Experiments

Selection bias: What randomization can and can’t prevent

A graphic of a bar chart with an arrow pointing upward.

Your randomizer can work correctly while the final experiment comparison becomes unreliable. People can enter different reporting paths, lose their identifiers, or disappear from outcome collection after assignment. The result still looks like an A/B test, but the rows left in the analysis may no longer support the question you meant to answer.

Random assignment is a powerful protection against selection into treatment. It does not guarantee representative recruitment, complete follow-up, or valid post-treatment filters. Understanding that boundary helps you distinguish an allocation problem from a problem observing the people you allocated.

The most consequential question is often what randomization lets you learn when some outcomes remain unknown. A larger sample or a more sophisticated statistical engine cannot supply those missing facts without additional evidence or assumptions.

Understand the protection randomization provides

Assignment should not depend on potential outcomes

Selection bias concerns systematic distortions introduced by how observations enter or remain in the comparison relevant to your question. The selection-bias definition emphasizes the relationship between the studied group and the population of interest. Terminology varies across fields, so name the actual selection process rather than assuming the label explains it.

With a common assignment probability, proper random assignment makes treatment allocation independent of the units’ potential outcomes. When assignment probabilities vary across prespecified strata, that independence is conditional on those strata. Product teams do not have to identify every motivation, prior preference, or unmeasured characteristic before assigning users. The allocation mechanism avoids deliberately routing more promising users into treatment and less promising users into control.

This protection concerns the units eligible for that randomization. It also relies on the implemented experiment matching the design: the assigned treatment must be well defined, identity must support the assignment, outcomes must be measured appropriately, and interference between units must be addressed where relevant.

Balance is an expectation, not a perfect copy

Randomization does not create two identical lists of people. A finite experiment can have chance differences in baseline behavior even when the mechanism is sound. The methodological basis of randomized comparisons is comparability on average under the design, rather than exact equality of every measured or unmeasured characteristic in every sample.

Investigate evidence of implementation failure, but do not repeatedly rerun allocation until the observed outcome looks attractive. Prespecified blocking or stratification can be legitimate design choices; changing allocation in response to favorable results is a different procedure.

Randomization also does not mean the observed difference equals the true effect exactly. Sampling or randomization uncertainty remains even with complete measurement. A confidence interval addresses uncertainty under its assumptions; it does not certify that recruitment, joins, filters, or outcome observation were valid.

There is another limit to what is observed. Each unit receives one assigned experience, so you do not observe that same unit’s outcome under the alternative at the same time. Randomization supports an aggregate comparison under the design; it does not reveal an individual causal effect for every user. Keeping this distinction clear prevents a clean experiment average from becoming an unsupported claim that every customer benefits.

Name the treatment effect you want to estimate. The effect of offering an experience and the effect among people who actually use it are different questions. Random allocation directly supports the first comparison under its assumptions; a user’s decision to adopt is not another random allocation.

Define the population and unit the design covers

Random assignment is different from random sampling

Assignment decides which experience an enrolled unit receives. Sampling concerns which units enter the study from a population. Randomizing a volunteer beta group can support a comparison inside that beta, while leaving a separate question about whether its result applies to customers who did not volunteer.

A narrow study is not automatically a failed study. An Android-only feature may appropriately be tested among eligible Android users. A launch claim about all platforms needs additional support. State the population, setting, and time period in the decision before debating whether the sample is representative enough.

The broader selection-versus-sampling comparison explains these categories and post-treatment subgroup examples. For an experiment review, the immediate task is to identify which population your assignment actually covers and whether the final analysis preserves it.

Choose identity and eligibility before treatment changes them

Eligibility should describe who can meaningfully receive the tested experience, using information that the treatment cannot already have changed. A supported device or an account state measured at entry can define a relevant population. Successfully using the new experience is usually an outcome, not an equivalent entry criterion.

Choose the randomization unit to match the intervention. Assigning individual employees to different collaboration settings may create interference when they work together. Assigning accounts can address some shared-experience requirements, but changes the number of independent units and the analysis needed. Randomization-unit design guidance makes identity stability and network effects part of the experiment design, not merely SDK configuration.

In implementation, preserve the assigned unit and variation through the outcome join. A device identifier that disappears after login can create missing outcomes rather than genuine non-conversions. A changing identity can also expose one person to multiple variations. Record the identity and timing assumptions with the experiment configuration, then verify the delivered data follows them.

Define the denominator first

Use the KPI playbook to distinguish binary, continuous, and ratio metrics, and connect each measurement to the experiment decision it is meant to support.

Read the KPI Playbook

Missing outcomes can defeat an apparently favorable result

Complete-case rates describe the people still observed

Suppose a hypothetical experiment assigns 100 units to each arm. The binary outcome window has ended for everyone, but a collection problem leaves some outcomes unknown. There are 18 known successes in each arm:

Outcome availabilityControlTreatment
Assigned units100100
Units with observed binary outcomes9060
Known successes among observed outcomes1818
Units with unknown outcomes1040
Success rate among observed outcomes20%30%

The complete-case comparison is 18/90 versus 18/60, an observed difference of +10 percentage points. It compares the units whose outcomes were available. It does not establish the success-rate difference for all 100 assigned units in each arm.

The larger missing fraction in treatment is a warning about observation. It is not proof of a particular direction or magnitude of bias. Equal missing fractions would not settle the issue either: equally many missing users could have different outcomes or different reasons for being unobserved.

Missingness does not inevitably create bias in every analysis. Purely random loss can mainly reduce precision; other missingness processes can select people with systematically different outcomes. The missing-value count alone does not identify which process occurred. Record the reasons for loss and what was known about affected units before treatment, rather than treating an arbitrary percentage threshold as a universal certificate of validity.

Intention-to-treat analysis with missing outcomes requires a defensible approach to those unknown outcomes. Keeping everyone’s original assignment preserves the target comparison, but the words “intention to treat” do not recover information that was never observed. Automatically assigning every missing value zero imposes an outcome assumption.

Ask what the unknown values could change

Because the example’s outcome is binary, we can calculate simple logical completion ranges. If none of control’s ten unknown outcomes were successes, its full-cohort sample rate would be 18%. If all ten were successes, it would be 28%.

Treatment has forty unknown outcomes, so its corresponding range is 18% to 58%. Across possible completions of these missing entries, the all-assigned sample rate difference could run from −10 percentage points to +40 points.

These are logical missing-outcome completion bounds, not a confidence interval and not exact bounds on a population causal effect. They hold the observed entries fixed and vary the unknown binary values. Even if every missing outcome were recovered, inference about an underlying effect would retain uncertainty from the experiment’s design.

The exercise reveals how little the complete-case +10-point result establishes without assumptions about observation. Filling all missing values with zero happens to produce 18% in both arms here. That zero difference is one completion, not a demonstrated absence of an effect.

Attrition analysis can use sensitivity checks to make missing-data assumptions visible. A useful next step is to recover outcomes from reliable records, investigate why collection failed, and assess conclusions under plausible alternatives. The most favorable completion is not a substitute for that investigation.

Preserve the comparison after assignment

Different exclusions require different repairs

Distinguish incomplete observation from a known zero. An account that was fully observed and did not purchase can legitimately contribute zero to a purchase metric. An account that no longer joins to the purchase system has an unknown outcome unless other evidence establishes what happened.

Telemetry-loss research shows why lost records can affect power and bias comparisons. Increasing traffic through the same broken collection path need not fix the problem. It may narrow the reported interval around a distorted complete-case comparison.

Inspect the metric and Fact Table definitions, including identifiers, join types, filters, timestamp alignment, and outcome availability. A left join followed by a filter on an outcome-table field can silently become an exclusion rule. A metric name does not reveal which people survived that rule.

Post-treatment filtering is another mechanism. Restricting analysis to purchasers or successful feature users can select different people in each arm when treatment changes those behaviors. Identical SQL applied to both arms does not make selected membership independent of treatment. Use such a view for a clearly limited descriptive purpose or an appropriate causal analysis; do not assume it inherits the original randomized comparison.

Triggering needs a causal and implementation argument

Analyzing a relevant triggered population can be defensible when the trigger identifies comparable eligibility across assignments. The criterion is not simply whether logging happens before or after the assignment timestamp. Ask whether the tested treatment can affect who satisfies the condition or whether that condition is recorded.

