A number of customers have called out that volume monitoring in Databricks -- specifically when using byte-count -- is alerting too often. Specifically, there are too many alerts for small size decreases related to optimize/vacuum operations on these tables.

We've shipped some adjustments to dramatically reduce these undesirable alerts. The changes should reduce the false positives by about 75%, while only reducing the overall number of byte-count alerts by about 16%. A very targeted improvement!

For many enterprises, attributing the cost of Monte Carlo back to their lines of business is a necessary step. This process, commonly called a "chargeback," comes in 2 flavors:

1. Charging back the consumption of Monte Carlo credits to the lines of business. This is relevant for our customers on consumption pricing plans, and we released much better support for this a few months ago (detailed here).
2. Charging back the consumption of warehouse resources from Monte Carlo to the lines of business. We're releasing much better support for this today.

Two big improvements we've shipped today, that align with #2 above:

• monitor_id as a query tag. When a monitor runs, Monte Carlo will pass a query tag containing the monitor_id, so that warehouse consumption can be easily attributed back to a specific monitor. For now, this is limited just to Snowflake & BigQuery.
• Limit the connections available to a user via Authorization Groups. Admins have long been able to create multiple connections to their data sources. But now, they can control which users have access to use each connection via Authorization Groups. This makes it much easier to track the total warehouse consumption from Monitors or Data Profiler coming from a particular team.

Read more about both of these improvements here.

Note, these improvements are limited to our Enterprise product tier.

In the last few quarters, the ability to select training data has become a widely adopted way that customers tune the ML thresholds in our monitors.

Many power users have noted the challenges of using this feature within segmented Metric Monitors. Specifically, the selections they make would be applied to all segments within the monitor, with no ability to limit it just to a single segment. This made it hard to fine-tune individual segments... a frustrating limitation.

We've now released controls for users to pick if their selections should be applied to all segments or just this segment. We've also released similar functionality for Custom SQL monitors that use variables, as well.

Several weeks ago, we released a series of changes that tightened the thresholds when monitoring the volume (row count) of a table.

In some cases, these made thresholds more sensitive or noisy than desirable, putting some customers at risk of alert fatigue. Today, we released adjustments to the model to reduce much of that noise. In many cases, this will cause volume thresholds to moderately widen / loosen.

Slack users can now initiate Monte Carlo's Troubleshooting Agent directly from Slack alerts. Upon completion, a troubleshooting summary is provided along with a link to open the findings in Monte Carlo for further investigation!

• One-click troubleshooting: "Troubleshoot alert" button on all Monte Carlo Slack alerts • Thread-based interactions: Agent responses stay organized in the alert thread • Real-time status updates: Clear feedback when troubleshooting starts and completes • Seamless handoff: Summary results in Slack with "Open in Monte Carlo" link for detailed findings

Collaboration just got easier! You can now tag teammates directly in alert comments using @username to bring them into the conversation.

When you mention someone, they’ll receive an email notification with a link to the alert - making it simple to jump in, review context, and take action.

Good news! Navigation in Monte Carlo just got easier. You will have access to all your tabs on a left-side bar in the new experience. Watch a video walk-through

How it works, top to bottom:

• Expand / collapse: use the chevron buttons next to the domain selector.
• Create monitor: one-click action to create a monitor from any part of the product.
• Key product pages: follows the same ordering you are accustomed to.
• Bottom panel: access your user profile menu, onboarding and notifications.

Agent instructions enables you to provide custom context about your data platform that troubleshooting agent can leverage during alert analysis.

Navigate to Settings → AI Agents to add custom context about your data platform.

Use-cases might include:

• Platform-specific rules: Define naming conventions, data sources, and architectural patterns • Custom context: Share information about your ETL tools, schema organization, and data lineage • Shared workspace instructions: Context is shared across your entire Monte Carlo account, not just individual users

Examples:

• “All tables that start with ‘sfdc_’ come from Salesforce” • “We use dbt core to run all tables in the ‘prod’ schema but the dbt integration is not yet connected to Monte Carlo” • “Our staging tables are refreshed every 4 hours during business hours only”

You can now manually trigger metric monitors either from the UI (via the Run button) or through the API. This gives teams flexibility to run monitors exactly when their data pipelines finish loading, avoiding wasted compute from partial runs and ensuring results reflect the latest data.