Personal Finance Decision Tree
You got two job offers the same week. Offer A pays $150K salary with guaranteed Total Compensation of $180K per year. Offer B pays $120K salary plus Equity Compensation worth $200K at face value, paid out over four years, at a PE-Backed company targeting a sale in three years. You can't just compare the dollar amounts - you need to map out the sequences of things that could happen, assign probabilities, and work backward to see which path has the higher Expected Payoff. That's a decision tree.
A decision tree lays out sequential choices and uncertain outcomes as a branching diagram, attaches probabilities and dollar values to each branch, then uses Expected Value to collapse the tree backward from endpoints to a single number per option - so you can apply a decision rule and pick.
A decision tree is a diagram with two kinds of nodes:
Every path through the tree ends at a terminal value - the dollar outcome if that exact sequence of decisions and events plays out.
You already know Expected Value: multiply each outcome by its probability and sum. A decision tree is just Expected Value applied recursively across a sequence of decisions and uncertainties. You start at the terminal values and fold backward, replacing each chance node with its Expected Value and each decision node with whichever branch wins under your decision rule.
The result: a single Expected Payoff for each option available to you right now, incorporating everything that might happen downstream.
Most real choices aren't one-shot bets. They're sequences:
Without a decision tree, people flatten these into a single gut-feel comparison. They anchor on the first branch and ignore what happens two or three steps out. Operators who run a P&L can't afford that - a choice that looks cheap at step one can cascade into a massive Error Cost at step three.
Decision trees also expose your Outside Option at every node. If you map the full tree and one branch dominates regardless of what the world does, you've found a Dominant Strategy - and you can stop agonizing.
Step 1: Draw the structure.
Start at the left with your first decision. Branch right for each option. After each option, ask: what uncertain thing happens next? Draw a chance node. After each chance resolution, ask: do I face another choice? If yes, draw another decision node. Repeat until every path reaches a terminal value.
Step 2: Assign terminal values.
These are the Cash Flow or net worth outcomes at each endpoint. Use real numbers - Total Compensation over your Time Horizon, Net Present Value of an investment, or the Expected Total Cost of a commitment like a rent-vs-buy decision.
Step 3: Assign probabilities at every chance node.
Probabilities across branches leaving any single chance node must sum to 1.0. Use your base case estimates. If you don't have good numbers, that's a signal you need more information before deciding - the tree makes that ignorance visible.
Step 4: Fold backward ("roll back").
Start at the rightmost nodes:
Replace each node with its rolled-back value. When you reach the root, each initial option has a single number.
Step 5: Decide.
Apply your decision rule. If Option A's Expected Payoff is $185K and Option B's is $162K, and your decision rule says "pick the higher EV unless the downside Tail Risk exceeds X," check that condition and commit.
Use a decision tree when:
Do not use a decision tree when the problem is a single-stage comparison with no sequential structure. There, Expected Value on its own is the right tool.
You're deciding between renting at $2,400/month ($144,000 over 5 years) or buying a condo for $400,000 with an $80,000 down payment, a $320,000 mortgage at a 6.5% mortgage rate, and a monthly payment of $2,023. Your Time Horizon is 5 years. You estimate two scenarios for the local market: 60% chance of moderate Appreciation (home worth $440,000 at year 5) and 40% chance of a Market Downturn (home worth $360,000 at year 5). If you buy and a downturn hits, you face a second decision at year 3: hold through year 5, or sell early and rent the remaining 2 years. Assume 6% selling costs on any sale.
Draw the tree. Root decision: Rent vs Buy. The Buy branch leads to a chance node (60% Appreciation, 40% Market Downturn). The downturn branch leads to a second decision node at year 3: Hold through year 5, or Sell at year 3 and rent for 2 years.
Assign terminal values (net cost over 5 years). First, two derived quantities from standard Amortization on the $320,000 loan at 6.5% over 30 years: remaining principal balance is ~$299,600 after 60 payments (year 5) and ~$308,500 after 36 payments (year 3). You can verify these with any amortization calculator - the balance goes down on an amortizing mortgage.
Rent path: $2,400 x 60 months = -$144,000 total housing cost.
