Export Pulse Results to Your Warehouse in Warehouse Native

Access and interpret Statsig Warehouse Native experiment results in the console, including Pulse, scorecard, drill-down, and exported analysis views.

In Warehouse Native, Statsig writes two datasets to your warehouse: an exposures table with each unit's first exposure per experiment, and a statsig_daily_results SQL View with every Scorecard metric lift and its metadata. Statsig writes exposures to the forwarding table you configure in your project, and writes a new copy of the Scorecard results to the SQL View each time you load an experiment. Use the SQL View to query results across experiments in your warehouse. Use the console's Export Report only when you need a one-time Pulse Summary CSV file, as described in Report types.

Export Pulse results to your warehouse

Exposures

Statsig writes exposures to the table configured in your project setup. To find the table's location, go to Settings > Data Connection. The table is at {Database Name}.{Schema Name}.{Exposures Forwarding Table Name}, for example experimentation.statsig.exposures.

Results

The results SQL View contains the same results you can view in the console: experiment metadata (such as team, tags, target duration, and settings like CUPED and sequential testing), each metric's metadata (such as metric tags), and all metric lifts. To enable it, go to Project Settings > Data Connection > Export. Statsig then creates the SQL View in your warehouse and a matching metric source in your Statsig project.

Each time you load an experiment, Statsig exports a new copy of the Scorecard metric results. To tell result versions apart, use the ds column, which contains the timestamp when Statsig wrote the data.

Project settings data connection export interface

Schema of the results export table

The default view name is statsig_daily_results. The metric source that Statsig creates in your project has the same name.

Column	Type	Description
`ds`	timestamp	The time when Statsig wrote the data
`experimentName`	string	Name of the experiment
`experimentCreator`	string	Creator of the experiment
`experimentTeam`	string	Team conducting the experiment
`experimentTags`	array of strings	Tags associated with the experiment
`experimentStartTs`	number	Start timestamp of the experiment, in milliseconds
`experimentEndTs`	number	End timestamp of the experiment, in milliseconds
`targetExposures`	number	The target number of exposures for the experiment
`targetDuration`	number	The target duration of the experiment
`actualDuration`	number	The actual duration the experiment ran, from start date to decision date. For analyze-only experiments, the number of days from the configured start date to today
`controlGroupName`	string	Name of the control group in the experiment
`testGroupName`	string	Name of the test group in the experiment
`useCUPED`	boolean	Whether the experiment applied CUPED
`useSequential`	boolean	Whether the experiment applied sequential testing
`metricName`	string	Name of the metric the experiment measures
`metricType`	string	Type of metric
`metricTags`	array of strings	Tags associated with the metric
`higherIsBetter`	boolean	Whether a higher value of the metric is better
`isVerifiedMetric`	boolean	Whether the metric is verified
`metricTeam`	string	Team responsible for the metric
`absoluteDelta`	number	The absolute change in the metric value between control and test groups
`absoluteDeltaCI`	number	Confidence interval for the absolute delta
`relativeDelta`	number	The relative change in the metric value between control and test groups
`relativeDeltaCI`	number	Confidence interval for the relative delta
`absoluteDeltaPValue`	number	P-value associated with the absolute delta metric result
`toplineAbs`	number	The absolute topline metric value for the experiment
`toplineAbsCI`	number	Confidence interval for the absolute topline metric
`toplineRel`	number	The relative topline metric value for the experiment
`toplineRelCI`	number	Confidence interval for the relative topline metric
`projectedTopline`	number	Projected topline metric value based on current data
`projectedToplineCI`	number	Confidence interval for the projected topline metric
`projectedToplineRel`	number	Projected relative topline metric value based on current data
`projectedToplineRelCI`	number	Confidence interval for the projected relative topline metric
`controlUnits`	number	The number of control group units
`testUnits`	number	The number of test group units
`controlTotal`	number	Total value for the control group metric
`testTotal`	number	Total value for the test group metric
`controlMean`	number	The mean value for the control group
`testMean`	number	The mean value for the test group
`sequentialTestingAbsoluteDeltaCI`	number (optional)	Confidence interval for the absolute delta with sequential testing enabled
`sequentialTestingRelativeDeltaCI`	number (optional)	Confidence interval for the relative delta with sequential testing enabled
`sequentialTestingAbsoluteDeltaPValue`	number (optional)	P-value for the absolute delta with sequential testing enabled

Report types

The report types in this section are the outputs of the console's Export Report action, described in Exporting Pulse reports. There are three types:

