Experiments

Feature Flags

Analytics

How to migrate from PostHog to GrowthBook: A step-by-step guide

Oct 4, 2026

min read

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

Replacing a flag SDK can change who sees a feature. Replacing an analysis pipeline can change which outcomes count. A migration has to control both.

Moving from PostHog to GrowthBook is a migration of application behavior and measurement contracts, not just a configuration export. The work includes identifying which capabilities move, preserving the meaning of identities and variants, connecting exposure records to business metrics, and proving a recovery path.

You can migrate experimentation while retaining PostHog for product analytics, session replay, or other workflows you still need. You can also analyze externally assigned experiments in GrowthBook before changing the allocator. Neither approach requires treating every part of your current stack as obsolete.

This guide uses a staged migration of one string-valued checkout flag. It includes a local mapping fixture, a warehouse audit pattern, and a current Node.js SDK integration. Adapt the example to your application and warehouse. The central rule is to keep one authoritative assignment path for each running experiment and to verify behavior before expanding production traffic.

1. Define what moves and freeze a migration contract

Separate assignment, analysis, and product analytics

List the responsibilities PostHog currently owns. These may include feature evaluation, experiment configuration, event capture, identity resolution, metrics, dashboards, and operational tools. Mark each as moving, remaining, or needing a separate decision.

PostHog feature flags already support targeting and product-release workflows linked to person and group data. Keeping those connections can be useful. GrowthBook becomes a relevant migration target when you want its experimentation workflow, data architecture, release controls, or deployment model. Evaluate the GrowthBook and PostHog fit against your actual requirements rather than assuming one product must replace the other's entire suite.

Distinguish 3 possible projects. Moving analysis means GrowthBook reads assignment and outcome data while the incumbent still assigns users. Moving flag delivery means the application evaluates GrowthBook flags. Moving analytics means replacing some event collection, reports, or exploration workflows. They have different acceptance criteria and can happen at different times.

Start with one low-risk, reversible capability that has an owner, clear targeting, and a maintained baseline. Avoid making an active high-value experiment or a complex account-entitlement rule your first cutover. Finish ongoing experiments under their existing allocator where feasible, then start new experiments on the new path.

Inventory behavior, not only flag names

For each candidate, record the key, type, configured default, application fallback, environments, ordered rules, targeting properties, group or person scope, rollout allocation, payloads, prerequisites, and code references. A matching key does not prove matching behavior.

For experiments, add the experiment identifier, variation identifiers, baseline, allocation, exposure event, analysis unit, start date, metrics, windows, statistical method, and stopping plan. PostHog's experiment configuration links experiments and variant behavior; inspect the actual configuration instead of assuming every test uses a particular pair of variant names.

Record side effects too. The checkout example may trigger a different API call, write a different schema, or enqueue work. A flag rollback only restores a code path if that path can still handle the resulting state. Name the owner who can approve the move and the signal that would cause a rollback.

Existing concept	Migration decision	Evidence before cutover
Flag key and value type	Preserve or explicitly map names and types	Known inputs produce expected application values
Person or group targeting	Define equivalent application attributes	Representative users and accounts match intended eligibility
Experiment variants	Map labels, IDs, and the baseline	Exposure records join to the correct variation
Metric definition	Recreate behavior, window, and denominator	Known-case records reconcile
Disable behavior	Choose the configured default and code fallback	Both are tested independently

Do not copy only the visible rollout percentage. Rule order, conditions, identity, and hash inputs can affect which specific units receive treatment. Percentage equality is an aggregate property, not a continuity guarantee.

Save an executable mapping fixture

Keep a small reviewed manifest alongside the migration work. The example below describes a proposed mapping; it is not an import payload for either product. Save it as migration-map.json and replace the synthetic details with the contract for your staging exercise.

