Explanation

Explanation reads the Digital Twin to answer why, not what-if. It takes a relationship that already exists in the model and decomposes it: how strongly one variable drives another, which drivers matter most, and the pathways an effect travels along. The product sums it up in one line: understand drivers and impacts for outcomes.

Where a prediction tells you an outcome, an intervention tells you what one change would do, and an optimization finds the best combination to aim for, Explanation steps back and accounts for the causal structure itself. Nothing is changed and no new scenario is run — the tool reads the trained twin and explains it.

For the workflow that produces a Digital Twin in the first place, see Step 5: Build Digital Twin.

---

Overview

Open a Digital Twin from the Digital Twins list, open the Simulate tab, click New Simulation, and choose Explanation from the type picker.

The Explanation setup opens on the mode step: Directional, Discovery, or Impact. The mode chosen here decides which variables the next step asks for.

Every Explanation run follows the same five steps.

1. Pick the mode. The mode frames the question and decides which variables you supply next:

• Directional — explain a specific cause → effect relationship.

• Discovery — find the strongest drivers of an effect.

• Impact — see what outcomes a cause influences.

2. Select the variables. A Cause variable is the lever; an Effect variable is the outcome. The mode decides which you need — Directional asks for both, Discovery for the effect alone, Impact for the cause alone.

3. Add segments (optional). A segment is a filter — one or more conditions, such as tier = premium — that the simulation runs separately, so you can compare how an explanation differs across subgroups. Suggest Segments proposes them from the twin; + Add segment defines one by hand. Leave the section empty to run across the whole population.

4. Run the simulation. The Configuration Summary restates the run as a plain-English Research question. Validate checks the configuration; Run Simulation launches it against the chosen twin version.

5. Review the results. A finished run shows a green Completed pill, a duration, Export PDF and Edit Config, and opens with the Configuration Summary and an AI Summary. Everything below that depends on the mode.

---

Detailed explanation

The mode chosen in step 1 changes both the inputs you supply and the result you get back. Each of the three is shown below with its own setup (input) and result (output).

Option 1 — Directional

Explain a specific cause → effect relationship. Use Directional when you already have a hypothesis — "does Contract affect Churn?" — and want it quantified. It asks for both a Cause variable and an Effect variable.

Input. Both a Cause variable (Contract) and an Effect variable (Churn) are set; Step 3 Segments is left empty to run across the whole population.

The result leads with Causal Relationship Analysis — a headline card for the pair, here Contract → Churn: a sustained Contract change moves the probability of Churn: Yes from 22.7% to 35.5% (+12.7 percentage points), with a confidence rating, a result-reliability note, and an effect range. Below it, Categorical Outcome Analysis breaks the shift down by outcome value as Delta Probabilities. Causal Pathways then decomposes the total effect into a Direct Effect and the ranked indirect paths it travels (Path 1, Path 2, …), each with its share and effect size, drawn as a Pathway Contributions donut. A Segment Analysis section closes the page.

Output. The headline card quantifies the single relationship, Contract shifting Churn: Yes from 22.7% to 35.5%, with the AI summary spelling out the effect size and its significance above it.

Deeper in the same result: Causal Pathways attributes 97.8% of the effect to the direct path, and Segment Analysis breaks the effect out by contract type.

Option 2 — Discovery

Find the strongest drivers of an effect. Use Discovery when you know the outcome you care about but not what moves it most. It asks only for the Effect variable.

Input. Discovery needs only the Effect variable (Churn); the Segments section is identical across all three modes.

The result is a Key Driver Analysis — a ranked bar chart of each driver's Relative contribution (%) to the Causal Effect on Churn, with Contract, InternetService, and OnlineSecurity topping the list.

Beneath the chart, every driver is a row — Driver, Rank, Contribution, Range, Confidence — and each one expands into the same deep analysis a Directional run produces for a single pair. Discovery is, in effect, Directional run across every driver at once and ranked.

Output. The ranked drivers: each row carries its contribution, range, and confidence, and expands into the full pathway breakdown.

Option 3 — Impact

See what outcomes a cause influences. Use Impact when you know the lever and want to trace everything downstream of it. It asks only for the Cause variable.

Input. Impact needs only the Cause variable (Contract).

The result is an Impact Analysis — a Downstream Effects chart plotting the Causal impact from the chosen cause across the variables it reaches, and an Affected Variable table giving each one a Contribution (split into direct and indirect), a Range, and a Confidence rating. It is the mirror image of Discovery: one cause, many effects, rather than one effect, many causes.

Output. Contract's downstream footprint is dominated by TotalCharges at 98.3%; every other affected variable, including Churn, carries a sliver of the total impact.

---

Past simulations

Past runs appear in the Simulate tab's run list inside the twin, and View all simulations on the twin's Home opens the full history. The workflow is the same as for any other simulation type — see Prediction › Past simulations.

---

Other Simulation Types

• Prediction — predict an outcome for a specific input.

• Intervention — change a single variable and observe propagation.

• Optimization — find the input combination that maximizes or minimizes a target.

• Best Action — find the minimum change needed to reach a target outcome.

• Root Cause Analysis — diagnose the cause of a specific abnormal value.

• Anomaly Scan & Diagnosis — scan every variable for anomalies and diagnose each one.

See Step 6: Run Simulations — general overview.