Buy + Appreciation (sell at year 5): Out of pocket = $80,000 down + $121,380 mortgage payments ($2,023 x 60) = $201,380. Sale proceeds = $440,000 - $26,400 selling costs (6%) - $299,600 mortgage payoff = $114,000. Net cost: $201,380 - $114,000 = -$87,400.
Buy + Downturn + Hold (sell at year 5): Same out of pocket = $201,380. Sale proceeds = $360,000 - $21,600 selling costs - $299,600 payoff = $38,800. Net cost: $201,380 - $38,800 = -$162,600.
Buy + Downturn + Sell at Year 3: Home value at year 3, interpolating linearly from $400,000 purchase price toward the downturn target of $360,000: $400,000 - (3/5)($40,000) = $376,000. Sale proceeds = $376,000 - $22,560 selling costs - $308,500 payoff = $44,940. Mortgage payments for 3 years = $72,828 ($2,023 x 36). Rent remaining 2 years = $57,600 ($2,400 x 24). Net cost: $80,000 + $72,828 + $57,600 - $44,940 = -$165,500.
Fold backward. At the year-3 decision node (downturn path): Hold = -$162,600 vs Sell Early = -$165,500. Decision rule says minimize cost, so Hold wins. Replace that decision node with -$162,600. At the chance node: 0.60 x (-$87,400) + 0.40 x (-$162,600) = -$52,440 + (-$65,040) = -$117,500 expected cost for the Buy path.
Compare at root. Rent = -$144,000 expected cost. Buy = -$117,500 expected cost. Buy is ~$26,500 cheaper in expectation. But the downturn branch at -$162,600 is $18,600 more expensive than renting - if your Risk Tolerance can't absorb that downside, your decision rule might override the EV advantage.
Insight: The tree revealed two things a flat comparison would miss: (1) selling early in a downturn is strictly worse than holding - a Dominant Strategy at that node eliminates a branch and simplifies your plan, and (2) buying has a meaningful $26,500 Expected Value edge, but renting at $144,000 sits between buying's best case ($87,400) and worst case ($162,600). The spread is $75,200. Whether you buy depends on your Risk Tolerance for the specific downturn downside, not just the Expected Payoff.
Current job: $160K salary, stable. Startup offer: $110K salary plus Equity Compensation with face value of $300K, paid out over 4 years. You estimate a 25% chance the startup hits its sale target (your equity becomes worth $300K), a 50% chance of a partial outcome (equity worth $75K), and a 25% chance the company fails (equity worth $0). Time Horizon: 4 years. You also face a chance at your current job: 70% chance of normal trajectory (no raise beyond inflation) and 30% chance of promotion to $190K in year 2.
Terminal values over 4 years. Stay + No Promotion: $160K x 4 = $640K. Stay + Promotion: $160K x 1 + $190K x 3 = $730K. Leave + Full Exit: $110K x 4 + $300K = $740K. Leave + Partial: $110K x 4 + $75K = $515K. Leave + Fail: $110K x 4 + $0 = $440K.
Fold back the Stay branch. 0.70 x $640K + 0.30 x $730K = $448K + $219K = $667K.
Fold back the Leave branch. 0.25 x $740K + 0.50 x $515K + 0.25 x $440K = $185K + $257.5K + $110K = $552.5K.
Compare. Stay = $667K vs Leave = $552.5K. The startup needs to clear a $114.5K EV gap. That's not close. Even if you bump the full-exit probability to 40% - redistributing the remaining 60% across partial and failure in their original 2:1 ratio (40% partial, 20% failure) - you get 0.40 x $740K + 0.40 x $515K + 0.20 x $440K = $296K + $206K + $88K = $590K - still $77K behind staying.
Insight: The tree quantifies the opportunity cost of leaving. The startup's equity sounds impressive at face value ($300K), but once you assign honest probabilities and compare against the base case of staying - including the promotion upside you'd give up - the gap is large. You'd need either a much higher exit probability or a much larger equity grant to justify the move. Sensitivity Analysis on the exit probability tells you exactly what number would flip the decision.