GrowthBook’s activation-metric settings explicitly warn that activation should be independent of assignment. Where a feature exists only in treatment, a carefully designed counterfactual logging approach can record who would have encountered it in control. That requires equivalent logic and validation, not just a treatment-only click event with a new name.

Passing a sample ratio mismatch check is useful evidence about allocation counts, not a guarantee that every inclusion rule is valid. Selection can change the composition of two equally sized groups. Review reasons for missingness and baseline characteristics alongside counts, rather than forcing arm sizes to match by discarding extra observations.

No statistical repair is universal. Weighting and imputation require assumptions and relevant observed information. A group absent from the data cannot be studied merely by assigning it a larger weight. A missingness mechanism related to unobserved outcomes cannot be declared harmless because a model returned a precise estimate.

For example, an imputation model may assume observed baseline information is sufficient to account for who has missing outcomes. That assumption needs a substantive argument; the unobserved values themselves cannot be inspected to verify it. Compare plausible departures from the assumption and report whether the practical decision survives. Do not choose the missing-data model because it produces the preferred launch recommendation.

Report the evidence randomization actually preserved

Track assignment separately from outcome observation

Maintain counts for eligible units, assigned units, recorded exposures, observed outcomes, and the final analysis population. Explain losses between stages by arm and by relevant baseline groups. The ledger should distinguish a known outcome from unavailable measurement, and an original exclusion from a later analytic restriction.

The data-quality requirements for online experiments include completeness, consistent assignment records, reliable joins, and timing. An allocation check alone covers only part of that path. A clean A/A test can reveal existing problems but cannot prove a future treatment will leave exposure and outcome collection unchanged.

Preserve the query or metric revision behind the result. If the team repairs a join or recovers delayed records, state whether the correction reconstructs the intended original population. If it does not, explain the resulting scope and whether a new run is needed. Avoid relabeling a selected survivor cohort as the original experiment.

Design outcome observation before increasing exposure. For an important transaction metric, a reliable backend record may allow recovery when a client event fails, provided identity and timing still join correctly. Verify that the recovery method is available in both arms and measures the intended outcome. A second event stream is useful evidence only when its own coverage and meaning are understood.

Make missing-data assumptions part of the launch decision

Alongside the experiment results, report outcome availability and the assumptions used for missing values. Transparent randomized-trial reporting treats participant flow, exclusions, and missing-data handling as information needed to judge a result. The same statistical discipline is useful for product experimentation when the participants are accounts rather than clinical-trial patients.

For the hypothetical experiment, the immediate conclusion is that the favorable complete-case result does not identify the all-assigned sample comparison without more information. The next action is to investigate observation, recover outcomes where possible, and evaluate the remaining uncertainty. It is not to declare a winner from 20% versus 30%, or a tie from zero-filled missing outcomes.

Randomization gives you a defensible starting contrast. Preserving it requires that the data and analysis continue to describe the units and outcomes the decision concerns. Before launching from a randomized result, be able to explain not only how people received a variation, but also whose outcomes are known, whose are missing, and what your conclusion assumes about the difference.

Inspect the metric definition

Open the read-only metric search workflow to find a fact metric and inspect its aggregation, filters, identifiers, and associated fact table. Running the workflow requires configured GrowthBook API access; the instructions do not execute an audit when opened.

View the Metric Search Workflow

Table of Contents

Related Articles

See All Articles
Experiments

T-statistic explained: What the t-value tells you about your results

Nov 13, 2026
x
min read

An experiment report shows a t-value of 2.4. Is the product change large? Is it beneficial? Does it have a 95% chance of working? That single number cannot answer those questions.

A t-statistic compares an estimated quantity with a hypothesized value, using the estimate’s standard error as the measuring stick. It compresses the size of the observed departure and the precision of the estimate into one signed number. Understanding the two parts is more useful than memorizing a cutoff.

For experiment readers, the essential habit is to unpack the t-value before interpreting it. Establish what was subtracted, which uncertainty estimate sits underneath it, and what reference distribution applies. Then return to the effect in product units and the decision it needs to support.

Read t as a difference divided by uncertainty

Identify the numerator and its direction

The general structure is:

t = (estimated quantity - hypothesized value) / estimated standard error

In an independent treatment–control comparison, the estimated quantity might be the treatment mean minus the control mean. With a null difference of zero, a positive t means the estimated treatment mean is higher. Reverse the subtraction order and the sign reverses.

“Higher” does not necessarily mean “better.” More purchases may be desirable; more errors usually are not. A negative latency difference can support an improvement, while a positive error-rate difference can indicate harm. Read the direction in the metric’s own terms before assigning a business meaning to the sign.

The hypothesized value also matters. The usual zero-difference comparison asks whether the population mean difference is zero. Testing another specified difference changes the numerator. Neither calculation independently establishes an equivalence or non-inferiority claim; those require the appropriate hypotheses, margins and decision procedure.

Distinguish standard error from standard deviation

An observation standard deviation describes spread among observed unit outcomes. A standard error describes uncertainty in an estimator, such as a sample mean or a difference between means. They can share units while answering different questions.

For an ordinary mean of independent, identically distributed observations with finite variance, the estimated standard error is the sample standard deviation divided by the square root of the sample size. The standard-error calculation therefore changes when the number of independent observations changes, even if their measured spread stays the same.

This is why dividing a mean difference by the observation standard deviation does not give the usual two-sample t-statistic. It omits how precisely the means were estimated. The distinction between standard error and standard deviation is the first denominator check to make when a t-value looks surprising.

Unit-free does not mean effect size

Both numerator and standard error have the same measurement units, which cancel in their ratio. Converting a revenue analysis from dollars to cents multiplies both by 100 and leaves t unchanged. That is a useful consistency check for a positive linear change of units.

However, t does not describe the size of the product effect independently of precision. An observed difference of one dollar can have a small or large t-value depending on the standard error. Comparing t-values across two experiments is therefore not a reliable way to rank their business impact.

The statistic measures departure from the null in estimated standard-error units. Keep the raw difference beside it so the unit-free summary does not displace the quantity the team actually needs.

The calculation also needs a usable, positive estimated standard error. A zero or undefined denominator is a reason to inspect the data and model, rather than manufacture a finite t-value or call the result certain. Software may reject such inputs; for example, the R t-test reference documents an error for effectively constant input data.

Match the denominator to the study design

Different designs estimate different uncertainty

The formula’s short form hides a substantial choice: how was the standard error calculated? Three common cases illustrate why the design must come first.

DesignEstimated standard errorReference degrees of freedom
Ordinary one-sample meanSample SD divided by √nn − 1 under the classical model
Ordinary paired mean differenceSD of within-pair differences divided by √number of pairsNumber of independent complete pairs − 1 under the classical model
Independent Welch comparison√(s² treatment / n treatment + s² control / n control)Welch–Satterthwaite approximation using both group sizes and variances

In the paired case, the relevant spread is the spread of differences within pairs. Substituting the two arms’ separate standard deviations discards the relationship the design uses. The paired t calculation treats the differences as its observations; the independent information comes from the pairs.

For two independent groups, Welch’s calculation does not require equal population variances. The pooled Student calculation adds that assumption and uses a different variance estimate. The Welch-versus-Student comparison helps distinguish those cases. A label reading only “t-test” can conceal the choice.

Separate a reference model from an arithmetic ratio

You can calculate a difference divided by an estimated standard error without proving that the resulting statistic has the reference distribution you intend to use. Exact classical one-sample and paired t inference relies on the corresponding independent normal-observation model; in the paired case, this concerns the differences.

Welch’s reference distribution is an approximation with estimated degrees of freedom. It is not generally exact even for unequal-variance normal populations. Non-normal outcomes can support useful large-sample mean inference under appropriate conditions, but there is no sample-size threshold that makes every heavy tail or dependence problem disappear.

If accounts are randomized, individual events from one account are not automatically independent units. A standard error that ignores the account structure may understate uncertainty. More rows in an exported file do not by themselves justify a smaller denominator or larger degrees of freedom.

