflix uses scenario planning for content strategy: they model "what if streaming growth slows?" vs "what if linear TV collapses faster?" and position their content portfolio to survive multiple futures. Shell famously used scenario planning in the 1970s to anticipate the oil crisis — while competitors were caught off guard, Shell's scenario teams had already planned for supply disruption and emerged as the strongest major oil company. Toyota's decision to invest heavily in hybrids in the 1990s was informed by sensitivity analysis: they tested their strategy against scenarios where gas prices stayed low, rose moderately, or spiked — and the hybrid bet was robust across most scenarios.
Mini Project
Business problem: A mid-size e-commerce company must decide between three strategic investments for the next year: (1) Build a mobile app ($500K, uncertain adoption), (2) Expand to international markets ($1M, regulatory complexity), or (3) Invest in AI-powered personalization ($750K, technology risk). Budget allows only one.
Objective: Apply multiple decision frameworks to arrive at a well-reasoned recommendation.
Task 1
Task 2
Task 3
Task 4
Task 5
Task 6
Task 7
| Deliverable | What it proves |
|---|---|
| RAPID map | You can clarify decision roles and accountability |
| Weighted scoring | You can compare alternatives systematically |
| Decision tree | You can map options and probabilities visually |
| Expected value | You can make risk-adjusted quantitative choices |
| Sensitivity analysis | You can identify and test critical assumptions |
| Scenario plan | You can prepare for multiple possible futures |
| CEO recommendation | You can synthesize frameworks into a clear business case |
Key Takeaways
- RAPID clarifies who Recommends, Agrees, Performs, Provides Input, and Decides
- Decision trees map options, probabilities, and payoffs — forcing explicit assumptions
- Weighted scoring compares alternatives against weighted criteria — simple and powerful
- Cost of Delay quantifies the economic impact of waiting — turns prioritization into economics
- Expected value probability-weights all outcomes — but always check the downside
- Scenario planning prepares for multiple futures when probabilities are unreliable
- Sensitivity analysis reveals which assumptions the decision depends on most
- MCDA provides mathematical rigor for complex multi-criteria trade-offs
Key Terms & Glossary
| Term | Meaning |
|---|---|
| RAPID | A role-clarifying framework in which someone Recommends, another Agrees, someone else Performs, others give Input, and exactly one person has the D for Decide. |
| Decision tree | A branching map of a decision showing choice nodes, chance nodes, probabilities and payoffs, used to calculate expected values across options. |
| Expected value | The probability-weighted average of all possible outcomes of a choice; useful for long-run decisions but never a promise about a single bet. |
| Cost of delay | The economic loss of not delivering something now, used in weighted-shortest-job-first prioritisation to rank competing work by value per unit of time. |
| Scenario planning | Developing several plausible futures and testing a strategy against each, used when probabilities cannot be reliably assigned. |
| Sensitivity analysis | Varying the key assumptions of a decision to see how much each would need to change before the recommendation flips. |
Learning Resources
Decide and Deliver — Bain & Company (RAPID)
The book that introduced the RAPID framework for clarifying decision roles and accountability.
Multi-Criteria Decision Analysis (Wikipedia)
Comprehensive overview of MCDA methods including AHP, TOPSIS, and PROMETHEE with comparisons.
The Art of Strategy — Avinash Dixit & Barry Nalebuff
An accessible introduction to game theory and strategic decision-making frameworks.
WSJF Prioritization (SAFe Framework)
Detailed guide to Weighted Shortest Job First prioritization using Cost of Delay.
The Art of the Long View — Peter Schwartz
The foundational book on scenario planning, with case studies from Shell, the military, and business.
Chapter Quiz
Chapter Quiz — Decision Frameworks & Models
Five questions on RAPID roles, expected value, the downside test, cost-of-delay prioritization, and sensitivity analysis.
Question 1 of 5
In the RAPID framework, the person with final authority to say yes or no is:
Knowledge Check
End-of-Topic Exam
Hint: The two classic pathologies have an almost-empty D or too many Agrees.
Model answer: Model answer shown for a purchase of a new CRM: the sales analyst Recommends based on the vendor comparison; finance Agrees on budget; the implementation partner Performs; the sales managers give Input on requirements; and the VP of Sales is the single D. Mapping this prevents the most common failure — everyone assuming someone else decides (no D) or several people quietly holding veto power (too many Agrees). If two people claim the D role, the issue must be escalated so one accountable owner is appointed before the meeting proceeds, because ambiguity here guarantees delay and later second-guessing.
Hint: EV is an average over many repetitions; a single roll can still destroy you.
Model answer: The expected value is (0.30 x £10M) + (0.50 x £2M) + (0.20 x minus £3M) = £3M + £1M minus £0.6M = £3.4M. A company near bankruptcy should refuse because EV is a long-run average over repeated identical bets: this is a single roll, and the 20% failure branch wipes out the capital the company cannot survive losing. When the downside is existential, expected value must be paired with a worst-case test — survival comes before average returns.
Hint: 2030 involves deep uncertainty; scenarios differ from forecasts in what they are used for.
Model answer: A single forecast assumes we can put reliable probabilities on a distant, uncertain future, which invites overconfidence; scenario planning instead builds several internally consistent futures and asks whether the plan is robust to each, so the board is prepared rather than predicted. For a four-scenario matrix you would pick the two most uncertain and most influential drivers as axes — for an energy company, for example, the speed of regulation versus the price trajectory of renewables — then write a named one- or two-paragraph narrative for each quadrant and stress-test the plan against all four.
Hint: The plausible range includes the break-even point, which tells you how load-bearing the assumption is.
Model answer: Because the break-even point (4%) sits inside the plausible range (2% to 15%), the decision is highly sensitive to this one assumption: at 2% the investment fails, at 15% it is a winner, so the recommendation can flip within a plausible variation. The correct use of sensitivity analysis is to flag this assumption as load-bearing and gather better data on it before committing — for example, run a smaller controlled test to estimate the lift — rather than signing off on the base case of 10%.