# Pulse (Beta)

> Pulse lets you describe, in plain language, what to watch in your data. Sisense turns it into a saved analysis that runs on a schedule, records each result, and notifies you when your condition is met.

*Source: https://docs.sisense.com/main/SisenseLinux/pulse.htm*

---

# Pulse (Beta)

Last updated: September 23, 2026

|  |  |
| --- | --- |
| [Tier](https://www.sisense.com/pricing/#pricing) | [Deployment](https://docs.sisense.com/main/SisenseLinux/introduction-to-sisense-cloud-managed-services.md#ComparisonofManagedCloudandSelfHosted) |
| Launch     Grow    Enterprise | Cloud     On-Prem |

**Availability:** Sisense 2026.4.0 and later · Beta.

- Pulse requires Sisense Intelligence with a configured LLM. Administrators control availability on the **Sisense Intelligence** admin page.

Pulse lets you watch something in your data and get notified when it changes, without needing to know in advance which number to watch.

You describe what you care about in your own words, and Sisense turns it into a **Pulse**: a saved analysis that runs on a schedule, records what it finds each time, and notifies you when your condition is met.

Classic Pulse alerts continue to work unchanged alongside Pulses.

## Prerequisites

- **Sisense Intelligence must be enabled** with a configured LLM. See [Setting Up Your LLM](https://docs.sisense.com/main/SisenseLinux/llm-setup.md).
- **You need Designer or Admin permissions** to create a Pulse.
- **You must have access to the data** you want to watch. You can create a Pulse only over data you can already see, and the Pulse respects that access every time it runs.

## Creating a Pulse

To create a Pulse:

1. Open the [Assistant](https://docs.sisense.com/main/SisenseLinux/ai-assistant.md).
2. Describe what you want to watch, in plain language. For example:

   - *"Tell me when weekly revenue in any region drops more than 10% from the week before."*
   - *"Let me know if a product falls out of our top 5 by units sold."*
   - *"Watch whether actual spend goes more than 15% over plan."*
3. The assistant compiles your description into a Pulse and shows you what it built: the data it will query, what it will look for, and how often it will run.
4. **Review the Pulse before you confirm it.** Check the measure, filters, condition, and schedule. This is the cheapest point at which to fix a misunderstanding.
5. If something is wrong, say so in the conversation (for example, *"use last year's data instead"* or *"check it daily"*), and the assistant revises the Pulse.
6. Confirm. Sisense creates the Pulse and runs it immediately.

**Note:**

**The first run usually can't reach a conclusion.** Most Pulses compare against what they saw previously, so the first run records a starting point rather than a result. The run history states this explicitly.

### When Sisense Cannot Build Your Request

Some requests fall outside what a Pulse can currently watch. When that happens, the assistant tells you directly and suggests the closest alternative it can build, rather than creating something that only approximates your request.

## What a Pulse Can Watch

A Pulse can watch for the following patterns:

| Pattern | Example |
| --- | --- |
| **A value crossing a line** | *"When weekly revenue drops below $50K"*, optionally with a breakdown of what drove the change |
| **A ranking changing** | *"When a region moves out of the top 5"* |
| **Actuals against plan** | *"When spend goes 10% over budget"* |
| **A trend** | *"When growth turns negative"* |
| **A forecast** | *"When the projection for next month falls short of target"* |
| **An anomaly** | *"When transactions look unusual compared with recent weeks"* |

You can combine a small number of these patterns in one Pulse. For example, you can watch for a value crossing a line together with a breakdown of which category moved most.

## Understanding the Results

Each run produces one of four outcomes:

| Outcome | Meaning |
| --- | --- |
| **Met** | Your condition was satisfied, and a notification is sent. |
| **Not met** | The Pulse ran, and your condition was not satisfied. |
| **Inconclusive** | The Pulse ran, but the data wasn't sufficient to reach a reliable conclusion. |
| **No data** | The query returned nothing to evaluate. |

**Inconclusive is a real answer, not an error.** For patterns that fit a line to your data (trend and forecast), Sisense calculates the result together with its confidence range and compares that range against your threshold:

- If the range sits entirely past your threshold, the condition is met.
- If the range sits entirely short of your threshold, the condition is not met.
- If the range straddles your threshold, the change can't be distinguished from normal variation. The Pulse reports **Inconclusive** rather than presenting an unreliable number as a finding.

When a result is based on a fitted line, the notification can include the confidence range and the number of data points used. For example: *"growth of 5.3% per week (95% CI: 4.1–6.5%)"*.

If your data can't support the question you asked, the assistant tells you **while you're setting up the Pulse**, not after it has been running for a month. You can still proceed if you choose.

### Viewing History

Sisense records every run, whether or not it met the condition. To see each run, its outcome, and the values behind it, open the Pulse from the **Pulse** page and select **History**.

## Scheduling

How often a Pulse runs depends on the data it watches:

- **ElastiCube data**: The Pulse runs after each successful build.
- **Live models**: The Pulse runs on a schedule you set. The minimum interval is **15 minutes**.

In Beta, you can also set **cron-style schedules** for more specific timing, and **pause** a Pulse for a period without deleting it.

Sisense excludes incomplete periods automatically. A Pulse that compares weekly figures uses the last completed week, so it never compares a partial current week against full ones.

You can also run a Pulse on demand at any time to see what it would find right now. On-demand runs don't affect what scheduled runs compare against.

## Notifications

When a Pulse meets its condition, it sends a notification that includes what was measured, the values behind the result, and a plain-language explanation.

The available channels are the same as for classic Pulse alerts: the in-product notification center, email, mobile, Slack, Zapier, and webhooks.

**Note:**

Configure Slack, Zapier, and webhook destinations on the Pulse itself after you create it, using the same **Notifications** settings as any other Pulse alert. You can't set them up while describing the Pulse to the assistant.

## Managing Your Pulses

All Pulses appear on the **Pulse** page alongside classic alerts. Pulses created through the assistant carry a distinct icon.

From the **Pulse** page, you can:

- See the current value and the result of the most recent run.
- Open **History** to review every run.
- Select **Turn Off** to stop a Pulse without deleting it.
- Delete a Pulse.

For more information about managing tiles on the Pulse page, see [Managing Tiles in Sisense Pulse](https://docs.sisense.com/main/SisenseLinux/managing-tiles-in-sisense-pulse.md).

In Beta, the tile also shows the query definition and schedule, so you can see what a Pulse does without opening the assistant.

To change what a Pulse watches, ask the assistant to revise it.

**Note:**

Changing the underlying query resets the comparison history, so the next run records a new starting point.

## Limitations in Beta

- You can create and revise a Pulse only through the assistant. There is no in-product editor and no entry point from a dashboard or widget.
- You configure Slack, Zapier, and webhook destinations after you create a Pulse, not during creation.
- External systems cannot query a Pulse or receive its outcomes programmatically. This capability is expected to be included in the GA release.
- Patterns that compare against previous runs need history before they can reach a conclusion. Anomaly detection in particular needs several runs before its results are meaningful.
- A notification may occasionally be delivered more than once for the same occurrence.
