> For the complete documentation index, see [llms.txt](https://docs.algenta.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.algenta.ai/guides/what-if.md).

# Run a what-if sweep

A what-if sweep varies one variable across a range while holding everything else fixed, then runs a full simulation at each step — in parallel — so you can see exactly how the outcome moves. `POST /v1/simulate/what-if` returns the sweep plus a computed sensitivity direction and elasticity, the numbers behind a tornado chart. For long simulations, `POST /v1/simulate/stream` streams progressive results over Server-Sent Events so you can show live progress instead of waiting for the final answer.

{% hint style="info" %}
What-if operates on **expert-mode** simulation requests (`"mode": "expert"`), because it needs named variables with explicit distribution parameters to sweep. See [Simulation models](/concepts/simulation-models.md) for the full expert schema.
{% endhint %}

## Before you start

You need:

* An API key sent as `Authorization: Bearer $ALGENTA_API_KEY` on a verified account.
* An expert simulation you want to probe: a `variable` name and the `param` of its distribution to sweep (`mean`, `std`, `low`, `high`, `mode`, or `value`).

## Sweep one variable

{% stepper %}
{% step %}

### Build the sweep request

The body wraps a base expert simulation in `request`, then names what to vary:

| Field           | Default | Meaning                                                                              |
| --------------- | ------- | ------------------------------------------------------------------------------------ |
| `request`       | —       | The base expert `SimulateRequest`.                                                   |
| `variable`      | —       | Name of the variable to sweep.                                                       |
| `param`         | `mean`  | Which distribution parameter to move: `mean`, `std`, `low`, `high`, `mode`, `value`. |
| `range_pct`     | `0.30`  | Sweep width as a fraction of the base value (`0.30` = ±30%).                         |
| `n_steps`       | `11`    | Number of steps (3–51; odd is symmetric around the base).                            |
| `runs_per_step` | `5000`  | Monte Carlo runs per step (100–100,000).                                             |
| {% endstep %}   |         |                                                                                      |

{% step %}

### Send it to `/v1/simulate/what-if`

{% tabs %}
{% tab title="cURL" %}

```bash
curl -sS -X POST "https://api.algenta.ai/v1/simulate/what-if" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "request": {
      "mode": "expert",
      "simulation": {
        "variables": [
          {"name": "revenue", "distribution": "normal", "params": {"mean": 120000, "std": 30000}},
          {"name": "cost",    "distribution": "fixed",  "params": {"value": 70000}}
        ],
        "objective_function": "revenue - cost"
      },
      "runs": 5000
    },
    "variable": "revenue",
    "param": "mean",
    "range_pct": 0.3,
    "n_steps": 9,
    "runs_per_step": 5000
  }'
```

{% endtab %}

{% tab title="Python" %}

```python
import os
import requests

BASE = os.environ.get("ALGENTA_API_URL", "https://api.algenta.ai")

resp = requests.post(
    f"{BASE}/v1/simulate/what-if",
    headers={"Authorization": f"Bearer {os.environ['ALGENTA_API_KEY']}"},
    json={
        "request": {
            "mode": "expert",
            "simulation": {
                "variables": [
                    {"name": "revenue", "distribution": "normal", "params": {"mean": 120000, "std": 30000}},
                    {"name": "cost",    "distribution": "fixed",  "params": {"value": 70000}},
                ],
                "objective_function": "revenue - cost",
            },
            "runs": 5000,
        },
        "variable": "revenue",
        "param": "mean",
        "range_pct": 0.3,
        "n_steps": 9,
        "runs_per_step": 5000,
    },
)
resp.raise_for_status()
sweep = resp.json()

print(sweep["sensitivity_label"], "elasticity=", sweep["elasticity"])
for step in sweep["steps"]:
    print(step["label"], "EV=", round(step["mean"]), "PoL=", step["probability_of_loss"])
```

{% endtab %}

{% tab title="TypeScript" %}

```typescript
const base = process.env.ALGENTA_API_URL ?? "https://api.algenta.ai";

const res = await fetch(`${base}/v1/simulate/what-if`, {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.ALGENTA_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    request: {
      mode: "expert",
      simulation: {
        variables: [
          { name: "revenue", distribution: "normal", params: { mean: 120000, std: 30000 } },
          { name: "cost",    distribution: "fixed",  params: { value: 70000 } },
        ],
        objective_function: "revenue - cost",
      },
      runs: 5000,
    },
    variable: "revenue",
    param: "mean",
    range_pct: 0.3,
    n_steps: 9,
    runs_per_step: 5000,
  }),
});
if (!res.ok) throw new Error(`what-if failed: ${res.status}`);
const sweep = await res.json();
console.log(sweep.sensitivity_label, sweep.elasticity);
```