Plan for more precise estimates

Explore variance-reduction methods and their tradeoffs before your next test. Match each method to the outcome you intend to estimate.

Explore Variance Reduction

Hold the observed effect fixed and change precision

Work through an independent mean comparison

Consider a hypothetical fixed-horizon experiment with one completed revenue outcome per independent user. Control has mean revenue of $10; treatment has $11. Each arm has sample standard deviation $4 and 100 users. The target is the arithmetic mean difference, treatment minus control, and the null difference is zero.

The estimated effect is $1 per user. Using the independent unequal-variance standard error, square each group’s standard deviation, divide by its sample size, add those terms and take the square root. Here that is √(16/100 + 16/100), or about $0.5657.

Divide the $1 difference by $0.5657 to get t ≈ 1.7678. The estimated difference lies about 1.77 estimated standard errors above the zero-difference null. The sample standard deviation is still $4; the $0.5657 figure measures uncertainty in the difference between the two estimated means.

For these equal sample sizes and equal observed variances, the Welch–Satterthwaite calculation gives 198 degrees of freedom. Under the conventional Welch approximation, the two-sided p-value is approximately 0.07864 and the 95% confidence interval for the mean difference is approximately −$0.1155 to $2.1155.

Compare two hypothetical studies

Now consider a separate hypothetical study with the same group means and standard deviations, but 400 independent users per arm. The observed $1 difference has not grown. Its estimated standard error is smaller.

Summary100 users per arm400 users per arm
Observed mean difference$1$1
Sample SD in each arm$4$4
Estimated SE of the difference$0.5657$0.2828
t-statistic1.76783.5355
Welch reference df198798
Conventional two-sided p-value0.078640.0004304
Conventional 95% CI for the difference[−$0.1155, $2.1155][$0.4448, $1.5552]

Quadrupling each group’s sample size halves the estimated standard error when these observed standard deviations stay fixed. It doubles t for the same observed effect. The larger statistic reflects greater estimated precision here, not a larger observed revenue gain.

These are teaching summaries for separate studies. They are not an instruction to keep extending one experiment until significance appears. In a real extension, means and variances can change, and the stopping rule affects how the inference should be calibrated. An observed t-value also does not directly tell you the power of a future study.

Translate t using the appropriate reference distribution

Degrees of freedom and tails determine the threshold

A t-value needs a reference distribution before you can obtain a p-value or critical value. The Student t family includes a degrees-of-freedom parameter; smaller values generally give heavier tails. The same absolute statistic can therefore produce different tail probabilities at different df.

For a prespecified two-sided 5% test, the relevant critical value is the 97.5th percentile of the reference t distribution. It approaches about 1.96 as df becomes large, but 1.96 is not a universal t cutoff. Nor does every test use n − 1. Degrees of freedom depend on the calculation and may be fractional for Welch.

A one-sided alternative uses a specified tail and a different critical value. A negative statistic is not evidence in favor of an upper-tail alternative simply because its absolute value is large. Choose the direction before seeing outcomes; taking whichever tail looks favorable changes the decision rule.

A p-value conditions on the null model

For a conventional two-sided t-test, the p-value is the probability, under the null model and assumptions, of a statistic at least as extreme in either direction as the observed t. It is not the probability that the null is true, that chance caused the result, or that a future experiment will succeed.

The ASA’s p-value principles also distinguish statistical significance from effect magnitude and practical importance. A threshold crossing does not determine whether the expected benefit pays for the change. If you need the broader conditional-probability interpretation, the t-test p-value guide develops it separately.

Read the interval in the original units

The corresponding conventional interval is the estimated difference plus or minus the relevant critical t-value times its estimated standard error. For the same two-sided procedure at 5%, a 95% interval excluding zero agrees with rejecting a zero-difference null. That agreement requires compatible test, interval and adjustment settings.

Confidence describes the repeated-use coverage of the method under its assumptions. It does not assign a 95% posterior probability to the fixed true effect lying inside the particular computed interval. The confidence-interval construction makes the relationship between standard error, critical value and width explicit.

In the first study, the interval includes a small loss and a substantial gain. In the second, it excludes zero but still spans different levels of business value. A minimum worthwhile effect helps interpret that range. Neither a large p-value nor an interval containing zero proves equality; equivalence requires an analysis designed for that claim.

Audit the inputs before acting on a t-value

Trace an unexpected statistic to its components

Start with the estimated effect and null value. Confirm the subtraction order, currency or time units, eligible population and observation window. A swapped group order explains a sign change; it does not explain a change in a correctly calculated two-sided p-value.

Then inspect the denominator. Check group sizes and sample variances at the actual independent unit. Look for duplicated joins, repeated events counted as independent observations, incomplete follow-up and treatment-dependent missingness. The arithmetic can be internally consistent while the data fail to represent the intended experiment.

Inspect the analysis settings too. A ratio metric, regression adjustment or variance-reduction method can have a standard error different from the simple independent-mean formula above. Do not attempt to reproduce a dashboard by substituting top-line traffic counts into the wrong calculation. GrowthBook’s statistical effect calculations describe its estimators and frequentist reference methods, including distinct metric types.

Keep the decision rule attached to the evidence

The worked examples use conventional fixed-horizon inference. Repeatedly inspecting that calculation and stopping at a favorable threshold changes its false-positive behavior. A planned sequential analysis is a different inferential procedure; GrowthBook documents an asymptotic confidence-sequence approach with its own settings. It does not repair incorrect units or missing outcomes.

Likewise, selecting the largest t among many metrics, variants or segments creates a testing problem beyond a single comparison. Preserve the planned comparisons and applicable corrections rather than treating the most impressive number as an isolated result.

A useful readout names the estimated effect, interval, test, df, direction and sample sizes, followed by the measurement checks and decision threshold. The team can then assess what the evidence supports and whether that effect is worth shipping. Use the t-value to understand precision relative to the null; keep the launch decision grounded in effects on users and the business.

Review the result in context

Open the official experiment-analysis workflow to inspect saved results, health checks and analysis settings with configured API access. It can refresh a result snapshot when needed; stopping or rollout is separate.

View the Analysis Workflow
Experiments

How to run A/B tests in Strapi

Nov 13, 2026
x
min read

A published Strapi document is content. An A/B test also needs an assignment, a delivered experience, and an outcome you can attribute.

The integration becomes unreliable when those pieces disagree. An older example may read data.attributes, while your Strapi 5 endpoint returns flat fields. A numeric database ID may be mistaken for a document ID. An internal service call may return a draft even though a similar REST request defaults to published content.

This guide builds an explicit Strapi 5 content test using a small server-rendered Node application and GrowthBook. It validates the complete published content pair before enrolling a visitor, preserves a stable browser identifier, and records an exposure only after browser execution reaches the selected page.

The code also fingerprints the reviewed strings so a later edit can be rejected before assignment. The local event collector demonstrates the measurement boundary; a real experiment still needs your existing signup flow, durable analytics delivery and verified warehouse joins.

Define the Strapi 5 delivery contract

Pick a content model you can inspect

The existing GrowthBook Strapi integration guide demonstrates Dynamic Zone components selected through a frontend SDK. The reference here uses two ordinary string fields and server-side evaluation. It does not require a separately installed Strapi experimentation plugin or depend on an unverified plugin compatibility claim.

Start with a hypothesis about a meaningful outcome: an outcome-focused headline will increase completed trial signups per eligible offer-page visitor. Keep the offer and signup destination consistent if headline copy is the intervention. If the change includes a larger component package, describe that package as the treatment.

A plain-text pair makes the first integration easier to inspect. You can verify the exact response, render the selected value safely, and compare the approved strings. When extending the model to components or Dynamic Zones, validate their full structure and use a reviewed variation mapping. Reordering an array must not silently redefine which experience belongs to control.

For example, an editorial interface might place an alternative card first because it is easier to find, while the frontend still interprets index zero as control. The content can remain valid JSON and render perfectly while the experiment labels have effectively switched. Use explicit variation keys or a constrained, reviewed ordering contract when extending the simple pair. Check the rendered package against the experiment mapping, not just that every component has a title.

