Contents
Use this when you have multiple options and need to choose one — not by gut feel, but by systematically scoring each option against the factors that actually matter. The Decision Matrix forces you to separate what you value from how well each option delivers on those values, turning a tangled multi-variable comparison into arithmetic.
Section 1
What This Tool Does
You're choosing between three enterprise software vendors. Or four candidates for a VP role. Or five cities for a new distribution centre. Each option has strengths. Each has drawbacks. The strengths and drawbacks live on different dimensions — cost, speed, risk, cultural fit, scalability — and those dimensions don't matter equally. Your brain knows this. It cannot, however, hold six options across eight unevenly weighted dimensions and produce a reliable ranking. Nobody's can. What the brain does instead is latch onto the dimension that feels most urgent in the moment, or the option that a trusted colleague endorsed last week, or the alternative that avoids the most frightening downside. These are heuristics, not analysis. They work often enough to feel reliable. They fail catastrophically when the stakes are high and the options are close.
The Decision Matrix — also called a Weighted Scoring Model — is the antidote to that failure mode. The mechanism is almost embarrassingly simple: list your options as rows, your evaluation criteria as columns, assign a weight to each criterion reflecting its relative importance, score each option against each criterion, multiply scores by weights, and sum. The option with the highest total wins. That's it. No proprietary methodology, no certification required, no software necessary. A spreadsheet. A whiteboard. The back of an envelope, if the decision is small enough.
The simplicity is the point. The Decision Matrix doesn't make the decision for you — it makes your reasoning visible, decomposed, and auditable. When you choose a vendor "because it felt right," nobody can interrogate that reasoning. When you choose a vendor because it scored 4.2 versus 3.8 on a weighted matrix, anyone can ask: Why did you weight "integration speed" at 25%? Why did Vendor B get a 3 on scalability instead of a 4? These are productive questions. They surface disagreements about values and evidence rather than letting those disagreements hide behind competing intuitions.
The tool's real power isn't computational — any calculator can multiply and add. It's structural. The matrix forces you to do three things that unaided judgment routinely skips: define your criteria before you evaluate options, weight those criteria before you see which option benefits from which weighting, and evaluate each option on each dimension independently rather than forming a global impression that contaminates every individual assessment. That sequencing — criteria, then weights, then scores — is a cognitive debiasing protocol disguised as a spreadsheet.
Stuart Pugh formalised a version of this approach in the 1980s as the "Pugh Matrix" for engineering design selection at the University of Strathclyde. Benjamin Franklin described a cruder version in a 1772 letter to Joseph Priestley — his "moral or prudential algebra" of listing pros and cons, then striking out items of equal weight from opposing columns. The underlying logic is older than either: decompose, weight, score, aggregate. What changes across eras is the rigour of the weighting and the discipline of the scoring. The modern Decision Matrix, as used in product management, procurement, site selection, and hiring, inherits from all of these traditions while adding the crucial step of explicit numerical weights.
Section 2
How to Use It — Step by Step
Instructions on the left. Worked example — "Which of three cities should we choose for our second fulfilment centre?" — on the right.
Step 1 — Define
List your options and confirm they are genuinely comparable
Before building the matrix, verify that you're comparing like with like. Every option should be a plausible answer to the same question. If one "option" is actually a different strategy (e.g., "don't open a second fulfilment centre at all"), it doesn't belong in the matrix — that's a prior decision. Aim for 3–6 options. Fewer than three and you don't need a matrix. More than six and the scoring burden becomes so heavy that quality degrades. If you have twelve options, screen them down to a shortlist first using simpler criteria (must-haves, deal-breakers) before building the weighted matrix.
Worked example
Fulfilment centre site selection
After an initial screen eliminating cities that lack adequate warehouse inventory or sit outside the target logistics corridor, three finalists remain: Columbus, OH, Nashville, TN, and Reno, NV. Each can serve a meaningful portion of the customer base. Each has available industrial real estate. All three are genuine contenders — no obvious winner, which is exactly when the matrix earns its keep.