{
  "sourceFlag": "checkout-layout",
  "targetFlag": "migration-checkout-layout",
  "targetExperiment": "checkout-layout-gb-v1",
  "unit": "user_id",
  "safeValue": "control",
  "values": ["control", "treatment"],
  "fixtures": [
    {"id": "migration-user-control", "expected": "control"},
    {"id": "migration-user-treatment", "expected": "treatment"}
  ]
}

The source and target values here intentionally share labels. A real migration can map different labels, but the mapping must be explicit and applied consistently in the application and analysis.

Save the following as check-map.mjs, then run node check-map.mjs. It uses Node's strict assertions to catch basic contract errors before any remote configuration changes.

import assert from "node:assert/strict";
import { readFileSync } from "node:fs";

const map = JSON.parse(readFileSync("migration-map.json", "utf8"));
assert.equal(map.unit, "user_id");
assert.ok(map.values.includes(map.safeValue));
assert.equal(new Set(map.values).size, map.values.length);
assert.equal(new Set(map.fixtures.map(x => x.id)).size, map.fixtures.length);
for (const fixture of map.fixtures) {
  assert.ok(fixture.id);
  assert.ok(map.values.includes(fixture.expected));
}
console.log("Migration mapping fixture passed.");

Expected output is Migration mapping fixture passed. A pass checks the local fixture's consistency. It does not confirm that either platform has been configured to match it. Preserve the manifest version so later test evidence refers to a known contract.

2. Make identities and warehouse records agree

Choose the identifier before copying targeting rules

Decide whether the experiment assigns users, accounts, devices, or sessions. The example uses a stable signed-in application user ID. That ID must be present in both the evaluation context and the outcome data you intend to analyze.

PostHog's identification behavior can link anonymous and identified activity. Passing one current ID into GrowthBook does not automatically reproduce that historical identity graph. Inspect how logged-out traffic, login transitions, logout resets, multiple devices, and merged profiles are represented in your data.

For a first migration, limiting eligibility to stable signed-in identities can reduce complexity when the use case permits it. If anonymous behavior is essential, define a persistent identifier and an explicit rule for what happens at login. If collaboration is account-scoped, use a stable account assignment and ensure the metric query analyzes that unit correctly.

The evaluation-context contract provides a useful way to think about identity and targeting inputs. It does not standardize every provider's hashing or preserve a previous platform's assignment. Using the same ID and flag key in both products is insufficient evidence that the same people will receive treatment.

Define a canonical exposure record

GrowthBook needs to know which unit was in which experiment, which variation it received, and when. Use data source and assignment-query configuration to map your actual warehouse records into that contract.

For the example, a modeled exposure table could contain user_id, timestamp, experiment_id, and variation_id, with a separate platform or phase field where needed. These are fields you define in your pipeline, not a claim about PostHog's raw export schema.

Keep the experiment namespace distinct across assignment phases. For example, preserve a legacy experiment ID for existing assignments and use checkout-layout-gb-v1 for a new GrowthBook experiment. Mixing both under one identifier can hide treatment switching and differences in eligibility.

Verify the warehouse join with a small fixture before connecting production tables. This read-only PostgreSQL example uses a left join so unmatched outcome identities remain visible. Review join semantics when adapting it to your warehouse.

WITH exposures(user_id, variation_id) AS (
  VALUES ('user-a', 'control'), ('user-b', 'treatment')
), outcomes(user_id, amount) AS (
  VALUES ('user-a', 10), ('user-c', 20)
)
SELECT
  o.user_id,
  e.variation_id,
  o.amount,
  CASE WHEN e.user_id IS NULL THEN 'unmatched' ELSE 'matched' END AS status
FROM outcomes o
LEFT JOIN exposures e ON e.user_id = o.user_id
ORDER BY o.user_id;

Expected results include user-a matched to control and user-c unmatched. An inner join would remove the unmatched row and conceal that gap. In real data, investigate the cause rather than automatically treating unmatched events as failures or excluding them without explanation.

Also audit duplicated exposures, duplicate event ingestion, timestamps, missing variation labels, and multiple variations per unit. These checks need the full experiment and time context; repeated legitimate exposures are not necessarily duplicate assignments.