Use the document API that your code actually calls

Strapi 5 REST responses expose normal content fields directly under data. The document endpoint uses documentId; the numeric id in a response is not the identifier this guide puts in the URL. Some plugin endpoints have different response conventions, so this contract applies to an ordinary generated collection endpoint.

The production request explicitly sets REST status to published. This is also the REST default when Draft & Publish is enabled. The internal Document Service API defaults differently, to draft. Do not transfer that internal default—or an older v4 parameter—into this REST implementation.

Keep locale explicit as well. The requested document and locale together identify the content version being delivered. This guide checks the returned locale instead of borrowing another CMS's fallback behavior or assuming a valid document ID means every translation exists.

The document concept groups localized and draft/published variations around a document identifier. That does not make the document identifier an immutable snapshot of the text. Keep source identity, publication state and approved content as separate fields in the experiment record. Changing one is a deployment decision even when the others remain familiar.

Define assignment and measurement before launch

Use one stable assignment unit. The example creates an anonymous first-party browser ID; account-level interactions may require account assignment instead. Plan the anonymous-to-authenticated transition so signup measurement does not lose the identifier or enroll one person under several unrelated IDs.

Choose a primary business outcome, a small diagnostic funnel and relevant guardrails. A headline can increase button clicks without increasing completed or qualified signups. Estimate eligible traffic and the effect that would change your decision before starting a test that cannot resolve the question in a useful period.

Separate content ownership from experiment operation. An editor approves the published pair, an engineer owns delivery and fallback behavior, and the analytics owner verifies exposure and outcome data. Name who can change the tested content and how a change starts a new reviewed comparison.

Model, publish and inspect the content pair

Create a localized collection in development

Use an existing Strapi 5 test application with Draft & Publish enabled for the collection. Configure internationalization and the intended locale if the content is localized. The Content-type Builder belongs to the development workflow; deploy schema changes through your normal application process.

Create a collection named Experiment page with singular API ID experiment-page and plural API ID experiment-pages. A minimal schema at src/api/experiment-page/content-types/experiment-page/schema.json is:

{
  "kind": "collectionType",
  "collectionName": "experiment_pages",
  "info": {
    "singularName": "experiment-page",
    "pluralName": "experiment-pages",
    "displayName": "Experiment page"
  },
  "options": {"draftAndPublish": true},
  "pluginOptions": {"i18n": {"localized": true}},
  "attributes": {
    "controlHeadline": {
      "type": "string",
      "required": true,
      "pluginOptions": {"i18n": {"localized": true}}
    },
    "variantHeadline": {
      "type": "string",
      "required": true,
      "pluginOptions": {"i18n": {"localized": true}}
    }
  }
}

For an existing model, merge the relevant changes rather than replacing its other fields. Restart or rebuild according to your development workflow, create a document, enter both approved headlines and publish the intended locale. Copy its actual documentId from the API response or your document inspection workflow.

Review model changes and content changes through their appropriate paths. Adding a field affects application schema and deployment; editing a headline changes a content entry. Give the editor clear field descriptions so control means the approved baseline and variant means the approved experimental alternative. A required field helps authoring, but the renderer still validates the actual response from the deployed endpoint.

With internationalized content, inspect the specific localized pair that will be eligible for the experiment. Translating control but leaving treatment in an unintended language changes more than the hypothesis describes. Keep locale targeting and content approval aligned, and decide in advance whether results across languages belong to one planned analysis or separate comparisons.

Because both alternatives are plain strings, this request needs no relation population. Components, relations and media require an explicit population strategy when added later; populate=* is not a substitute for defining the treatment's dependencies and permissions.

Use a read token on the rendering server

Create a scoped Content API token for the renderer. A read-only token supports find and findOne; a Custom token can narrow the relevant endpoint access. Keep it on the application server. You do not need to expose an admin token or broadly enable public-role access for this authenticated server request.

Read-only describes allowed operations, not a universal guarantee that a caller can never request drafts. Fix production status and locale in trusted code. Keep editor preview routes and any draft-capable configuration separate from live enrollment and analytics.

When extending the endpoint with a language switcher, select from an approved locale set rather than forwarding every browser query parameter to Strapi. A request parameter named status should not turn the production renderer into a draft-content proxy. Keep the collection path and publication choice fixed, and validate the returned document and locale even after constructing a valid request.

Check permissions on the actual route you use. A working admin interface does not prove the rendering token can read the generated Content API endpoint. Conversely, an access problem is not a reason to grant broad public access to the entire model. Confirm the token's scope and the endpoint response independently of experiment assignment.

Create a reference Node project, or integrate the files into your existing application. Use a modern Node runtime compatible with the locked dependencies, such as Node 22 or later:

npm init -y
npm install express@5 cookie-parser @growthbook/growthbook@2.0.0
mkdir public

Add server configuration. STRAPI_ORIGIN is the CMS origin, including a port when needed, without /admin or /api appended:

# .env.local
STRAPI_ORIGIN=http://localhost:1337
STRAPI_DOCUMENT_ID=your-document-id
STRAPI_LOCALE=en
STRAPI_READ_TOKEN=your-read-token
STRAPI_EXPECTED_SHA256=
GROWTHBOOK_CLIENT_KEY=sdk-your-test-connection-key
PORT=3000

Use the appropriate HTTPS origin outside your local development setup. Store credentials through your deployment's configuration mechanism, not in public HTML or query parameters.

Validate the published response and reviewed fingerprint

The following reader fetches one document by documentId, with explicit locale selection. It rejects a missing, draft, wrong-locale or malformed pair before the SDK can assign anyone:

// strapi-content.mjs
import { createHash } from "node:crypto";

export async function getPublishedPair() {
  for (const name of ["STRAPI_ORIGIN", "STRAPI_DOCUMENT_ID", "STRAPI_LOCALE", "STRAPI_READ_TOKEN"]) {
    if (!process.env[name]) throw new Error(`Missing ${name}`);
  }
  const origin = new URL(process.env.STRAPI_ORIGIN);
  if (!["http:", "https:"].includes(origin.protocol) || origin.pathname !== "/" ||
      origin.search || origin.hash || origin.username || origin.password) throw new Error("Invalid CMS origin");
  const documentId = process.env.STRAPI_DOCUMENT_ID;
  const locale = process.env.STRAPI_LOCALE;
  const url = new URL(`/api/experiment-pages/${encodeURIComponent(documentId)}`, origin);
  url.searchParams.set("status", "published");
  url.searchParams.set("locale", locale);
  const response = await fetch(url, { headers: { Authorization: `Bearer ${process.env.STRAPI_READ_TOKEN}` },
    signal: AbortSignal.timeout(3000) });
  if (!response.ok) throw new Error(`Strapi delivery status ${response.status}`);
  const payload = await response.json();
  const data = payload?.data;
  if (!data || Array.isArray(data) || data.documentId !== documentId || data.locale !== locale ||
      typeof data.publishedAt !== "string" || !Number.isFinite(Date.parse(data.publishedAt)) ||
      typeof data.controlHeadline !== "string" || !data.controlHeadline.trim() ||
      typeof data.variantHeadline !== "string" || !data.variantHeadline.trim()) {
    throw new Error("Invalid published Strapi content pair");
  }
  const canonical = JSON.stringify([origin.origin, documentId, locale,
    data.controlHeadline, data.variantHeadline]);
  const sha256 = createHash("sha256").update(canonical).digest("hex");
  const expected = process.env.STRAPI_EXPECTED_SHA256;
  if (expected && (!/^[0-9a-f]{64}$/i.test(expected) || expected.toLowerCase() !== sha256)) {
    throw new Error("Strapi content differs from the reviewed pair");
  }
  return { copy: { control: data.controlHeadline, variant: data.variantHeadline },
    source: { instance: origin.origin, documentId, locale, publishedAt: data.publishedAt, sha256 } };
}

Leave the optional expected hash blank for the first read-only inspection, with Strapi running and the document published. Run:

node --env-file=.env.local --input-type=module -e "import {getPublishedPair} from './strapi-content.mjs'; console.log(await getPublishedPair());"

Confirm the actual approved pair, source origin, document and locale. Set STRAPI_EXPECTED_SHA256 to the returned digest before enrollment. The SHA-256 calculation fingerprints those exact strings and source dimensions. It does not retrieve an older Strapi version or grant access to anything.

documentId, updatedAt and publishedAt are useful context, but they do not freeze a treatment. The fingerprint rejects changes to the reviewed pair before assignment. Identical strings can retain the same digest after republication; publication metadata is recorded separately. Extend the fingerprint when additional content becomes part of the treatment.