Step 2 — Criteria
Identify 5–8 evaluation criteria that capture what actually matters
This is the step most teams rush through, and it's the step that determines everything. Bad criteria produce a precise answer to the wrong question. Good criteria are specific, measurable (or at least assessable), and collectively exhaustive — they cover every dimension that would influence your satisfaction with the decision a year from now. Avoid overlapping criteria (e.g., "cost" and "affordability" are the same thing). Avoid criteria so vague they can't be scored ("quality," "fit"). Test each criterion: if two options scored identically on everything else, would a difference on this criterion change your choice? If not, drop it.
Worked example
Criteria for the fulfilment centre
The team identifies six criteria: Labour availability (can we hire 200+ warehouse workers within 90 days?), Lease cost per sq ft (annual), Proximity to customer base (% of orders deliverable in 2 days), State/local tax incentives, Carrier network density (number of major carriers with local hubs), Expansion potential (can we double capacity at the same site within 3 years?). Each criterion is specific enough to score. None overlaps with another.
Step 3 — Weight
Assign percentage weights to each criterion — they must sum to 100%
Weighting is where your strategy becomes explicit. Two teams facing the same decision with the same options might weight criteria entirely differently — a bootstrapped company optimising for cash preservation will weight lease cost heavily; a venture-backed company optimising for speed will weight proximity to customers and carrier density. That's not a flaw. It's the feature. The weights encode your priorities. Assign weights before you score options. This is critical. If you score first and weight second, you'll unconsciously adjust weights to favour the option you already prefer. Use round numbers (5%, 10%, 15%, etc.) to avoid false precision. If the team can't agree on weights, that disagreement is the most valuable output of the exercise — it means you haven't aligned on strategy.
Worked example
Weighting the fulfilment criteria
The company is scaling rapidly and has raised a Series C — speed to serve customers matters more than cost minimisation. Weights: Proximity to customer base: 30%, Labour availability: 20%, Carrier network density: 20%, Lease cost: 15%, Expansion potential: 10%, Tax incentives: 5%. Total: 100%. The CEO initially pushed for lease cost at 25%, but the logistics VP argued that a cheaper site 500 miles from the customer cluster would cost more in shipping than it saved in rent. The weight debate surfaced a strategic misalignment that would have remained hidden without the matrix.
Step 4 — Score
Rate each option on each criterion using a consistent scale
Use a 1–5 or 1–10 scale. Define what each score means for each criterion before you start scoring — otherwise "a 4 on labour availability" means something different to every person in the room. Score one criterion at a time across all options, not one option at a time across all criteria. This row-by-row approach forces comparative judgment ("Is Columbus better or worse than Nashville on labour availability?") rather than holistic impression ("How do I feel about Columbus overall?"). If you have data, use it. If you're estimating, be explicit about the uncertainty. Where team members disagree on a score, discuss the evidence, then either converge or average.
Worked example
Scoring the three cities
Scoring on a 1–5 scale, criterion by criterion. Proximity to customer base (30%): Columbus 5, Nashville 4, Reno 2 — Columbus sits within two-day ground shipping of 47% of the US population. Labour availability (20%): Columbus 4, Nashville 4, Reno 3 — Reno's labour market is tighter due to competing warehouse demand from major e-commerce players. Carrier network density (20%): Columbus 5, Nashville 4, Reno 3 — Columbus is a top-five US logistics hub. Lease cost (15%): Columbus 3, Nashville 3, Reno 4 — Reno's industrial rents are lower. Expansion potential (10%): Columbus 4, Nashville 3, Reno 5 — Reno has abundant land. Tax incentives (5%): Columbus 3, Nashville 4, Reno 4 — Tennessee and Nevada both offer competitive packages.
Step 5 — Calculate and Interpret
Multiply each score by its weight, sum across criteria, and interrogate the result
The arithmetic is trivial. The interpretation is everything. Look at the total scores — but also look at the component scores. An option that wins overall but scores a 1 on a criterion weighted at 15% has a real vulnerability. An option that loses by a thin margin but dominates on your highest-weighted criterion deserves a second look. Run a sensitivity check: if you shifted the top weight by ±5%, would the winner change? If yes, the decision is fragile and the weight debate matters more than the scores. If the top two options are within 5% of each other, the matrix is telling you they're effectively tied — and you should decide based on factors the matrix didn't capture, or gather more data to break the tie.
