ModelChoice is a professional decision-analysis add-in for Microsoft Excel. This page is the comprehensive feature reference: every analytical report, every authoring capability, every integration. For the product overview, see the ModelChoice main page. For the trial download, scroll to the bottom of this page or use the menu.
How you build a model and watch it solve itself.
ModelChoice draws decision trees as native Excel shapes alongside data cells on the same worksheet. Click to insert a starter tree; add decision options, chance branches and terminal payoffs through a node-editor dialog or by clicking on existing nodes. Branch values, probabilities and payoffs are real Excel cells — link them to any cell in any workbook. There is no separate tree file or external editor.
Every time you change an input — a probability, a branch value, a linked cell — ModelChoice re-solves the tree using backward induction and refreshes the displayed expected values. The optimal decision path is highlighted in real time. There is no "run" button: rollback recalculates the way an Excel formula does.
Every tree edit is undoable. Add a branch by mistake, change a probability, paste a subtree from the wrong place — Ctrl+Z reverses each step, and you can re-apply with Ctrl+Y. The full edit history is preserved per tree, separate from Excel's own undo stack, so experimentation is risk-free.
ModelChoice treats a subtree (a node plus all its descendants) as a first-class object. Copy a subtree from one tree and paste it onto a terminal in another. Append a symmetric subtree of decisions or chances under any terminal. Replace one subtree with another in place. The data cells that back each node travel with it.
Right-click on any tree shape — decision, chance, terminal — for a context menu of relevant actions: edit, add branch, remove, collapse/expand, copy subtree, paste subtree. The menu is shape-aware (different options for different node types) and works exactly like Excel's native right-click experience.
Every angle on the same model, from a single ribbon click.
The core analysis: backward induction from terminals to root, producing the optimal strategy and its expected value. Runs continuously as inputs change. The basis on which every other report builds. Optionally substitutes utility values for raw EV when a utility function is configured.
Most spreadsheet decisions are made on the mean alone. The Risk Profile report shows the full probability distribution of outcomes for each strategy — every possible payoff and its probability. The Cumulative Risk Profile gives the CDF view with percentiles. Side-by-side comparison of strategies on their tails is the language that risk-averse stakeholders and regulators understand.
Vary each input one at a time over a defined range, plot the result. Generates tornado diagrams ranking inputs by impact, spider charts showing the path of EV across the variation range, and kink-detection plots identifying where the optimal strategy flips. Use it to focus refinement effort on the inputs that actually move the answer.
Vary two inputs simultaneously to see how they interact. Output is a strategy-region chart: a 2-D grid coloured by which strategy is optimal at each (x, y) combination. Use it when you suspect two assumptions move together — e.g. price and volume, success probability and timing.
How sure are we? For every input, ModelChoice finds the exact threshold at which the optimal strategy flips to a different choice ("flip distance"). A 0–100 robustness score summarises overall fragility. Probabilistic-reversal Monte Carlo gives the probability that the recommendation is wrong given joint uncertainty across all inputs. The six-sheet report is the defensible answer to the boardroom question "how confident are we in this?"
Sometimes the question is the other way round: what would have to be true for a different decision to become optimal? Force to Outcome searches the input space for single-parameter and multi-parameter recipes that make a specified strategy optimal (or hit a target EV). Useful when a decision is politically pre-determined and the analyst needs to surface the implicit assumptions behind it.
Enumerate and rank every feasible strategy by expected value, showing the gap from the optimal choice. Useful when stakeholders have a favourite that isn't the EV-maximising option — quantifies exactly how much it costs. The accompanying Policy Tree report renders just the optimal subtree, with non-chosen branches stripped, for clean executive presentation.
A flat, audit-ready enumeration of every strategy in the tree with its complete decision path and expected value. Designed for regulated environments where every option must be visible and traceable. Exportable as a separate workbook for archival.
The Expected Value of Perfect Information (EVPI) is the maximum amount you should ever pay to eliminate all uncertainty before deciding — it caps what any study, pilot or expert opinion could be worth. The Expected Value of Imperfect Information (EVII) gives the realistic value of a specific test with known accuracy. Together they answer the go / no-go on commissioning research before commitment.
A single-page summary fusing rollback result, sensitivity headlines, risk-profile percentiles, robustness score, and value-of-information findings into a board-ready narrative. Generated in seconds; rerun in seconds when an input changes. Designed to be copied or screenshotted straight into a steering-committee deck.
