One option is certain, the other is a gamble. How do you compare them?
A safe choice and a risky one do not sit on the same scale, so the argument goes in circles and the loudest view wins. A decision tree puts them on the same scale. You lay out the choices, the things that could happen after each one with the odds you honestly believe, and what you are left with at the end of every path. The tree then works backwards and says what each choice is worth today.
Method guide for Google Sheets Decision trees Pro engine
What a decision tree is
A tree has three kinds of node. Decision nodes are what you control. Chance nodes are what you do not, and the branches under one carry probabilities that add up to one. End nodes carry the payoff along that path. Solving it is called rollback: at a chance node you take the probability-weighted average of the branches, at a decision node you take the best branch, and you repeat until you reach the root.
What comes out is an expected value: what you would average if you faced this decision many times. You face it once. That is why the tree also returns a risk profile, the full list of outcomes you could actually land on with the probability of each, and why the profile deserves as much attention as the average.
The questions it answers
- Which option is worth more once the odds are priced in?
- What am I paying for certainty by taking the safe one?
- What is the worst realistic outcome of the choice the average recommends?
- How much would better information be worth, before I pay for it?
When you do not need a tree
Skip a tree when one option wins on every branch. If it is better whatever happens, the arithmetic is ceremony. Skip it when the uncertainty is a continuous quantity rather than a handful of outcomes, because that is a simulation, not a tree.
Skip it when you cannot put honest numbers on the branches. A tree built on invented probabilities returns a confident answer that only launders the guess. And expected value is the wrong rule for a bet you cannot survive losing: if one branch would end the company, read the risk profile and decide on that instead of the average, or re-roll the tree at your stated risk tolerance, so a risky average and a sure thing are compared honestly.
A worked example
An offer of $120,000 is on the table. You can take it, or counter at $135,000 and find out what happens.
| Accept the offer | $120k, certain |
| Counter, they agree | 50 percent, $135k |
| Counter, they hold firm | 40 percent, $120k |
| Counter, the offer sours | 10 percent, a $105k backup |
| Goal | maximize the payoff |
The answer
Rolling the tree back to the root:
- Accept
- $120k
- Counter, expected value
- $126k
- Best move
- Counter
- Worst branch
- $105k 10% of the time
Rollback values the counter at $126,000 against $120,000 for accepting, so at these odds asking is worth about $6,000. The risk profile is the other half of the answer: the counter lands on $135k half the time, on $120k in four cases out of ten, and on the $105k backup in one. Nothing about the average removes that last branch. A tree is honest about it in a way a headline expected value is not, and you can argue with the three probabilities instead of arguing with a feeling.
Trees earn their keep when a decision comes in stages. In the safe project against the risky one, a project that disappoints 75 percent of the time still rolls back to $1.0M against $625K for the safe project. All of the edge is the stage gate: forced to launch whatever the prototype shows, the same project is worth only $175K and loses. The right to spend $500K, look, and walk away is worth $825K by itself, and no weighted average of the outcomes would ever have found that.
Figures from the shipped Job Offer template, rolled back by the same code the add-on runs.
What you get back in the sheet
- A node table on the sheet, six columns: id, parent, type, label, probability, value. No canvas to fight with, and the tree is a range you can edit, copy and review.
- Rollback to maximize a payoff or minimize a cost.
- An optimal policy table saying what to choose at every decision, not only the first one.
- The risk profile under that policy: each distinct outcome, its probability, the running cumulative probability, and the path that leads there.
- Rolled-back values on every node, so you can see where the value came from rather than trusting one number at the root.
- A probability sweep with flip points: one chosen probability swept across its range, with the exact values where the best decision changes, so you know how wrong your estimate is allowed to be.
- A ceiling on what research could be worth: the value of perfect information, and of an imperfect study, priced before you commission either.
- Rollback at your risk tolerance: state how much risk you can carry and the tree is re-rolled on that basis, so a risky average and a sure thing are compared honestly.
