You can now set Change Alerts to track relative shifts in your metrics. Instead of relying on fixed thresholds, these alerts notify you when a metric moves up or down by the percentage or amount you choose.
What You Can Do Now
Create alerts that trigger on % increases or decreases
Catch major swings like a 20% drop in signups or a 50% jump in errors
Use Change Alerts with Threshold Alerts to cover both relative and topline changes
Getting Started
In the left product menu, open Topline Alerts.
Create a new alert and choose your desired Condition Type.
When to Use Each Alert Type
Threshold - use to monitor against a fixed limit. ("Alert me when total daily signups drops below 1000")
Change - use to monitor absolute shifts. ("Alert me when daily signups drop by 200 compared to yesterday")
Change (%) - use to monitor percentage shifts ("Alert me when daily signups drop 20% compared to yesterday")
Sample Ratio Mismatch (SRM) happens when the share of users in experiment groups is different from what you expected. For example, if you set up a 50/50 split between control and treatment but the actual traffic is 60/40, that’s an SRM.
The SRM p-value is a statistical measure that tells you whether the observed imbalance could have happened by chance.
A p-value above 0.01 generally means the imbalance is within expected random variation.
A p-value below 0.01 suggests the imbalance is unlikely due to chance and may warrant investigation.
View SRM results and p-values across experiment groups in Metrics Explorer
Group results by different properties to identify potential causes of imbalance
Start from experiment exposure diagnostics and click on suggested properties to pre-apply them as group-bys in Metrics Explorer
Metrics Explorer applies the SRM formula across experiment groups and shows the resulting p-value. From there, you can add group-bys (such as country, platform, or custom properties) to spot where imbalance is happening.
Experiment diagnostics also highlight properties that may be driving the imbalance. Clicking the icon next to one of these properties takes you into Metrics Explorer with that property already grouped, so you can continue the investigation seamlessly.
This workflow makes it faster to detect and understand exposure imbalances. By moving directly from diagnostics to group-by analysis, you save time and get clearer visibility into which properties are linked to the imbalance.
Sample Ratio Mismatch debugging is available now across Cloud and Warehouse Native.
You can now add manual annotations directly to Drilldown charts in Metrics Explorer. This lets you document notable moments in your data and see them again whenever the same metrics are viewed.
Click any data point on a Drilldown chart to add a custom annotation
Apply an annotation to the metric you clicked, or extend it to additional metrics
See annotation icons whenever a chart’s date range and metrics overlap with saved annotations
Edit existing annotations, including description, date, time, and associated metrics
Each annotation is tied to a point in time and one or more metrics. When you load a Drilldown chart that includes both, an annotation icon appears. Click the icon to view or expand the note. You can adjust the description, date, time, and metrics at any point.
Annotations help you connect changes in the data to events in the real world. For example, you can tag the day a feature shipped or note an outage that caused a traffic dip. These markers appear on charts whenever the same metrics are analyzed, so you never lose the context.
Conversion Drivers are now available in Warehouse Native and Cloud. They surface the most significant factors influencing funnel conversions or drop-offs, helping you quickly understand why users convert or drop off.
Identify high-impact drivers of conversion or drop-off
Analyze event properties, user properties, and intermediary events
View summaries with conversion rate, share of participants, and impact
Drill into a driver for conversion matrices and correlation coefficients
Group funnels by any surfaced driver with one click
Conversion Drivers analyze columns from the metric source used in the first step of the funnel. For best results, configure your metric source as a multi-event metric source on the setup page and ensure all funnel steps come from that source.
From a funnel, click a step and select “View Drop-Off & Conversion Drivers.” You’ll see a ranked list of factors with conversion likelihood, conversion rates, and share of participants. Clicking into a factor opens detailed comparisons and lets you regroup the funnel by that property.
Funnels show what your conversion rate is. Conversion Drivers explain why, so you can investigate drop-offs, explore new funnels, and validate which user groups or behaviors matter most.
Conversion Drivers are available now for all Warehouse Native customers. For Cloud customers, read more about how Conversion Drivers work on Cloud.
We’ve expanded the Decision Framework feature beyond templates.
Now, you can directly configure and manage decision frameworks for each experiment. This gives teams a place to codify decision-making so that users can quickly move to action at the conclusion of an experiment.
To add a decision framework to your experiment select “Add Decision Framework” from the experiment menu.
You can now generate personal Console API keys in Statsig. These keys are automatically scoped to your role, ensuring the same access restrictions you already have. Each key is tied to its owner, making it easier to track usage and maintain clean audit logs.
Why it matters:
Simplifies multi-user projects by giving every user their own key
Provides clear ownership visibility for better security and compliance
Admins can control the ability to generate personal keys in the organization settings
We've added two more endpoints to our Console API for Dynamic Configs. Now you can archive and unarchive a Dynamic Config in your project programmatically!
Access the endpoints here: https://docs.statsig.com/console-api/dynamic_configs
You can now enable sampling in all major chart types to speed up queries on large datasets—while still getting directionally accurate results.
Use user-level sampling in Funnel, Distribution, Retention, and User Journey charts
Use event-level sampling in Metric Drilldown
Toggle sampling on or off in chart settings
See when sampling is active, and disable it at any time for exact results
Sampling is off by default. When toggled on, it only applies under high-volume conditions:
Warehouse Native: Sampling applies if metric sources exceed 100K rows/day or row counts can’t be determined. For User Journeys, sampling is always applied when toggled on.
Cloud: Sampling applies if the event volume in the query exceeds 100K. For Journeys, we look at total event volume across the company.
In Drilldown, event-level sampling is used for high-volume events unless the variance is too high, in which case we fall back to full data.
Sampling helps you move faster through exploratory workflows. In early results, User Journey query times dropped by over 60% when sampling was applied.
It’s a small precision tradeoff for a much faster iteration loop.
Experiment exposure events are now supported in Metrics Explorer on Warehouse Native. You can select them like any other event, filter or group by properties (variant, metadata), and tie rollout data directly to product metrics.
More details here: Exposures in Metrics Explorer
Admins can now mark specific cohorts and dashboards as verified. This signals that they are the trusted, official versions while also protecting them from accidental edits.
Mark cohorts and dashboards as verified to indicate they are the approved versions
Prevent edits to verified entities unless you are an admin
Clone verified cohorts and dashboards to create your own editable versions
Cohorts: Mark as verified when creating a new cohort or by editing an existing one
Dashboards: Mark as verified from the settings cog in the top right of the dashboard page
Teams can align on a single source of truth for key cohorts and dashboards while still allowing individuals to explore their own versions without risking changes to the verified originals.
This keeps shared analysis reliable and consistent.