Treat updating the expected digest as a review step. First inspect the new delivered pair, confirm why it changed, and decide whether the current experiment can continue under a documented phase or needs a new experiment. Copying the latest digest merely to clear an unavailable response defeats the check. Preserve the approved strings alongside the digest so a reviewer can understand the change without reconstructing it from a hash.

The digest covers the two headline strings and source dimensions in this implementation. It does not cover an unrelated image, price, signup destination or application deployment. If those become part of the intervention, extend the reviewed package and its validation. A useful content contract is explicit about both its coverage and the changes that require another review.

Review the experiment before launch

Open the experiment-design workflow to turn your content hypothesis, assignment unit, and metrics into a reviewable specification. Configure API access as directed and review the plan before launch.

Open Experiment Design Instructions

Render the assigned page with request-local context

Link the content values to an experiment

Create a string feature strapi-offer-copy, defaulting to control. Link an experiment with tracking key strapi-offer-copy-v1, mapping the two string values to the intended variations. Set hash attribute id, the correct SDK environment, targeting and goals in the experiment configuration.

The content value chooses a string from the pair. The experiment and variation identifiers describe the assignment for analysis. Preserve experiment.key and result.key; do not substitute the headline or assume a variation key must equal the content value.

The GrowthBook Node SDK can share feature payload data while keeping each visitor's context in a scoped instance. The server initializes through the request path and checks success, using the SDK's caching and refresh policy rather than retaining one startup promise forever. Test stopping and refreshing on warm instances; a shared cache is not an instant global update guarantee.

Keep the browser identity stable and the output safe

Create the server below. The existing /signup application flow is a prerequisite: implement or replace that link before measuring completed signups. The collector is a local diagnostic endpoint until connected to durable analytics.

// server.mjs
import express from "express";
import cookieParser from "cookie-parser";
import { randomUUID } from "node:crypto";
import { fileURLToPath } from "node:url";
import { GrowthBookClient } from "@growthbook/growthbook";
import { getPublishedPair } from "./strapi-content.mjs";

const app = express();
app.use(cookieParser());
app.use(express.static(fileURLToPath(new URL("./public/", import.meta.url))));
const experimentKey = "strapi-offer-copy-v1";
const gb = new GrowthBookClient({ apiHost: "https://cdn.growthbook.io",
  clientKey: process.env.GROWTHBOOK_CLIENT_KEY });
const uuid = /^[0-9a-f]{8}-[0-9a-f]{4}-4[0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$/i;
const identity = (req) => typeof req.cookies.gb_strapi_id === "string" && uuid.test(req.cookies.gb_strapi_id)
  ? req.cookies.gb_strapi_id : null;
const html = (value) => value.replace(/[&<>"']/g,
  (char) => ({ "&": "&amp;", "<": "&lt;", ">": "&gt;", '"': "&quot;", "'": "&#39;" })[char]);
const jsonData = (value) => JSON.stringify(value).replace(/</g, "\\u003c");

app.get("/experiment-session", (req, res) => {
  res.set("Cache-Control", "private, no-store");
  if (!identity(req)) res.cookie("gb_strapi_id", randomUUID(), { httpOnly: true,
    secure: process.env.NODE_ENV === "production", sameSite: "lax", path: "/",
    maxAge: 180 * 24 * 60 * 60 * 1000 });
  res.redirect(303, "/offer?session=1");
});

app.get("/offer", async (req, res) => {
  res.set("Cache-Control", "private, no-store");
  const id = identity(req);
  if (!id && req.query.session !== "1") return res.redirect(303, "/experiment-session");
  try {
    const content = await getPublishedPair();
    const assignments = [];
    let selected = "control";
    if (id) {
      const initialized = await gb.init({ timeout: 2000 });
      if (initialized.success) {
        const scoped = gb.createScopedInstance({ attributes: { id, locale: content.source.locale },
          trackingCallback: (experiment, result) => {
            if (experiment.key !== experimentKey) return;
            if (result.value !== "control" && result.value !== "variant") {
              console.error("Invalid Strapi assignment", { experiment_id: experiment.key,
                variation_id: result.key, user_id: id, value: result.value });
              return;
            }
            assignments.push({ experiment_id: experiment.key, variation_id: result.key, value: result.value });
          } });
        const result = scoped.evalFeature("strapi-offer-copy");
        if (result.value === "control" || result.value === "variant") selected = result.value;
        else console.error("Invalid Strapi feature value", result.value);
      } else console.error("Feature payload unavailable", initialized.source);
    }
    const context = { assignments: assignments.filter((item) => item.value === selected),
      content_variant: selected, source: content.source };
    res.type("html").send(`<!doctype html><html lang="${html(content.source.locale)}">
      <head><meta charset="utf-8"><title>Trial offer</title></head><body>
      <main data-content-variant="${selected}"><h1>${html(content.copy[selected])}</h1>
      <a href="/signup">Start a trial</a></main>
      <script type="application/json" id="strapi-exposure">${jsonData(context)}</script>
      <script src="/experiment-view.js" defer></script></body></html>`);
  } catch (error) {
    console.error("Strapi offer unavailable", error.message);
    res.status(503).type("html").send("<h1>This offer is temporarily unavailable</h1>");
  }
});

app.post("/experiment-view", express.json({ limit: "4kb" }), (req, res) => {
  const id = identity(req);
  const event = req.body;
  const source = event?.source;
  const fields = ["experiment_id", "variation_id", "event_id", "content_variant"];
  const sourceFields = ["instance", "documentId", "locale", "publishedAt", "sha256"];
  if (!id || !event || typeof event !== "object" ||
      fields.some((key) => typeof event[key] !== "string" || !event[key].trim() || event[key].length > 200) ||
      event.experiment_id !== experimentKey || !["control", "variant"].includes(event.content_variant) ||
      !source || typeof source !== "object" || sourceFields.some((key) =>
        typeof source[key] !== "string" || !source[key].trim() || source[key].length > 300) ||
      !/^[0-9a-f]{64}$/i.test(source.sha256) || !Number.isFinite(Date.parse(source.publishedAt))) {
    return res.sendStatus(400);
  }
  console.log(JSON.stringify({ event: "experiment_view", user_id: id,
    received_at: new Date().toISOString(), ...Object.fromEntries(fields.map((key) => [key, event[key]])),
    source: Object.fromEntries(sourceFields.map((key) => [key, source[key]])) }));
  res.sendStatus(204);
});

app.listen(Number(process.env.PORT || 3000));

The cookie parser can produce non-string values, so identity is validated before assignment. Express cookie maxAge uses milliseconds. The redirect establishes the cookie before the offer request; its one-attempt marker avoids looping when persistence fails. Adapt identity and consent behavior to the application's existing policy.

The headline is escaped as HTML text. The JSON data element separately escapes less-than signs so a closing-script sequence cannot terminate it. These are different output contexts. Do not use this plain-string renderer as a general HTML or Rich Text renderer.

Keep failure behavior visible

A missing document, wrong locale, draft payload or changed fingerprint fails before enrollment. A failed feature-payload initialization can retain valid control content outside enrollment. Unexpected raw assignment values are logged and do not become misleading control-view events.

Fetching both alternatives in one validated payload makes this eligibility check independent of the assigned arm. A different architecture that first assigns treatment and then fetches only its content has another failure problem: a missing treatment resource can selectively remove treatment visitors from observed exposures. If larger packages require that architecture, preserve assignment and delivery failures explicitly and design the analysis population before launch. Do not infer that a missing view means the visitor never belonged to the comparison.

Preserve diagnostics and investigate the affected interval. Filtering failed treatment deliveries from analysis can select the comparison population after assignment. A safe display fallback does not make a broken experiment trustworthy; repair or halt it when delivery or configuration is unreliable.

The personalized response uses private, no-store. Shared caching of the complete validated pair is a separate choice from caching a visitor's selected HTML. Verify the actual reverse proxy and CDN behavior rather than assuming a response header alone describes every cache in the deployment.

