✋ Holdout Analysis in Product Analytics

You can now filter or break down any Product Analytics chart by holdout group, making it easier to measure the combined impact of multiple features.

What You Can Do Now

• Filter any chart to only include users in a specific holdout group

• Break down metrics by holdout status to compare behavior between held-out users and exposed users

How It Works

Holdouts are used to evaluate the aggregate effect of multiple features—not just individual experiments. A holdout group is a set of users who are intentionally excluded from a group of features or experiments to serve as a baseline. Now, you can use that same grouping to filter or break down any Product Analytics chart.

To apply, use the filter or group-by menu on a chart and select the relevant holdout.

Impact on Your Analysis

Holdout analysis helps you answer questions like:

• What’s the total impact of all features launched in the last quarter?

• Are users in the holdout group retaining or converting differently than exposed users?

It gives you a high-level view of product changes—beyond individual experiments—using the same familiar Product Analytics workflows.

🎯 Cohort Filters on Dashboards

You can now filter entire dashboards using behavioral cohorts—alongside existing property-based filters.

What You Can Do Now

• Apply a behavioral cohort as a filter to any dashboard

• Combine cohort filters with property filters for scoped, layered analysis

• Easily compare how different cohorts interact with your product across multiple charts

How It Works

From the dashboard filter panel, select a saved cohort to apply it globally. All charts on the dashboard will update to reflect data only for users in that cohort. You can still apply property filters in parallel.

This makes it simple to compare behaviors across user groups like “First-time Users,” “Power Users,” or “Users Who Churned Last Month.”

Impact on Your Analysis

Cohort filters unlock more targeted analysis across dashboards, allowing you to focus on the patterns and behaviors of specific user groups without editing each chart individually.

Statsig is proud to announce CURE, an advanced implementation of CUPED that allows customers to perform more complex regression adjustment using user-inputted categorical or numerical covariates.

This means users can run experiments even faster on Statsig, and extends the power of CUPED by making it available for new user experiments or metrics without any pre-exposure data available. This is compatible with simple adjustments to the regression - e.g. adding categorical covariates like region - or complex use cases like using predicted outcomes as a covariate.

Refer to our blog post for more information on CURE as well as the docs. This feature is currently available on Warehouse Native, and will be applied to experiments started after today. We'll be following up with CURE on Statsig Cloud.

📜 Helm Charts for Forward Proxy

Helm Charts are a simple way to deploy Kubernetes resources - like NPM, but for Kubernetes. Today we're releasing Helm Charts for the Statsig Forward Proxy - making it easy to add the proxy to your services. The Forward Proxy provides a centralized point of access for Statsig rulesets in your infrastructure resulting in lower cost, improved performance, and an extra layer of resilience. Helm charts make proxy deployment worlds easier - allowing you to deploy with simple commands like "helm install statsig-forward-proxy statsig/statsig-forward-proxy". You can also configure various options for Forward Proxy setup. Try out the Helm Chart today by visiting our Docs!

Gradual Experiment Rollouts

When you conclude an A/B test and decide to ship the winning variant, typically a 100% of the experiment's traffic shifts to the winning variant. Statsig now lets you ramp this gradually if you don't want a sudden, large shift like this.

Typical use case : When you're testing 5 variants; each gets 20% of traffic. When you make a ship decision, the winning variant goes from 20% of traffic to 100% of traffic. If you want to ramp traffic gradually to understand impact on CPU (or other resources), this feature lets you configure and schedule a multi-step, automated ramp.

Learn more here.

🔨 Rust Core SDK

Server Core is a full rewrite of our Server SDKs based on a shared, performance-focused Rust library. Today, we're launching the Rust library itself as a standalone, brand-new Rust SDK.

Performance & functionality exclusive to Server Core

So far, we've launched Node, Java, PHP, Python, and Elixir Core SDKs. Each of those depends on the Rust core, which today we're launching as its own SDK. Given its usage across numerous languages, we've invested in performance optimization of this Rust library, with substantial improvements vs. our original Rust SDK. In addition, Rust Core has new-to-Statsig features like Parameter Stores, Contextual Multi-Armed Bandits, and more.

Rust core is available on crates - read the docs to get started!

Aggregated Launch Impact

Quantifying your impact is critical. Statsig's Insights page provides you a clear view of how experiments impact a metric of interest. It allows you to focus on a single metric, identify which experiments are driving it most and estimate the aggregated impact. You can filter down to a team, or a tag. This is particularly useful to understand your team’s impact or set a reasonable goal for a future period.

You navigate to the Insights section on the Statsig console, or the insight tab for each metric, to check it out. Learn more here.

🚨 Topline Alerts

Want to monitor new sign-ups? Want to be alerted if feature performance is slow or a page fails to load? Topline Metric Alerts allow you to monitor a metric's topline value, independent of any Feature Gate or Experiment Rollout.

To configure a Topline Alert, head to Analytics -> Topline Alerts tab where you can find all your Topline Alerts and configure new ones. Choose the event you want to alert on and the aggregation you're most interested in, then craft a notification message and configure alert subscribers. To learn more, check out our docs here!

Note: this feature is in Beta and not broadly available (yet!) Please reach out to us if you'd like to be opted in for early access.

Topline Alert Configuration

Informed Bayesian for Smarter Experimentation

We're excited to start rolling out support for Informed Bayesian. This new feature allows you to integrate meaningful priors into Bayesian analyses, leading to faster insights and smarter decision making.

Reach out in Slack to opt your team in to this.

Native Contentful Integration

The Statsig Contentful app lets you create A/B/n tests and test different content blocks against each other - creating the experiment without leaving Contentful. Marketers can now optimize content, obtain insights, and iterate continuously right from within Contentful.

Learn more here.