- A written reading of the result. Every run ends with a card titled “What this means”: the probability that matters, the biggest driver, and what to tighten first. On the free tier the tool writes it from its own figures. On Pro you also get an AI reading of the same figures, written by a language model, and See what was sent shows the whole payload: “Ratios, shares, counts and cell references only. No cell values, no labels, no names, no formulas.”
Free or Pro
This is a Pro engine, one of five. Every free install includes five full-quality runs of any Pro engine: the same engine with nothing switched off, at any model size, on your own numbers. The five are one allowance shared across all five Pro engines, not five for each. A run counts only once it has produced a report, so a cancelled or failed run costs you nothing.
After that, this engine asks you to upgrade and nothing else does. Every statistics, forecasting and machine-learning tool stays free on every plan, and optimization and what-if stay free with limits set by the method rather than the plan: 2,000 decision cells on Simplex LP, 32 changing cells in a scenario. Nothing you have built stops working. Pro also removes the “Made with Sortia” footer from generated report tabs.
The other four Pro engines are Monte Carlo risk simulation, Schedule Risk (Monte Carlo CPM), Critical Chain and Optimization under uncertainty. Pro is $199/year, and a Day Pass covers seven days for $9 if you have a single decision to make.
Templates that use it
Each one loads into your sheet with the numbers already in place, and every figure on its page was computed by the tool itself.
- The worked example aboveJob Offer: Counter or Accept?
- A certain settlement against a verdict, where the answer is genuinely closeSettle or Go to Trial?
- What a stage gate is worth, pricedFund the Safe Project or the Risky One?
- The ceiling price of information, and why a cheap survey can be worth nothingIs That Market Study Worth It?
- The break-even price of never tryingShould You Take the Career Leap?
- Sign the round or wait for the milestone, read against the down-round branchRaise Now, or Wait for the Milestone?
The other 25 models in the library that run decision trees:
- Which apartment should you take?Which Apartment? Score It
- Which job offer really pays more?Two Job Offers: Price the Tradeoffs
- Should you hold your price, or cut first?Price War: Hold or Cut?
- Can the panel agree on an order, if not on scores?Rank the Candidates
- Which office wins once you price the commute?Where Should the Office Be?
- What is staying put actually costing you?Switch Cloud Providers, or Stay?
- Should you announce your format before they do?Standards War: Which Format to Back?
- Neither side wants to blink. What happens?Deadline Standoff: Hold Firm or Concede?
- Big advance or double the royalty rate?Advance vs Royalty
- Ship at 80% confidence, or run one more validation round?Ship Now or Test More
- What pre-money valuation can your startup actually justify?What Is the Startup Worth to an Investor?
- Should you counter, or take the offer on the table?Counter or Accept
- Drill blind or pay to learn the odds first?Is the Test Worth Paying For?
- What is a pilot actually worth before you commit?Pilot First, or Go Big?
- Is the franchise worth giving up a steady salary?Franchise, Independent, or Keep the Job?
- Should you chase this tender at all?How Should Two Bidders Price One Tender?
- Is a pitch you probably lose still worth entering?Should We Pitch for This Account?
- Book the same room again or take the bigger one?Should We Run It Again Next Year?
- Is the bigger fee the better client?Which of Two Clients Should Get the Slot?
- What does outsourcing cost once you price the risks?Outsource or Keep It In House?
- Which tenders should you not price at all?Score the Sites Before You Bid
- Build the course yourself or license one?Is the New Course Worth Building?
- Tool up or keep paying the supplier?Make It or Buy It?
- Should you drop the slow line?Should We Drop This Line?
- What is the engagement worth once scope grows?Take the Engagement or Pass?
Try it in your own sheet
- Open Sortia in Google Sheets and choose Start from a template.
- Pick one of the models above, and it loads with the inputs filled in.
- Change the assumptions to fit your situation and press Run.
Never used Google Sheets? Start here goes the whole way, in seven steps, and assumes nothing.
Other methods: Monte Carlo Schedule risk Critical chain Optimization under uncertainty Statistics Machine learning Forecasting Optimization What-if analysis