Preserve history without pretending the allocator moved

Importing an externally assigned experiment into GrowthBook requires the assignment data to exist in a linked data source. Configure and test the query, then review the inferred experiment metadata, including variations, allocation, and start date.

This can let you analyze a legacy experiment using warehouse outcomes. It does not copy PostHog feature rules into GrowthBook or change which system serves users. Keep the original allocator authoritative while that experiment runs.

Do not expect identical result screens just because both analyses read similar events. Metric definitions, eligibility, windows, handling of repeat exposures, statistical methods, and data freshness can differ. Reconcile raw counts and metric construction before comparing a headline uplift or probability.

Prove the data path before importing history

If the needed events are already in your warehouse, identify their modeled tables and owners. If they exist only in the incumbent system, choose and validate a supported export or delivery path before treating them as available to GrowthBook. Historical backfills and ongoing delivery can have different completeness and latency characteristics.

Use the actual schema to build the assignment query, including the correct identifier type and variation mapping. Check a bounded interval with known test subjects. Inspect the earliest and latest event times, unmatched identities, missing variation values, and repeated records. Record the data cutoff and any known backfill gaps.

Test the warehouse connection with the permissions intended for normal operation. A connection that works only with a broad administrator credential has not established the production access model. If a query fails because a view is missing, a schema is inaccessible, or a type cast is invalid, fix that specific dependency before importing the experiment. Do not replace missing data with an invented event or a convenient count.

Finally, preserve the original event time separately from ingestion time. Late delivery can affect freshness without changing when the user was exposed. A backfill received today should not silently turn an old exposure into a new experiment participant.

Check the migration architecture

Compare the responsibilities you want to move with GrowthBook's experimentation and feature management model before choosing the first cutover.

Compare the Platform Fit

3. Integrate one GrowthBook flag in staging

Create the target with a tested baseline

Create a staging GrowthBook feature using the target key from the manifest, with a string value and a default of control. Follow feature key, type, environment, and default-value configuration. Use a dedicated staging SDK connection and confirm the environment before testing.

For this initial routing check, add an explicit rule serving treatment to migration-user-treatment. Let migration-user-control receive the baseline. Review and publish the staging configuration through your normal workflow. This is a controlled fixture, not an experiment allocation.

Then verify the 2 application paths. The baseline must render the existing checkout and the treatment must render the intended replacement. Test the default when no rule matches and the application fallback when configuration cannot be resolved. Typed fallback values are part of the evaluation contract; choosing one is separate from proving it is safe.

Keep administrative credentials out of application code. A GrowthBook SDK client key serves feature configuration; it is not the management API credential used to create or publish resources. Use environment-specific configuration and the deployment's documented credential handling.

Install and run the staging smoke check

In a small Node.js project, install the official SDK with npm install @growthbook/growthbook. Configure GROWTHBOOK_SDK_CLIENT_KEY for the staging SDK connection. Set GROWTHBOOK_API_HOST if the connection uses a different host. Use the host and key displayed for your actual connection, rather than copying production credentials into the fixture.

The current GrowthBook Node.js SDK uses a shared GrowthBookClient and a per-evaluation user context. Save the following as check-staging.mjs next to the manifest and run node check-staging.mjs after activating the staging rules.

import assert from "node:assert/strict";
import { readFileSync } from "node:fs";
import { GrowthBookClient } from "@growthbook/growthbook";

const map = JSON.parse(readFileSync("migration-map.json", "utf8"));
const clientKey = process.env.GROWTHBOOK_SDK_CLIENT_KEY;
assert.ok(clientKey, "Set the staging GROWTHBOOK_SDK_CLIENT_KEY");

const viewed = [];
const client = new GrowthBookClient({
  apiHost: process.env.GROWTHBOOK_API_HOST || "https://cdn.growthbook.io",
  clientKey,
  trackingCallback: (experiment, result, context) => {
    viewed.push({
      user_id: context.attributes.id,
      timestamp: new Date().toISOString(),
      experiment_id: experiment.key,
      variation_id: String(result.key),
    });
  },
});