When adding a shared content cache, include the CMS instance, document, locale, publication selection and reviewed fingerprint in the appropriate cache design. A cached English pair should not satisfy a request for a different locale, and a previously approved pair should not conceal a broken publication process indefinitely. Test invalidation and recovery using the same deployment configuration as the experiment.

The server selects one headline before producing HTML, so the browser does not need to rewrite that headline to implement this treatment. Measure the server response path under realistic load nonetheless: content fetching, payload initialization and application work still consume time. Keep loading and error guardrails appropriate to the page rather than promising that server-side evaluation has no performance cost.

Connect browser exposure to the outcome

Record a defined display boundary

The callback during evaluation queues assignments; it is not proof that HTML reached a browser. Use the deferred-tracking pattern to carry compact metadata into a client script that runs after parsing the page:

// public/experiment-view.js
(() => {
  const node = document.getElementById("strapi-exposure");
  if (!node || node.dataset.sent === "1") return;
  let context;
  try { context = JSON.parse(node.textContent); }
  catch { console.error("Invalid Strapi exposure data"); return; }
  if (!Array.isArray(context?.assignments)) return;
  node.dataset.sent = "1";
  for (const assignment of context.assignments) {
    if (!assignment || typeof assignment !== "object" ||
        !["control", "variant"].includes(context.content_variant) ||
        assignment.value !== context.content_variant || typeof assignment.experiment_id !== "string" ||
        typeof assignment.variation_id !== "string") continue;
    const event = { experiment_id: assignment.experiment_id, variation_id: assignment.variation_id,
      content_variant: context.content_variant, source: context.source, event_id: crypto.randomUUID() };
    if (!navigator.sendBeacon("/experiment-view", new Blob([JSON.stringify(event)],
      { type: "application/json" }))) console.warn("Strapi exposure was not queued");
  }
})();

This boundary establishes that browser execution reached the rendered page. It does not prove viewport visibility or reading. Use a symmetric visibility condition if that is the intended exposure definition. A successful beacon queue operation also does not prove network acceptance or durable ingestion.

The data-element marker prevents duplicate execution within one document. Reloads can produce additional displays, and collector retries need their own event-ID policy. Define whether analysis uses first exposure or another consistent rule, then make the assignment query match that definition.

Replace local logging with verified ingestion

The diagnostic collector reads identity from the HttpOnly cookie, validates the event/source fields and discards unexpected fields. It still receives browser-supplied observations. Do not use that metadata as an authorization or billing decision, and do not treat a console row as a warehouse record.

Connect your supported analytics ingestion with appropriate validation, origin controls, delivery monitoring, retries and deduplication. Apply consent and tracker-readiness behavior consistently across both arms. Compare assignments, browser events, accepted requests and warehouse rows when investigating loss; those counts represent different boundaries.

Configure the GrowthBook data source to read experiment key, variation ID, timestamp and assignment unit from the actual stored schema. Preserve the document/locale/hash provenance in the event or a linked experiment record so the content behind a result remains identifiable.

Validate that mapping with one controlled event before increasing traffic. Compare the browser request, accepted collector fields and resulting warehouse row. Field names can change through an analytics transformation, timestamps may be stored in different zones, and a custom identifier can be dropped by a tag or export. Finding the row and joining its outcome is stronger evidence than seeing a successful HTTP response from the local collector.

Join the business outcome to the same identifier

Define completed signup or another chosen outcome with Fact Tables and metrics. Verify that it joins to the browser assignment identifier or a tested authenticated-identity mapping. A signup on another origin does not automatically carry this cookie.

Inspect an actual controlled exposure and outcome together, including timestamps, duplicate handling and follow-up window. Recent visitors may not yet have had enough time to complete the outcome. Keep diagnostic clicks and qualified conversions separate so faster but less useful signals do not replace the intended decision.

Verify publication and maintain a clean test

Check HTTP behavior before interpreting results

With the reviewed hash configured, run the application and establish a local cookie jar:

node --env-file=.env.local server.mjs
curl -i -c strapi-test-cookies.txt http://localhost:3000/experiment-session
curl -s -b strapi-test-cookies.txt http://localhost:3000/offer
curl -s -b strapi-test-cookies.txt http://localhost:3000/offer

Expect a fixed redirect and consistent selected content for the same identifier while configuration remains stable. These requests inspect server HTML; they do not execute the browser exposure script. Use a real browser to inspect the beacon and collector, then verify the actual analytics and warehouse path separately.

Exercise both arms with controlled identities. A forced value can prove a rendering branch, but it does not prove ordinary randomized enrollment. Keep preview, forced and employee QA traffic outside production analysis.

Persistence is also narrower than sticky assignment across arbitrary configuration changes. The same browser ID can be used consistently while a changed experiment key, rule or allocation alters the experience. Freeze the intended configuration during the comparison, record deliberate changes, and use the platform's supported bucketing behavior when your requirements extend beyond a stable identifier.

Test the modern contract and its failure cases

Test conditionExpected behaviorWhat to investigate
v4-style nested responseReject missing flat fieldsMigration compatibility and parser version
Numeric row ID used as document IDNo successful content displayActual documentId and generated endpoint
Draft or wrong-locale payloadReject before assignmentStatus, localization and preview configuration
Reviewed string changesHash mismatch before enrollmentPublication approval and treatment version
SDK, cookie or event failureExplicit baseline/unavailable/diagnostic pathIdentity continuity and delivery loss

Also test HTML-significant characters and a closing-script sequence in approved QA content. They should remain text, and the data element should remain intact. Repeat cache and identity checks on the actual deployment: local success does not establish proxy behavior, public rendering or warehouse attribution.

Include a schema deployment in the regression check when field names or content structure change. A correct new schema in development can coexist with older published documents or a renderer expecting the previous shape. Verify the deployed model, the delivered payload and the rendering contract together. Keep a clear unavailable/diagnostic result for incompatible content rather than disguising it as a valid control observation.

Close the loop with an owned content decision

Before interpreting lift, investigate allocation anomalies, missing data and variation-specific failures. Agree on duration, outcome maturity, meaningful effect and a decision framework before seeing the result. An imprecise estimate can leave the content choice unresolved, and nonsignificance does not prove equivalence.

When the test ends, stop enrollment, verify payload refresh in deployed instances and publish the chosen headline as ordinary content. Preserve the approved pair, source document/locale, fingerprint and metric definition with the learning. Remove retired assignment and tracking branches through review so future tests begin with a clear owner and a clean content contract.

Coordinate the content promotion with the fingerprint gate. Turning the winning headline into the new control changes the reviewed pair, so approve that resulting baseline and its new digest deliberately. Test the publication/configuration sequence in staging; do not clear a mismatch by copying a new hash without checking what it represents. The final baseline should work independently of the retired experiment rule and its exposure code.

Run a clear content comparison

Connect published Strapi content to consistent assignments and the business outcomes your team already measures. Start with a small test and verify the complete delivery and measurement path.

Get Started With GrowthBook
Experiments
Analytics

What is warehouse-native experimentation and why does it matter?

Nov 11, 2026
x
min read

An experiment should use the same definition of revenue that your finance team trusts.

The checkout test says revenue increased. The finance dashboard says net revenue was flat. Product sees a winner; finance sees refunds, discounts, and failed payments that never reached the experimentation tool. Before anyone can decide what to ship, someone has to reconcile 2 versions of the business.

Warehouse-native experimentation addresses that architectural problem. It analyzes experiments using data and metric logic in your data warehouse, so exposure events can be joined to the same orders, subscriptions, support records, and other models used elsewhere in the company.

The benefit goes beyond storing experiment events in a familiar database. The warehouse becomes the measurement foundation. Your application still assigns and delivers variations. Your data pipeline records what happened. An experimentation layer queries the relevant warehouse data and applies statistical analysis to support a decision.

That separation matters. Warehouse-native does not mean a customer request must wait for a warehouse query, that every calculation happens inside SQL, or that owning the data automatically makes an experiment trustworthy. It changes where measurement draws its evidence, while leaving important work around identity, metric definitions, data freshness, and operating costs.

