Skip to content

Safety and risk analysis

Decision tree analysis for choices under uncertainty

Lay out decisions, uncertain outcomes and payoffs as a tree, and let rollback recommend the option with the best expected value, expected utility or lowest cost. Check how robust the choice is with sensitivity analysis and Monte Carlo.

Overview

A decision tree models a choice under uncertainty. Decision nodes are the options you control, chance nodes are uncertain events with probabilities, and terminal nodes hold the payoff or cost of each path. Rolling the tree back from the payoffs gives the expected value of each option and the recommended policy.

ModelGraphX evaluates the tree live as you edit. You can maximise value or minimise cost, apply per-branch arc values and discount factors, and use expected utility with an exponential utility function when risk attitude matters. Helpers normalise probabilities and fill in the remaining probability on a branch set.

Sensitivity analysis shows which inputs change the decision. Monte Carlo samples uncertain inputs and reports how often each option is optimal, so you can see whether a recommendation is stable or a close call.

Capabilities

Decision, chance and payoff nodes

Build trees from decision nodes, chance nodes and terminal payoff nodes.

Rollback

Compute expected values with arc values and discount factors, maximising value or minimising cost.

Expected utility

Use an exponential utility function with a risk-tolerance parameter for risk-averse decisions.

Probability helpers

Normalise, set the remaining probability or auto-balance a branch set.

Live recommendation

See the tree summary, the recommended policy and the best path highlighted on the canvas as you edit.

Sensitivity analysis

Vary one input at a time to see which assumptions change the decision.

Monte Carlo

Sample point, triangular, normal or beta inputs and get mean, p10, p50, p90 and decision stability.

Monte Carlo sensitivity

Rank inputs with standardised regression coefficients, Spearman rank and quantile measures.

Saved runs and dashboard

Reopen saved calculation runs and review results on a dashboard.

Reports

Report nodes, branches, rollback, the recommended strategy and sensitivity, exported to Word, PDF or PowerPoint.

Risk Registry links

Publish nodes to the Risk Registry and link them to a hazard scenario.

Templates

Start from simple or advanced templates and export diagrams as images or PDF.

Standards and methods

  • Decision analysis

ModelGraphX supports work done under these standards and methods; it does not certify compliance.

Who uses it

  • Risk managers comparing mitigation options
  • Engineers choosing between designs or maintenance strategies
  • Project and investment decision makers
  • Students of decision analysis

Works with

  • Risk Registry

    Decision outcomes can be published to the Risk Registry.

  • Hazard Management

    Nodes link to the hazard scenarios they address.

  • Event Tree Analysis

    Event trees quantify accident sequences; decision trees compare the options for dealing with them.

Frequently asked questions

What is rollback in a decision tree?
Rollback evaluates the tree from the payoffs back to the root: chance nodes take the probability-weighted average of their branches, and decision nodes take the best branch. The result is the expected value of each option and the recommended policy.
Can I model risk aversion?
Yes. Switch the objective to expected utility with an exponential utility function and set a risk tolerance, so options with large downside are penalised.
How do I know if the recommended decision is robust?
Run one-at-a-time sensitivity analysis, or Monte Carlo, which reports how often each decision is optimal across the trials. Monte Carlo depends on your plan.
Does AI draft decision trees?
Not in this module. You can ask the platform Agent to review a tree or explain its results, but decision trees are built by you.

Try it on a real project

Explorer is free, and new accounts get a 30-day trial of the paid capabilities.

All modules
© Archer Futura Technologies AB