const initialized = await client.init({ timeout: 3000 });
assert.ok(initialized.success, "Staging configuration did not load");

for (const fixture of map.fixtures) {
  const value = client.getFeatureValue(map.targetFlag, "__missing__", {
    attributes: { id: fixture.id },
  });
  assert.equal(value, fixture.expected, `Unexpected value for ${fixture.id}`);
  console.log(JSON.stringify({ id: fixture.id, value }));
}
console.log(JSON.stringify({ localExposureReceipts: viewed }));

Expected values are control and treatment for the corresponding synthetic users. An initialization failure or unexpected value should stop the exercise. Check the connection, environment, published rules, value type, and attribute names before changing the expectations.

The initial forced-value fixture may produce no experiment receipts, because a force rule is not an experiment rule. That is expected. The callback only becomes a measurement path when an applicable experiment is evaluated. The array above is local diagnostic storage, not a durable event pipeline.

Repeat the routing check with an unknown synthetic ID, an omitted targeting property, and the safe default. Also run it against an intentionally invalid SDK key in a separate test configuration. The smoke check should fail initialization rather than produce a misleading pass. In the application, test the configured outage policy without exposing a broken customer experience.

Add explicit cases for every rule you migrate. For example, test an eligible paid account and an ineligible free account, a missing account ID, and the production boundary that should never use a staging connection. These fixtures should be maintained with the mapping contract, not created from a few convenient users who happen to receive the expected value.

Wire exposure delivery before measuring a real experiment

For a new experiment, configure the relevant GrowthBook experiment and feature rule, then evaluate at the actual application exposure boundary. Send the callback's receipt through your existing approved event pipeline, using the same identity and variation mapping your assignment query expects.

Do not copy the local array into production and assume events are retained. Verify delivery, retries, deduplication, timestamp handling, and warehouse arrival. Preserve enough metadata to distinguish platform and experiment phase without collecting unnecessary personal information.

After removing the temporary forced fixture rules, verify normal experiment assignment with a suitable synthetic staging population. Check that records contain the new experiment ID and the intended variation IDs. Repeated calls should not create an accidental double-counting problem in the modeled assignment data.

Keep shadow evaluations out of the production experiment exposure stream when they do not serve a user experience. Evaluating both providers for comparison can be useful for ordinary flag behavior, but logging both as genuine exposures changes what the analysis population represents.

4. Validate new experiments and cut over deliberately

Compare contracts before comparing estimates

Rebuild the selected metrics using GrowthBook Fact Tables and metric windows. Start with a small set: the outcome that decides success, a few useful diagnostics, and guardrails for unacceptable harm. Record numerator, denominator, identity, time bounds, filters, and preferred direction.

Reconcile known records. For revenue, distinguish an order event from settled net revenue. For conversion, decide whether the metric means at least one qualifying action per assigned user or a count of repeated events. For retention, give every included cohort the follow-up period that the definition requires.

Compare totals over the same frozen data interval. If one system uses events ingested up to noon and another includes later data, a difference does not yet establish a calculation defect. Save the query settings and data cutoff with the reconciliation receipt.

Review the analysis method and stopping plan as a separate contract. Moving to another engine does not justify selecting whichever result is most favorable. Sequential inference requires a compatible procedure when decisions depend on repeated observation. Preserve the plan for a running legacy experiment or explicitly start a new analysis phase with its limitations documented.

Run a new A/A test and investigate unhealthy data

Before relying on a new allocation and exposure pipeline, run an A/A exercise with identical application behavior in both variations. Inspect the expected split, repeated identities, exposure timestamps, join coverage, and plausible metric values. One inconclusive A/A result is not proof that everything is correct.

Sample ratio mismatch is a symptom to investigate. It can arise from assignment, execution, logging, joins, or selection in analysis. Do not fix an unhealthy split by deleting inconvenient observations until the result looks balanced.