What warehouse-native experimentation actually means

Separate delivery from measurement

Think of an experiment as 2 connected paths. The delivery path determines whether an eligible user or account receives control or treatment. The measurement path records the assignment or exposure, connects it to subsequent outcomes, and estimates the difference between variations.

In a common implementation, a feature flag SDK evaluates a downloaded configuration in your application. Exposure and outcome events travel through an event pipeline into your warehouse. Analysis queries then join those records and produce the inputs needed for statistical inference. A warehouse outage can delay analysis without necessarily interrupting flag evaluation; the exact failure behavior depends on the delivery implementation and its cached configuration.

GrowthBook's delivery and analysis architecture illustrates this separation. Its normal bring-your-own-warehouse experiment analysis retrieves aggregate information such as sums and sums of squares, then uses the GrowthBook statistics engine to produce the report. Warehouse queries perform data selection and aggregation; saying all statistical inference executes inside the warehouse would overstate that boundary.

This also means you can evaluate a measurement layer separately from an assignment mechanism. A team might keep an existing assignment service while adopting warehouse-based analysis. The important requirement is a reliable record of who entered which variation and when, rather than a particular vendor generating that record.

Inspect the data path behind the label

“Warehouse-native” describes an architectural approach, not a complete specification. Ask which data is queried, which data is copied, what query results leave the warehouse, and what must be configured before an experiment can run.

These patterns are useful distinctions rather than mutually exclusive vendor categories:

Measurement patternPrimary evidence used for analysisWhat to inspect
Vendor event storeEvents copied into a vendor-controlled analytics storeEvent definitions, retention, export, and reconciliation with business models
Warehouse-connected importSelected warehouse data imported for downstream analysisImport schedule, transformation logic, copied fields, and historical updates
Warehouse-native analysisQueries against governed warehouse dataAssignment joins, metric logic, query output, permissions, and refresh behavior

An export back to your warehouse does not by itself establish that the platform analyzes your governed warehouse models. Conversely, returning aggregates to an analysis service does not make the warehouse irrelevant. The practical question is whether the experiment's evidence and definitions come from the data foundation you intend to govern.

For GrowthBook, the available data paths include connecting an existing warehouse and pipeline, using an Event Forwarder that writes events into your warehouse, or using a managed warehouse. Those paths have different ingestion and ownership boundaries. Choose the one you actually need before making assumptions about where raw events travel.

Why using warehouse data changes product decisions

Measure the business outcome beyond the click

A product event is often an early signal, not the entire outcome. A completed checkout can later become a refund. A new subscription can fail its first payment. A support interaction can appear resolved and reopen tomorrow. If your experimentation tool only receives the initial event, it may reward activity that does not create durable value.

Warehouse-based measurement makes it easier to join these stages when the underlying records and identifiers exist. For a pricing experiment, you might compare net revenue per assigned account after refunds and credits, while monitoring trial conversion and cancellation. For onboarding, you might measure activated accounts rather than the number of people who clicked the final setup button.

This is not a reason to make every metric complicated. Start with the decision. If the decision depends on paid activation, define paid activation clearly and trace the required records. Additional joins should answer a business question; joining every available table can make a metric slower and harder to interpret without improving the decision.

Reuse definitions without confusing populations

A shared warehouse model can reduce discrepancies caused by separate transformation pipelines. It does not mean an experiment report should exactly equal a company dashboard. Finance may report all recognized revenue in a calendar month. Your experiment may report revenue from eligible assigned accounts within 14 days of first exposure. Both can be correct while answering different questions.

Write down the differences in population, unit, event time, attribution window, currency conversion, and exclusions before reconciling totals. “Revenue” is a label. The actual definition includes all of those choices.

Metric definitions in a semantic modeling system can centralize business logic, but experiment analysis still needs the relevant population and variance information. A monthly revenue total is not enough to estimate uncertainty for a user-randomized experiment. Confirm that an integration preserves the grain and statistical inputs the analysis requires; do not assume every semantic-layer metric can be consumed unchanged.

GrowthBook's Fact Tables and metrics provide a way to build multiple metrics on shared warehouse query definitions. That makes a governed orders model useful across experiments while preserving explicit choices about the outcome being measured.

Consider a hypothetical subscription test with 2,000 assigned accounts in each variation. Treatment produces 40 additional initial purchases, but those purchases also generate additional refunds and support work. A click or gross-purchase metric cannot tell you whether the change improves contribution per assigned account. Connecting the relevant business models lets you ask that question with a consistent denominator and a defined observation window.

The warehouse does not make the definition automatic. Decide whether the metric includes tax, credits, currency conversion, and costs that arrive after the initial purchase. Define whether an account with no purchase contributes zero or is excluded; for revenue per assigned account, excluding it would answer a different question. Put those decisions in the metric definition rather than leaving each analyst to reconstruct them during a readout.

Keep exploratory analysis distinct from the original decision

Existing warehouse history can support adding a metric after an experiment starts, provided the necessary data was collected and retained. That is useful when an unexpected result raises a new question. It is not permission to keep searching for a favorable metric and present it as the original success criterion.

Record the primary metric and guardrails before launch. Label later slices and metrics as exploratory, preserve the original decision record, and confirm a promising new hypothesis in a subsequent experiment when needed. Access to more data improves investigation; it also makes selective storytelling easier if the analysis process is loose.

Historical reanalysis needs versioning as well. If a model changes how refunds are allocated, a refreshed experiment result can change even though user behavior did not. Keep the query or model revision and refresh timestamp with the result so the team can explain why the answer moved.

Turn your model into a metric

Open the metric creation workflow to define a fact metric from warehouse data. Configure GrowthBook API access first, then review the proposed definition before creation.

View The Metric Workflow

The data contracts that make results trustworthy

Record exposure at the right point

Assignment, evaluation, and actual delivery are related events, but they are not always identical. Your code can evaluate a feature before the user reaches the relevant screen. A request can be assigned treatment and fail before it returns an answer. Define which event makes a subject part of your analysis and ensure the implementation matches it.

The minimum experiment record usually includes an experiment identifier, variation identifier, assignment unit identifier, and timestamp. GrowthBook's assignment query contract formalizes those fields and supports identity joins when exposure and conversion use different identifiers.

Preserve repeated exposure records when required by your platform; they can help reveal crossover between variations. Do not deduplicate the source events by dropping inconvenient rows before you understand them. The analysis layer can have its own subject-counting rules while the underlying event history remains available for diagnosis.

Match the assignment unit to the metric grain

If the product experience is shared within an account, randomizing accounts may be more appropriate than randomizing individual users. The analysis then needs account-level outcomes or a method that correctly handles the cluster structure. Counting every seat as an independent randomized subject exaggerates the amount of independent information.

The same care applies to joins. Joining one order to several product events can multiply its value. Joining only purchasers removes assigned non-purchasers from a conversion denominator. A valid revenue-per-assigned-account metric normally needs the assigned population, including zero-revenue accounts, with outcomes attached at the intended grain.

Before trusting a statistical result, inspect row counts before and after joins, distinct assignment units, missing identifiers, multiple variations per subject, and outcome totals. A warehouse gives you the ability to investigate those checks directly. It cannot decide the correct unit or join logic on your behalf.

Test the measurement system before the hypothesis

An A/A test assigns equivalent experiences to different labels. Use it to exercise assignment, event collection, identity mapping, metric queries, and reporting before a consequential treatment. It can reveal problems, but a single nonsignificant result is not proof that the entire measurement system is sound.

Check for sample ratio mismatch, or SRM: an unexpected difference between observed and configured allocation proportions. Research on SRM diagnosis shows why this is an indicator of several possible implementation and data-quality problems. Investigate the cause rather than treating the check as a cosmetic warning.

Also compare event arrival by variation, missingness by device or geography, and the timing of identity changes. An apparently balanced assignment can coexist with biased outcome collection. These audits matter whether the warehouse is yours or someone else's.

Preserve the information your analysis needs

Pre-aggregation can improve query performance, but an arbitrary summary table may erase information needed for experimentation. A table with one row of monthly totals per variation cannot reveal users exposed to multiple variations or support every analysis of user-level variance. Decide what must remain available before reducing the data.