When a decision is more than just the mean.
Not every decision is a dollar decision. ModelChoice's MCDA mode lets you blend the financial payoff with weighted non-financial criteria — clinical efficacy, quality of life, ESG score, technology fit. Each criterion has a weight and an ordinal scale of discrete options (Poor / Fair / Good / Excellent). The composite score on a 0–1 scale is computed at every terminal and rolled back through the tree.
Defensible weights are derived from stakeholder pairwise comparisons using Saaty's 1–9 scale. ModelChoice's Analytic Hierarchy Process workflow walks the analyst through every pairwise comparison, then computes the weights from the matrix's principal eigenvector. A consistency ratio flags inconsistent stakeholder responses — the matrix is rejected if the ratio exceeds the standard 0.1 threshold.
Expected value assumes risk-neutrality. Real decision-makers are not. ModelChoice supports four classical utility functions — exponential, logarithmic, square-root, linear — for capturing risk-averse and risk-seeking behaviour. The analyst sets a risk-tolerance parameter; ModelChoice computes certainty equivalents (the guaranteed sum equally attractive to a risky strategy) and risk premiums (the implicit cost of bearing the risk).
Monte Carlo simulation via ModelRisk.
Decision trees model uncertainty using discrete branches. For variables that are better described continuously — oil price, demand, project duration, interest rate — ModelChoice integrates natively with Vose Software's ModelRisk, replacing a chance branch with a continuous distribution (VoseNormal, VosePERT, VoseLognormal, VoseEmpirical, and dozens more). Thousands of iterations through the tree produce a true probabilistic answer.
Expected Value Tracking gives a fast estimate of the optimal strategy and its EV under continuous uncertainty. Sampled Paths draws a complete random path through the tree each iteration and reports the realised payoff — the full empirical distribution of outcomes. Perfect Information Bounds computes EVPI under continuous uncertainty by always picking the truly best decision given each iteration's resolved outcomes.
Native Excel, JSON, PrecisionTree migration, automation API.
Organisations using PrecisionTree can import existing models with a single Import command. ModelChoice reads the PrecisionTree workbook directly, reconstructs the tree, and re-renders it with the broader analytical suite immediately available. The migration is the principal path for teams whose PrecisionTree licensing is changing or who simply want deeper analytical capability on a modern .NET 9 codebase.
Every aspect of a tree — structure, probabilities, payoffs, settings, MCDA configuration, AHP matrix, per-terminal criterion ratings — is fully serialisable to JSON. Export a tree, store it in version control, share it with a colleague, diff two versions, generate trees programmatically from a database. The JSON format is the foundation for any automation pipeline.
Every major analysis has a headless _Auto command callable from any COM-aware host via Application.Run. Build trees programmatically (or from JSON), run sensitivity, MCDA, robustness, Decision Brief — all without user interaction. Use it for batch portfolio analysis, regression testing, internal applications, or embedding ModelChoice in larger workflows. Excel UDFs expose every node value, probability, strategy EV, composite score, and criterion weight as a callable formula function.
Every significant value in a tree is exposed as a named range (MC_V_<nodeId>, MC_P_<nodeId>, etc.) and every formula is inspectable. No proprietary opaque file format, no black box. Sheet protection locks down completed models for distribution. The audit trail lives inside the workbook, alongside the data.
Languages, example models, documentation.
Full localisation in English, German, Spanish, French, Japanese, Portuguese, and Russian. Ribbon labels, dialog text, status messages, error messages, and the complete documentation site are all translated. Switch languages from Application Settings; the change applies immediately.
ModelChoice ships with approximately sixty fully-built example models drawn from real decision-analysis applications across every industry: oil and gas, pharma and life sciences, finance and investment, infrastructure and real estate, manufacturing and supply chain, technology, agriculture, public sector. Each example is a working tree — open, run any analysis, modify and adapt.
Tutorials for every analysis type. Reference pages for every UDF and ribbon command. Conceptual guides for decision analysis, MCDA, AHP, utility functions, and Monte Carlo simulation. A complete Automation API reference with Python, C# and VBA examples. An exact JSON file-format specification. All available both inside Excel (via the Help ribbon) and as a standalone documentation site.
The 15-day free trial is the complete product — every analysis, every example model, no feature restrictions.
Or — back to the ModelChoice overview › Request an online demonstration ›