Inspect representative users across both application paths. Confirm that a backend event uses the same ID as the evaluation context, that a login transition does not silently alter the unit, and that no client still records the old experiment ID for the new allocation.

For a real product experiment, plan power and duration around the actual metric and detectable effect. A successful migration smoke test provides no evidence that your business hypothesis has enough sample size.

If a metric differs, debug it in a stable order. First compare exposure counts and the selected population. Then compare event counts, identifier joins, metric windows, and per-unit aggregation. Only after those agree should you compare the statistical output. Changing several settings at once makes the cause harder to identify.

Choose an investigation policy before the production cutover. For example, a missing-variation value should block using that record for interpretation until its source is understood; an unexplained drop in join coverage should pause expansion. Set practical thresholds from your baseline and risk tolerance instead of copying a universal percentage from a migration checklist.

Cut over one authoritative path

Use a routing decision owned by the application to select the provider for the migrated capability. Roll out that provider choice through a controlled cohort where appropriate. Log which provider served the real experience so incident investigation can identify the active path.

Avoid splitting an ongoing experiment's allocation between independent PostHog and GrowthBook hashes and then pooling the data as one unchanged test. Either finish it under the incumbent, preserve an explicitly designed external assignment scheme, or end that phase and begin a new experiment. The last option needs attention to prior exposure and carryover, not just a new identifier.

Before expansion, prove that switching back restores the intended baseline for future requests. Check in-flight work and changed data state separately. If a new schema or side effect makes the previous code path incompatible, provider routing alone is not a sufficient rollback.

Keep a small cutover receipt with the application version, provider route, flag and experiment IDs, configuration revision, start time, intended population, observed values, metric health, and rollback owner. Then expand only after the agreed checks pass.

Keep that provider route stable at the appropriate unit boundary. A user should not silently alternate between providers on consecutive requests merely because a backend instance restarts or one lookup briefly fails. Decide whether provider selection is tied to a release cohort, a persisted assignment, or another explicit mechanism, and test the same user across application instances.

If the new provider is unavailable, follow the documented baseline policy. Automatically falling back to a different allocator can change the treatment mid-experiment. For an operational flag, a baseline response may be acceptable; for a running experiment, document how such cases are detected and handled in the analysis. Availability and causal continuity are separate requirements.

5. Retire old flag paths after the evidence holds

Remove dependencies in the right order

Search the application and worker code for old flag keys, experiment names, wrappers, environment variables, and tracking calls. Include mobile or other long-lived clients that may still depend on the incumbent configuration. Removing a server-side path does not remove a released client from the installed base.

Preserve records needed to interpret historical decisions. Archive the old contract, experiment configuration, result, and assignment metadata before cleaning up operational resources. Distinguish removing active evaluation from deleting historical analytics data; they are different actions with different retention requirements.

If PostHog still powers analytics or replay, keep that integration and remove only the migrated responsibilities. Verify that shared initialization or event code is not accidentally removed with the flag wrapper. Run the relevant user flows again after cleanup.

Cutover gate	Required evidence
Configuration	Flag types, defaults, targeting, and environment match the reviewed contract
Application	Both paths and provider fallback work for representative identities
Exposure	Real served experiences reach the warehouse with correct phase and variation IDs
Metrics	Known-case counts, windows, and denominators reconcile
Experiment health	A/A and data checks complete; unresolved issues investigated
Recovery	Scoped rollback tested, with state and in-flight work considered
Cleanup	Old references removed only after supported clients and retained products are accounted for

Review the operating cost of the final architecture

Compare current PostHog pricing and retained product usage with GrowthBook's current plan, required governance, warehouse compute, integration maintenance, and the period when both systems run. Do not assume the migration eliminates charges for capabilities you continue to use.

PostHog remains a reasonable choice when its integrated product workflows fit your team and your current implementation is serving the decision well. A migration is worthwhile when the target architecture solves a specific constraint and your team can validate the changed contracts. A new dashboard alone is a weak reason to accept assignment or data risk.