Report	Contents	Files	Approximate size
Exposures	All exposed users and their first exposures. Use it to join on your own internal data, run custom queries in your own warehouse, or verify who was in the experiment, which group Statsig assigned them to, and when Statsig first exposed them.	`<experiment>_first_exposures.csv`	1-25 MB
Pulse Summary	Precomputed summary experimental data for all metrics and test groups, including everything visible on Pulse.	`<experiment>_pulse_summary.csv`	10-100 KB
Raw Data	Raw exposures and metrics data at the user-day level, for inspecting data manually or recomputing your own statistics.	`<experiment>_first_exposures.csv` (the same file as the Exposures report) and `<experiment>_user_metrics.csv` (each user's calculated metrics for each day of their enrollment)	10 MB-1 GB

In Warehouse Native, you can export only the Pulse Summary, because Statsig stores the other two types of data only in your warehouse. Retention limits export availability: Statsig holds exposures data for up to 90 days after you conclude an experiment, and holds raw user-level metrics data for 90 days.

Pulse Summary file description for feature gates

Column name	Description
`name`	Name of the experiment or feature gate
`rule`	Name of the feature gate rule
`metric_type`	Category of the metric. Statsig computes different metric types differently, including how it computes them in Pulse
`metric_name`	The name of the metric. For event metrics, this is the name of the event
`metric_dimension`	The subcategory of the metric. For example, if you log `value` in LogEvent, then `value` shows up as a subdimension. `dimension = !statsig_topline` indicates that this row reflects an aggregate across all dimensions
`start_date`	The start date for this measurement
`end_date`	The end date for this measurement
`test_units`	The number of users in the test group
`test_mean`	The average value of this metric across test users (or participating units when applicable)
`test_stderr`	The standard error for the estimate of the mean for test users. Use this to compute confidence intervals
`ctrl_units`	The number of users in the control group
`ctrl_mean`	The average value of this metric across control users (or participating units when applicable)
`ctrl_stderr`	The standard error for the estimate of the mean for control users. Use this to compute confidence intervals
`abs_delta`	The absolute difference between the test and control mean (`test_mean` - `ctrl_mean`)
`abs_stderr`	The estimated standard error of `abs_delta`
`rel_delta`	The relative difference between test and control mean, sometimes called lift ((`test_mean` - `ctrl_mean`)/`ctrl_mean`)
`rel_stderr`	The estimated standard error of `rel_delta` (`abs_delta`/`ctrl_mean`)
`z_score`	The calculated Z-score

Pulse Summary file description for experiments

Column name	Description
`name`	Name of the experiment or feature gate
`rule`	Name of the feature gate rule
`experiment_group`	The group of users for which Statsig computed this metric. For a feature gate, this is pass or fail. For an experiment, this is the variant name
`metric_type`	Category of the metric. Statsig computes different metric types differently, including how it computes them in Pulse
`metric_name`	The name of the metric. For event metrics, this is the name of the event
`metric_dimension`	The subcategory of the metric. For example, if you log `value` in LogEvent, then `value` shows up as a subdimension. `dimension = !statsig_topline` indicates that this row reflects an aggregate across all dimensions
`start_date`	The start date for this measurement
`end_date`	The end date for this measurement
`units`	The number of users included in this metric estimate
`mean`	The average value of this metric across units (or participating units when applicable)
`stderr`	The standard error for the estimate of the mean. Use this to compute confidence intervals

First exposures file description

Column name	Description
`unit_id`	The unit identifier used in the experiment (for example, `user_id`, `stable_id`, or `org_id`)
`name`	The name of the gate or experiment
`rule`	For gates, the rule name
`experiment_group`	The group Statsig assigned the user to
`first_exposure_utc`	The UTC timestamp when Statsig first assigned the user to the experiment
`first_exposure_pst_date`	The date in PST when Statsig first assigned the user to the experiment
`as_of_pst_date`	The date Statsig generated this data
`user_dimensions`	JSON-formatted key-value pairs describing the user's attributes at the time of first exposure

Unit metrics file description

Column name	Description
`pst_ds`	The 24-hour window the data refers to. Statsig anchors all dates from 12:00 AM to 11:59 PM PST
`unit_id`	The unit identifier used in the experiment (for example, `user_id`, `stable_id`, or `org_id`)
`metric_type`	The category of the metric
`metric_name`	The name of the metric
`metric_dimension`	The name of the metric dimension. `!statsig_topline` is the overall metric with no slicing
`metric_value`	The numeric value of the metric
`numerator`	For some metrics, Statsig tracks the numerator
`denominator`	For some metrics, Statsig tracks the denominator

Was this helpful?