Worked example
The result — and what it reveals
Columbus: (5×0.30) + (4×0.20) + (5×0.20) + (3×0.15) + (4×0.10) + (3×0.05) = 1.50 + 0.80 + 1.00 + 0.45 + 0.40 + 0.15 = 4.30. Nashville: (4×0.30) + (4×0.20) + (4×0.20) + (3×0.15) + (3×0.10) + (4×0.05) = 1.20 + 0.80 + 0.80 + 0.45 + 0.30 + 0.20 = 3.75. Reno: (2×0.30) + (3×0.20) + (3×0.20) + (4×0.15) + (5×0.10) + (4×0.05) = 0.60 + 0.60 + 0.60 + 0.60 + 0.50 + 0.20 = 3.10. Columbus wins by a clear margin. Sensitivity check: even if Proximity drops from 30% to 20% and Lease Cost rises to 25%, Columbus still leads. The decision is robust. Reno's strengths (cost, expansion room) simply don't compensate for its distance from the customer base given this company's priorities.
Section 3
When It Works Best
✓
Ideal Conditions for the Decision Matrix
| Dimension | Best fit |
|---|---|
| Number of options | 3–6 options that have survived an initial screen. Fewer than three and the comparison is trivial. More than six and scoring fatigue introduces noise — people start satisficing on scores rather than genuinely evaluating. Use a quick pass/fail filter to get to a shortlist before deploying the matrix. |
| Criteria clarity | The team can articulate what "good" looks like across multiple dimensions. If you can't define your criteria, you're not ready for a matrix — you need Abstraction Laddering or Reframing to clarify what you're actually optimising for. The matrix is a scoring tool, not a goal-setting tool. |
| Stakeholder alignment | Multiple stakeholders with different priorities need to reach a shared decision. The matrix's greatest value is often not the final score but the weight-setting conversation, which forces implicit priorities into the open. Finance wants cost. Engineering wants scalability. The matrix makes that tension productive rather than political. |
| Decision reversibility | Most valuable for irreversible or expensive-to-reverse decisions: site selection, major vendor contracts, senior hires, platform migrations. For easily reversible decisions, the overhead of building a proper matrix exceeds the cost of choosing wrong and correcting. Match the rigour of the tool to the stakes of the decision. |
| Data availability | Works with hard data (lease rates, population density) and informed judgment (cultural fit, expansion potential) — but works best when you have at least some quantitative inputs to anchor the scores. A matrix built entirely on subjective estimates is better than no structure, but only marginally. Push for data wherever you can get it. |
| Decision type | Selection decisions — choosing one option from a set. Not suited for go/no-go decisions (use Cost-Benefit Analysis), sequencing decisions (use Impact-Effort Matrix), or diagnostic problems (use Ishikawa or 5 Whys). The matrix answers "which one?" not "should we?" or "in what order?" |
Section 4
When It Breaks Down
⚠
Failure Modes
| Failure pattern | What goes wrong | What to use instead |
|---|---|---|
| Criteria gaming | Someone who already has a preferred option reverse-engineers the criteria and weights to guarantee that option wins. They add criteria where their favourite excels and weight them heavily. The matrix becomes a rationalisation engine rather than an evaluation tool. The tell: criteria that only one option scores well on. | Set criteria and weights before revealing options, or have different people set weights and score options |
| False precision | Scoring to two decimal places on a subjective 1–5 scale. Weights at 17.3%. Declaring a winner by 0.02 points. The matrix produces numbers, and numbers feel objective, but the inputs are often estimates with wide uncertainty bands. A margin of victory smaller than the scoring uncertainty is meaningless — it's noise dressed as signal. | Run sensitivity analysis on close results; use coarser scales (1–3) when data is sparse |
| Missing the deal-breaker | A weighted average can mask a fatal flaw. An option scores 5 on five criteria but 1 on "regulatory compliance" — and the matrix ranks it first because compliance was only weighted at 10%. But a 1 on compliance isn't a weakness; it's a disqualification. The matrix treats all low scores as gradations when some are binary. | Apply must-have thresholds before scoring — any option below the minimum on a critical criterion is eliminated, not scored |
| Incommensurable values | Some decisions involve criteria that resist numerical comparison. How do you score "cultural alignment" on the same scale as "cost per square foot"? The matrix assumes all criteria can be reduced to a common scoring scale. When they can't — when you're comparing the measurable against the meaningful — the aggregation produces a number that obscures more than it reveals. | Hard Choice Model for decisions where values are genuinely incommensurable; use the matrix for the quantifiable dimensions only |