The tree forces you to fill in every node with a probability or dollar value. If you can't, that's a signal you need more information before committing - a decision tree makes ignorance visible before it becomes an Error Cost.
Always fold backward from terminal values. The optimal choice at each decision node depends on what's optimal downstream - money already spent doesn't change which branch is better going forward.
The tree's biggest gift is often what it eliminates - branches where one option is strictly worse reveal a Dominant Strategy at that node, simplifying your real choice set.
Forgetting to include the 'do nothing' branch. Every decision tree should have a status-quo option. Your base case of staying put, keeping the current job, or continuing to rent is a real branch with its own chance nodes and terminal values. If you only model the shiny new option, you're comparing it against an imaginary zero.
Using probabilities that don't sum to 1.0 at a chance node. If you assign 30% to success and 40% to failure, you've left 30% unaccounted for. That missing probability mass usually hides the most likely outcome - mediocrity - which is exactly the scenario that changes the Expected Payoff the most.
You have $10,000 to allocate. Option A: put it in a High-Yield Savings Account earning 4.5% APY (guaranteed). Option B: invest in index funds. You estimate a 60% chance of a good market year (12% Expected Return), 30% chance of a flat year (1% return), and 10% chance of a bad year (-15% return). After one year, if Option B lost money, you face a second decision: hold for another year (same probability distribution applies again) or sell and move proceeds to the savings account. Draw the tree and find the Expected Payoff of each root option over a 2-year Time Horizon.
Hint: For Option A, compound the 4.5% over 2 years. For Option B, you only face the hold-or-sell decision if the first year is bad (10% branch). In the other first-year branches (good or flat), assume you hold through year 2 with the same probability distribution. Compute terminal values, then fold backward.
Option A (savings): $10,000 x (1.045)^2 = $10,920. Guaranteed.
Option B tree, Year 1 outcomes: Good (60%): $11,200. Flat (30%): $10,100. Bad (10%): $8,500.
Year 2 from Good ($11,200): 0.60 x $11,200 x 1.12 + 0.30 x $11,200 x 1.01 + 0.10 x $11,200 x 0.85 = $7,526 + $3,394 + $952 = $11,872 EV.
Year 2 from Flat ($10,100): 0.60 x $10,100 x 1.12 + 0.30 x $10,100 x 1.01 + 0.10 x $10,100 x 0.85 = $6,787 + $3,060 + $859 = $10,706 EV.
Year 2 from Bad ($8,500) - Decision node: Hold: 0.60 x $8,500 x 1.12 + 0.30 x $8,500 x 1.01 + 0.10 x $8,500 x 0.85 = $5,712 + $2,576 + $723 = $9,011 EV. Sell and save: $8,500 x 1.045 = $8,883. Hold wins ($9,011 > $8,883). Replace decision node with $9,011.
Fold back to root for Option B: 0.60 x $11,872 + 0.30 x $10,706 + 0.10 x $9,011 = $7,123 + $3,212 + $901 = $11,236 EV.
Compare: Option A = $10,920. Option B = $11,236. Option B has ~$316 higher EV, but with Variance. The bad-year branch could leave you at ~$7,225 (two bad years in a row) - a 28% loss. Apply your Risk Tolerance as your decision rule.
A decision tree extends the two concepts you already know into sequential problems. Your Expected Value skill gives you the math at each chance node - multiply and sum. Your decision rule skill gives you the discipline at each decision node - compare against your pre-committed threshold instead of rationalizing after the fact. Downstream, decision trees connect to Sensitivity Analysis, where you stress-test the probabilities at each chance node to find which assumptions matter most. They also feed into Net Present Value when terminal values occur at different points in time and need Discounting. And when two parties are making decisions that affect each other's payoffs, the tree becomes the foundation for Game Theory - your chance nodes become the other player's decision nodes.
Disclaimer: This content is for educational and informational purposes only and does not constitute financial, investment, tax, or legal advice. It is not a recommendation to buy, sell, or hold any security or financial product. You should consult a qualified financial advisor, tax professional, or attorney before making financial decisions. Past performance is not indicative of future results. The author is not a registered investment advisor, broker-dealer, or financial planner.