{% endtab %}
{% endtabs %}
{% endstep %}

{% step %}

### Read the sweep

The `200` `WhatIfResponse` gives one `step` per point, plus a summary of how much the outcome reacts:

```json
{
  "variable": "revenue",
  "n_steps": 9,
  "range_low": 84000,
  "range_high": 156000,
  "param_changed": "mean",
  "steps": [
    {
      "step": 0,
      "value": 84000,
      "label": "revenue=8.4e+04 (-30.0%)",
      "mean": 14000, "std": 29900,
      "p5": -35000, "p25": -6000, "p50": 14000, "p75": 34000, "p95": 63000,
      "probability_of_loss": 0.32,
      "score": 0.51
    }
  ],
  "increasing_direction": "better",
  "sensitivity_label": "High sensitivity",
  "elasticity": 1.02
}
```

* `steps` — mean, standard deviation, five percentiles, and probability of loss at each swept value.
* `increasing_direction` — whether increasing the variable helps (`better`), hurts (`worse`), barely moves the outcome (`insensitive`), or is non-monotonic (`mixed`).
* `sensitivity_label` — a plain-English band from `Insensitive` through `Very high sensitivity`.
* `elasticity` — percent change in expected value per percent change in the variable.
  {% endstep %}
  {% endstepper %}

## Stream a long simulation live

`POST /v1/simulate/stream` runs a normal simulation but emits results as they compute, over `text/event-stream`. You get named `progress` events with online statistics, then a single `final` event carrying the same `DecisionEnvelope` a normal `/v1/simulate` call returns. Set `checkpoint_every` (query parameter, default `1000`, minimum `100`) to control how often progress fires. The request body is a plain `SimulateRequest` — auto or expert.

{% tabs %}
{% tab title="cURL" %}

```bash
curl -N -X POST "https://api.algenta.ai/v1/simulate/stream?checkpoint_every=1000" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "expert",
    "simulation": {
      "variables": [
        {"name": "revenue", "distribution": "lognormal", "params": {"mean": 11.5, "std": 0.4}}
      ],
      "objective_function": "revenue"
    },
    "runs": 200000
  }'
```

{% endtab %}

{% tab title="TypeScript" %}

```typescript
const base = process.env.ALGENTA_API_URL ?? "https://api.algenta.ai";

const res = await fetch(`${base}/v1/simulate/stream?checkpoint_every=1000`, {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.ALGENTA_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    mode: "expert",
    simulation: {
      variables: [{ name: "revenue", distribution: "lognormal", params: { mean: 11.5, std: 0.4 } }],
      objective_function: "revenue",
    },
    runs: 200000,
  }),
});

const reader = res.body!.getReader();
const decoder = new TextDecoder();
for (;;) {
  const { value, done } = await reader.read();
  if (done) break;
  process.stdout.write(decoder.decode(value)); // raw SSE frames
}
```

{% endtab %}
{% endtabs %}

The wire is standard SSE: a heartbeat comment, repeated `progress` events, then one `final` event.

```
: stream-started

event: progress
data: {"type":"progress","runs_completed":1000,"runs_total":200000,"progress":0.005,"mean":98123.4,"std":40211.2,"probability_of_loss":0.07}

event: progress
data: {"type":"progress","runs_completed":2000,"runs_total":200000,"progress":0.01,"mean":98410.2,"std":40088.1,"probability_of_loss":0.068}

event: final
data: {"type":"final","result":{ /* full DecisionEnvelope */ }}

: stream-end
```

Each `progress` event carries online (Welford) statistics — `mean`, `std`, `probability_of_loss`, `runs_completed`, and `progress` (0–1). The `final` event's `result` is the complete decision envelope; close the connection once you receive it.

{% hint style="success" %}
**Expected result** — `/v1/simulate/what-if` returns a `WhatIfResponse` whose `steps` length equals `n_steps` and whose `sensitivity_label` and `elasticity` summarize the reaction. `/v1/simulate/stream` streams one or more `progress` events followed by exactly one `final` event containing the full `DecisionEnvelope`.
{% endhint %}

## Next steps

<table data-view="cards"><thead><tr><th></th><th></th><th data-hidden data-card-target data-type="content-ref"></th></tr></thead><tbody><tr><td><strong>Recommend an action</strong></td><td>Rank named actions and get a single recommendation.</td><td><a href="/pages/AharbnfV1rQPPkNynKqR">/pages/AharbnfV1rQPPkNynKqR</a></td></tr><tr><td><strong>Simulate a scenario</strong></td><td>Run a single decision simulation end to end.</td><td><a href="/pages/xo3kyzyPZXFb0ETopz2h">/pages/xo3kyzyPZXFb0ETopz2h</a></td></tr><tr><td><strong>Simulation models</strong></td><td>Distributions, objectives, and expert-mode options.</td><td><a href="/pages/kzBRwbzvypUVd7HsDqzw">/pages/kzBRwbzvypUVd7HsDqzw</a></td></tr></tbody></table>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.algenta.ai/guides/what-if.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.