Begin with GrowthBook's guided setup and a bounded staging capability. For a complex identity graph, account-level design, or experiment-history import, bring the manifest and acceptance criteria to a technical demo. That makes the conversation concrete: one flag, one exposure path, one set of outcomes, and evidence that the next step is ready.

Start with one verified migration

Set up GrowthBook for a staging capability, confirm the identity and exposure contracts, and expand only after the application and metric checks pass.

Get Started With GrowthBook

Example H2

See All Articles

Experiments

Feature Flags

CLI vs MCP: TLDR, it doesn’t matter as much as you think

Oct 21, 2026

min read

Your agent should know how to do the job whichever interface it uses.

Your coding agent can already run terminal commands. Your product manager's assistant can connect to remote tools. Both need to inspect the same experiment, understand the same metric, and prepare changes to the same feature flag.

That creates a familiar question: should you give them a CLI or an MCP server?

The usual answers mix several concerns. A CLI is easy to compose with scripts. MCP gives compatible clients a standard way to discover and invoke tools. Skills can supply the instructions that make either route useful. None of those statements tells you whether the agent knows your release process or can distinguish a draft from a live change.

For GrowthBook, the choice has become smaller. We've built the agent layer around a REST API and reusable workflow skills, then made those available through different interfaces. Your choice of CLI or MCP no longer needs to determine which product workflow your agent can understand.

It still determines how the agent connects, executes requests, and handles credentials. This article explains that distinction, shows the same disabled-flag operation through both routes, and gives you a way to evaluate the integration you actually intend to use.

The debate combines three different questions

Start with a specific job: prepare a release control for a new checkout flow. The agent needs to know which project owns checkout, which environments exist, what the default behavior should be, and whether the proposed flag key already exists.

Choosing an interface answers only part of that job.

How can the agent reach the product?

A command-line interface lets an agent invoke a program in an environment where it has shell access. That program can make authenticated API requests and return data. You can also run the same commands yourself, put them in scripts, and preserve output for inspection.

The Model Context Protocol architecture connects a host application to servers through clients. An MCP server exposes capabilities that a compatible host can discover and call. The host might be a coding environment, a desktop assistant, or another agent application.

These categories overlap. A terminal-based agent can also be an MCP client. A local MCP server can be a process launched by that client. A remote MCP connection can work without the host giving the model a general-purpose shell.

So “terminal agent” doesn't automatically mean “use the CLI,” and “MCP” doesn't automatically mean “remote service.” Ask what access the runtime provides before choosing the adapter.

What procedure should the agent follow?

An API can expose a create operation without teaching an agent when to use it. A syntactically valid feature flag might still have the wrong project, an ambiguous key, or a default value that doesn't match the application.

That's the job of a skill: reusable instructions for a recognizable task. The Agent Skills format gives those instructions a portable structure, including a skill entry point and supporting references. A workflow can describe the context to inspect, the sequence to follow, and the evidence to return.

For the checkout example, useful guidance says to resolve environments, check the permanent key, choose a value type, and create the flag disabled. Merely exposing a create command leaves those decisions to the agent's general knowledge.

Skills are also useful when the right answer is to take no action yet. “Create a flag” and “design an experiment” are related requests, but they can require different objects and sequencing. An agent needs enough product guidance to make that distinction.

What context makes the procedure relevant?

Even a good playbook needs the current product state. The agent should inspect the objects in your GrowthBook organization and, when it is changing application code, the repository's SDK initialization and tests.

Those are different sources of context. GrowthBook supplies flags, experiments, metric definitions, and other authorized product resources. Your coding runtime supplies the application files it can access. An MCP connection doesn't by itself grant repository access, and a CLI doesn't by itself explain your metric definitions.

Evaluate the combination: access to the product, a maintained procedure, and enough context to apply that procedure. A comparison that changes all three at once cannot isolate the effect of CLI versus MCP.