| Too many criteria | Teams add criteria to be "thorough" until the matrix has 15 columns. Each additional criterion dilutes the weight of every other criterion. With 15 criteria, even the most important one can only carry ~15% weight, which means nothing dominates — and the result converges toward a bland average that favours the most mediocre option. The "good enough at everything" option wins over the "exceptional at what matters most" option. | Cap at 5–8 criteria; use Pareto thinking to identify the 20% of factors that drive 80% of the decision quality |
| Static snapshot bias | The matrix evaluates options as they are today. It has no mechanism for incorporating how options might evolve. A vendor that scores a 3 on product capability today might be investing heavily and score a 5 in eighteen months. A city with a 4 on labour availability might face a 2 after a major employer opens a competing facility. The matrix is a photograph, not a film. | Scenario Planning to model how scores might shift under different futures; Decision Tree for sequential or contingent decisions |
The most dangerous failure mode is criteria gaming, because it's the hardest to detect and the most corrosive to trust. When a matrix is gamed, the output looks rigorous — weights sum to 100%, scores are documented, the arithmetic checks out. But the conclusion was predetermined. Everyone in the room senses it, even if they can't articulate exactly how the manipulation occurred. The result: the team loses faith in structured decision-making altogether, which is worse than never having used the tool. The protection is procedural, not technical. Separate the people who set criteria and weights from the people who score options. Or set criteria and weights in a session where the specific options haven't been revealed yet — force the team to articulate what matters in the abstract before they know which option benefits from which weighting. This single procedural change transforms the matrix from a rationalisation tool into an actual decision tool.
Section 5
Visual Explanation
Decision Matrix — fulfilment centre site selection. Weighted scores shown in parentheses. Columbus wins with a total of 4.30, driven by dominance on the two highest-weighted criteria.
Section 6
Pairs With
The Decision Matrix is an evaluation tool — it scores and ranks options you've already generated against criteria you've already defined. Its power increases dramatically when paired with tools that strengthen the inputs (better options, sharper criteria) or stress-test the outputs (sensitivity analysis, second-order consequences).
Use before
Issue Trees
Before you can score options, you need the right options. Issue Trees decompose a decision space into mutually exclusive, collectively exhaustive branches, ensuring you haven't missed a viable alternative. A matrix that evaluates three options when a fourth — the best one — was never considered produces a precise but wrong answer.
Use before
First Principles Thinking
Criteria selection is the matrix's most consequential step. First Principles Thinking helps you derive criteria from fundamental requirements rather than copying them from a template or a previous decision. "What must be true for this decision to succeed?" generates better criteria than "What did we evaluate last time?"
Use before
Abstraction Laddering
When the team can't agree on criteria, it's often because they're operating at different levels of abstraction. One person says "cost." Another says "total cost of ownership over five years including switching costs." Abstraction Laddering moves the conversation up and down until the team finds the right level of specificity for each criterion.
Use after
Pre-Mortem
The matrix picks a winner. The Pre-Mortem stress-tests it. "Assume we chose Columbus and it failed spectacularly eighteen months from now. What went wrong?" This surfaces risks the matrix didn't capture — political risks, execution risks, assumptions baked into the scores that might not hold.
Use after
[Second-Order Thinking](/mental-models/second-order-thinking)
A matrix evaluates first-order effects: how does each option perform on each criterion today? Second-Order Thinking asks what happens next. Choosing the cheapest vendor might trigger a quality problem that triggers customer churn that costs ten times the savings. The matrix can't see around corners. This tool can.