Work backward from the metric and statistical method. A proportion metric needs a defined population and an indicator for whether each subject met the condition. Revenue per account needs the relevant account-level amounts and an appropriate treatment of zero outcomes. Ratio and quantile metrics have additional requirements; they should not be reconstructed from whichever averages happen to appear in a reporting table.

Keep the raw or sufficiently detailed governed history available under the appropriate access policy. This does not mean every reader needs raw-event access. It means the system has a controlled route for diagnosing why an aggregate changed. Optimization should preserve the evidence needed to challenge a result, not only the number displayed in the final report.

The operational tradeoffs you still need to manage

Freshness is an entire pipeline

A result cannot include an outcome that has not happened or data that has not arrived. End-to-end freshness depends on event ingestion, transformations, warehouse availability, query scheduling, query runtime, and the outcome's own observation window.

If events arrive every few minutes but the financial model rebuilds nightly, net revenue is not a minute-by-minute metric. If cancellation is measured over 30 days, a dashboard refreshed every minute still cannot provide a mature 30-day outcome for yesterday's assignments. Display the latest source-data timestamp as well as the report refresh time.

GrowthBook exposes an experiment auto-update frequency setting. Choose it around the decision cadence and available data rather than assuming frequent refresh creates new information. Operational monitoring may need a separate low-latency error signal while the experiment measures slower business outcomes.

Query cost is part of experimentation cost

Using data you already store avoids some duplication, but warehouse compute is still a real cost. More experiments, metrics, dimensions, refreshes, and long observation windows can increase scans and concurrency. Compare the complete cost of the workflow, including platform fees, warehouse compute, pipeline work, and analyst time.

Start with bounded dates and selected columns. BigQuery query optimization guidance explains why scanning fewer columns and relevant partitions can reduce work. Verify the generated query and its execution details; adding a limit to a broad scan is not equivalent to pruning the source data.

In Snowflake, virtual warehouses provide query compute. Choose workload resources and operating settings deliberately so an experimentation refresh does not compete unpredictably with another critical workload. A larger warehouse or more concurrency may reduce waiting while increasing spend. Measure the effect instead of assuming the most powerful setting is the most efficient.

Optimize with a clear correction policy

Shared Fact Table queries, materialized models, and incremental processing can avoid repeated work. GrowthBook's query optimization options include several of these approaches, subject to metric and warehouse restrictions. Savings depend on the workload; there is no universal percentage that applies to every experiment program.

Caching also has conditions. BigQuery's cached query results can avoid recomputation for eligible repeated queries, but changes to data or query text and other exceptions can prevent reuse. Include cold runs and representative refreshes when estimating cost, not only the fastest cached result.

Incremental systems need a policy for corrections to older records. GrowthBook's Pipeline Mode writes reusable intermediate tables and documents limitations around historical changes and refresh behavior. If late data carries an older source timestamp, do not assume a new incremental update will incorporate it. Know when a full refresh is necessary and who initiates it.

Separate event time from ingestion time in this policy. A payment adjustment can arrive today while referring to an order from last week. If the incremental process only reads source rows whose event timestamps exceed its previous maximum, that adjustment can be missed. A full refresh may correct the result, but the team still needs to communicate why an earlier readout was restated.

Track the expensive parts of the workflow. Snowflake Query History provides a route for inspecting executed queries and their performance. Use the equivalent job or query history in your warehouse to connect failed refreshes and resource use to specific workloads. A cost estimate should include the correction path as well as the routine incremental update.

Data ownership needs explicit access boundaries

Scope the connection to the intended workflow

A warehouse connection should use an identity with the access required for its actual job. Restrict readable schemas or views, avoid unrelated personal fields, and give the workload an identifiable role or service account. Snowflake's access-control model offers object privileges and roles for implementing those boundaries.

Distinguish user-authored read queries from platform-managed writes. A read-only analysis connection and a pipeline that creates intermediate tables have different permission needs. If you enable a write-based optimization, scope the destination and retention policy deliberately; do not grant broad production ownership merely to make setup easier.

Test access with the identity the application actually uses. A query that works for an administrator may fail for the service account, while a service account with broader privileges may see more data than a human reviewer. Document the effective query identity along with the intended access policy.

Describe what crosses the boundary accurately

Warehouse-native does not mean nothing ever leaves the warehouse. Queries, connection metadata, aggregate results, exports, diagnostics, and event-ingestion services each have their own data flows. In a normal GrowthBook bring-your-own-warehouse experiment-results path, raw event processing happens in the warehouse and aggregates return for statistical analysis. Other selected features and data paths require their own review.

Aggregated does not automatically mean anonymous. A tiny segment or a sufficiently detailed grouping can still reveal sensitive information. Review which dimensions are queryable, who can export results, how small groups are handled, and what the interface allows a user to inspect.

Row-level access policies in BigQuery can restrict which rows an identity sees. That affects interpretation as well as security: two analysts using different effective policies may obtain different populations from similar-looking SQL. Preserve the relevant access context when reproducing a result.

Choose deployment as a separate decision

Connecting your warehouse is not the same as hosting the experimentation application yourself. Cloud and self-hosted deployments change responsibility for the application, credentials, networking, updates, and operational support. GrowthBook's deployment options make those choices explicit.

Evaluate the arrangement your team can operate reliably. A self-hosted system still needs upgrades, backups, monitoring, and an owner. A cloud service still needs suitable access controls and a clear data-flow review. Architecture can support a security requirement; it does not replace your organization's assessment of the configured system.

Start with one experiment you can reconcile

Use a bounded pilot to prove the contracts

Choose one product change, one assignment unit, a primary metric, and a small set of guardrails. Prefer a workflow whose events and business outcomes already exist in the warehouse. Avoid making your first pilot depend on a new identity graph, a new financial model, and a new assignment system simultaneously.

Write an acceptance plan before connecting the pieces. It should identify the exposure definition, source tables, identifier mappings, observation window, refresh cadence, access identity, and expected query workload. Include the person responsible for each component so a missing outcome has somewhere to go besides a general experimentation channel.

Run the analysis against a known time window and reconcile the assigned population and metric totals with an independent warehouse query. Use the same filters and attribution rules. If the numbers differ, inspect the population and grain before comparing statistical engines. A mismatch caused by a duplicated join will not be fixed by choosing a different confidence interval.

A concise pilot record can make the handoff concrete:

Acceptance questionEvidence to retainOwner to name
Did eligible subjects receive a stable variation?Assignment/exposure records and crossover checksApplication engineering
Do outcomes join at the intended grain?Identifier audit and reconciled metric totalsData or analytics engineering
Is the readout current enough for this decision?Source cutoff, refresh time, and maturity ruleExperiment owner
Can this workload operate at expected volume?Query history, refresh cost, and correction planData platform

Define a response when one of these checks fails. A stale model should delay interpretation or trigger a clearly labeled stale-data state. A missing source should not become a zero-valued metric. A failed refresh should preserve the last successful result with its original timestamp rather than presenting it as newly computed. These behaviors make the system usable under ordinary operational failures.

Preserve a result someone else can explain

Keep the experiment configuration, metric definitions, query/model versions, data cutoff, refresh time, health checks, and decision together. Record whether a metric was planned or added during investigation. When a later refresh changes the result, the team should be able to distinguish newly arrived outcomes from changed logic.

GrowthBook's experiment configuration connects delivery choices with analysis settings, while warehouse-native measurement lets your team use the data foundation it already maintains. The product is most useful when that foundation has clear contracts and owners.

Expand after the pilot demonstrates that you can explain the joins, reproduce the metric, observe data freshness, and estimate refresh cost. The reason to adopt warehouse-native experimentation is to make product decisions from a consistent, inspectable measurement system. The architecture earns its value when a surprising experiment result leads to a productive investigation instead of another argument over whose dashboard is right.

Start with the data you trust

Connect your experimentation workflow to GrowthBook and choose the warehouse data path that fits your team's infrastructure and operating responsibilities.

Get Started With GrowthBook

Ready to ship faster?

No credit card required. Start with feature flags, experimentation, and product analytics—free.

Simplified white illustration of a right angle ruler or carpenter's square tool.White checkmark symbol with a scattered pixelated effect around its edges on a transparent background.