Mental model
Reversible vs. Irreversible Decisions
Not every decision deserves a matrix. Jeff Bezos's distinction between Type 1 (irreversible, high-stakes) and Type 2 (reversible, low-stakes) decisions helps you calibrate when to invest in a full weighted evaluation and when to just decide and iterate. The matrix is a Type 1 tool applied to Type 1 decisions.
Section 7
Real-World Application
NASA — Mars rover landing site selection
The scenario
Selecting a landing site for a Mars rover is among the highest-stakes decisions in science and engineering. Get it wrong and you lose a multi-billion-dollar mission — not to mention a decade of scientific planning. For the Mars Science Laboratory mission (the Curiosity rover, launched in 2011), NASA needed to choose one landing site from over thirty initial candidates. Each site offered different scientific value, but the engineering constraints were brutal: the site had to be within a specific latitude band, below a certain elevation (for sufficient atmospheric braking), relatively flat (to survive landing), and free of large rocks and steep slopes. Scientific promise and engineering safety pulled in opposite directions.
How the tool appliedNASA convened a series of community workshops between 2006 and 2008 where planetary scientists and engineers evaluated candidate sites using a structured scoring approach that functioned as a multi-round Decision Matrix. Criteria included mineralogical diversity, evidence of past water activity, geological context, landing safety margins, and rover traversability. Each criterion carried explicit weights reflecting the mission's dual mandate: maximise science return while maintaining acceptable engineering risk. Sites were scored by independent teams — scientists scored science criteria, engineers scored safety criteria — preventing any single group from gaming the weights. The process narrowed thirty-plus candidates to four finalists (Gale Crater, Eberswalde Crater, Holden Crater, and Mawrth Vallis), then to one.
What it surfacedGale Crater won — and the reason was instructive. It didn't score highest on any single science criterion. Eberswalde had stronger evidence of a delta deposit. Mawrth Vallis had richer clay mineral signatures. But Gale scored well across nearly every science dimension and had a five-kilometre-high central mound (Mount Sharp) that exposed billions of years of geological history in a single traverse. The matrix's aggregation revealed that Gale's breadth of scientific opportunity, combined with acceptable engineering margins, made it the strongest overall choice. No individual scientist's intuition would have reached that conclusion — each specialist naturally favoured the site that best served their specific discipline.
The non-obvious factorThe separation of scoring responsibilities was the key design choice. Scientists couldn't inflate engineering safety scores to favour their preferred site. Engineers couldn't downweight science criteria to simplify their landing problem. The matrix structure enforced intellectual honesty by making it structurally impossible for any single stakeholder group to control the outcome. NASA also ran explicit sensitivity analyses — varying the science-to-safety weight ratio to see whether the winner changed. Gale Crater proved robust across a wide range of weightings, which gave the final decision unusual confidence. The matrix didn't just pick a site. It demonstrated why that site was the right pick under multiple plausible priority schemes — a level of decision transparency that would have been impossible with expert judgment alone.
Section 8
Analyst's Take
Faster Than Normal — Editorial View
The Decision Matrix is the most widely used structured decision tool in business, and also the most widely abused. Its accessibility is both its strength and its vulnerability. Anyone can build one in ten minutes. That ease of construction creates a dangerous illusion: the belief that having a matrix means having a rigorous process. It doesn't. A matrix is only as good as the criteria it evaluates, the weights it assigns, and the honesty of the scores it contains. Get any of those wrong and you've built an elaborate justification machine — one that produces a number, which feels like an answer, which ends the conversation prematurely.
The failure I see most often isn't criteria gaming (though that's common) — it's weight avoidance. Teams assign equal weights to every criterion because they can't agree on priorities, or because equal weights feel "fair." This is the worst possible default. Equal weighting is itself a strong claim: it asserts that every criterion matters exactly as much as every other. That's almost never true, and when it is, you probably don't need a matrix. The weight-setting conversation is the decision. If your team can't have that conversation — can't say "speed matters twice as much as cost to us right now" — the matrix will produce a result that satisfies nobody because it reflects nobody's actual priorities. I'd rather see a team argue for an hour about weights and never finish scoring than see them breeze through with equal weights and declare a winner.
The highest-leverage improvement: score one criterion at a time across all options, not one option at a time across all criteria. This sounds like a minor procedural detail. It's not. When you evaluate Option A across all six criteria, then Option B across all six, you form a holistic impression of each option that contaminates every individual score. You like Option A, so you unconsciously give it 4s where it deserves 3s. Scoring by criterion — "How does each option perform on labour availability?" — forces genuine comparison and breaks the halo effect. It's slower. It feels less natural. It produces dramatically more honest scores. Every matrix I've seen that produced a surprising, genuinely useful result used criterion-by-criterion scoring. Every matrix that merely confirmed what the loudest person in the room already believed used option-by-option scoring.
Section 9
Top Resources
01
Book
The best book on making strategy choices explicit — which is what the Decision Matrix ultimately demands. Lafley and Martin's "Where to Play / How to Win" framework provides the strategic logic that should drive your criteria and weights. Read this before building a matrix for any strategic decision; it ensures you're evaluating options against the right dimensions rather than the convenient ones.
02
Book
The scientific foundation for why the Decision Matrix works. Kahneman's chapters on anchoring, the halo effect, and substitution explain exactly which cognitive biases the matrix is designed to counteract — and which ones it can't. His discussion of "mediating assessments" in hiring (scoring candidates on individual traits before forming an overall judgment) is the direct ancestor of criterion-by-criterion scoring. Essential reading for anyone who wants to understand the tool's mechanism, not just its mechanics.
03
Book
Horowitz doesn't discuss decision matrices explicitly, but his chapters on CEO decision-making — particularly on hiring executives and making "wartime" choices — illustrate both when structured evaluation helps and when it becomes a crutch. His account of choosing between keeping or firing a friend-turned-underperforming-executive is a masterclass in the limits of scoring: some decisions involve incommensurable values that no matrix can resolve. Read it as the counterweight to over-reliance on structured tools.
04
Smart Choices: A Practical Guide to Making Better Decisions — John Hammond, Ralph Keeney & Howard Raiffa (1999)
BookThe most rigorous practical guide to multi-criteria decision-making available for a general audience. Hammond, Keeney, and Raiffa — all affiliated with Harvard — walk through the complete process of structuring decisions, identifying objectives, creating alternatives, and evaluating tradeoffs. Their "Even Swaps" method is a powerful complement to weighted scoring when criteria resist numerical comparison. Chapter 6 on tradeoffs is worth the price of the book alone.
05
Book
Cagan's framework for product prioritisation — particularly his guidance on opportunity assessment and feature scoring — is the Decision Matrix applied to product management. His insistence on separating "value" from "effort" in prioritisation mirrors the matrix's separation of criteria from weights. For product leaders who need to choose between competing roadmap bets, this is the most practical translation of weighted scoring into a tech-company context.
Decision Tools Library — Browse by phase
Why this matters next
decision contextsRaise and Allocate Capital
Raise and Allocate Capital uses this research to help founders treat fundraising and allocation as linked decisions about control, opportunity cost, survivability, and compounding.
mental modelsIncentives
Issue Trees applied the Incentives mental model
mental modelsSecond-Order Thinking
Issue Trees applied the Second-Order Thinking mental model
mental modelsFirst Principles Thinking
Issue Trees applied the First Principles Thinking mental model
mental modelsLeverage
Issue Trees applied the Leverage mental model
mental models5 Whys
Issue Trees applied the 5 Whys mental model
Continue exploring
SHCADTSP
Decision tool
Six Thinking Hats
Examine a decision from six cognitive perspectives — facts, emotions, risks, ben
Decision tool
Cost-Benefit Analysis
Quantify the economic trade-offs of each option to determine net value
Decision tool
Decision Tree
Map sequential outcomes with probabilities and expected values to find the highe
Decision tool
Scenario Planning
Construct multiple plausible futures and test how each option performs across th
More like this, in your inbox
I send a newsletter every week — free, no spam, unsubscribe anytime.