GrowthBook puts the durable parts below the interface

We separated the agent layer into responsibilities that can be reused. The product API exposes operations. Skills explain workflows. The interface connects the agent to those two layers.

This is the architectural shift behind GrowthBook's agent experience. Instead of making every client integration carry its own interpretation of experimentation, we can share the interpretation and change how it is delivered.

The API provides the operations

GrowthBook's REST API makes product objects addressable. An agent can inspect a project, retrieve a flag, or perform an authorized operation against those objects.

The rebuilt API described in GrowthBook 4.4 uses shared schemas to support types, validation, and API documentation. The CLI is generated from the API definition. That reduces the amount of separately maintained command logic needed to expose product operations.

The OpenAPI specification explains the underlying principle: a machine-readable description can represent an HTTP API's paths, operations, inputs, and responses. Generation is useful because it derives another interface from that contract.

Generation doesn't mean any installed binary knows every future endpoint. The generated client still has a version, and your GrowthBook server still has a version. A self-hosted instance can lag behind the public documentation. Treat API compatibility as a deployment fact you verify.

Skills provide the workflow guidance

GrowthBook's feature-flags skill routes requests to more specific workflows. Creating a flag, inspecting dependencies, changing targeting, requesting review, and publishing a revision are separate procedures.

That structure lets the agent load the guidance relevant to the current job. The skill authoring guidance on progressive disclosure supports this approach: keep the entry point focused and place detail in references that are loaded as needed.

The benefit is maintainability as well as context management. A correction to a flag workflow can be made in the maintained skill rather than repeated across a series of tool descriptions and client-specific prompts.

There are two shell concepts to keep distinct. The generated growthbook CLI is a terminal interface to the product API. The skills repository also includes a small gb-call helper for making REST requests. Installing a skill doesn't make that helper a globally available command; its location depends on the installation. Both can reach the API, but their configuration and invocation are different.

MCP provides a thin bridge

The current GrowthBook MCP implementation exposes four tools: growthbook_list_skills, growthbook_read_skill, growthbook_api_read, and growthbook_api_write.

An agent discovers a domain, reads the relevant workflow, then makes API calls. It doesn't need a dedicated MCP tool for every product endpoint. Calls described as gb-call GET in a skill map to the read tool; mutating methods map to the write tool with the same path and body.

The bridge does not run the shell helper. It forwards authenticated requests. Product-side permissions and validation remain important because a small tool surface can still reach many operations.

Shared guidance also has a delivery boundary: the MCP server bundles skills when it is built, using a pinned source revision. A newly edited skill in the repository needs to be included in a subsequent server build before that server supplies it. “Shared source” is a maintainability claim, not a promise of instant synchronization.

Explore the shared flag workflows

Open the feature-flags skill to see the procedures that guide creation, targeting, reviews, and publishing. Bring the same workflow knowledge to the agent interface your team uses.

View the Flag Skill

The same disabled flag can use either route

The practical question is whether the interface changes the job. Let's create a Boolean flag called checkout-redesign, with a false default and every environment disabled.

This is an API example for a sandbox project, not an application integration tutorial. It creates a control object; your application still needs a correctly configured GrowthBook SDK to evaluate it. The feature flag documentation covers the product concepts behind that integration.

Connect and verify before preparing the write

For the terminal route, install the current GrowthBook CLI, configure a Personal Access Token through its interactive setup, and inspect the available command syntax:

npm install -g growthbook
growthbook configure
growthbook whoami
growthbook features create --help
growthbook features get --help

Check that the active server and identity are the ones you intend to use. Prefer a tested, pinned CLI version in repeatable automation. Don't include a live token in a command you plan to share.

For an MCP route, the GrowthBook Cloud setup uses the hosted endpoint and browser OAuth. In a client using the documented mcpServers configuration shape, such as Cursor, add:

{
  "mcpServers": {
    "growthbook": {
      "url": "https://mcp.growthbook.io/mcp"
    }
  }
}

Complete the browser authorization and confirm tool discovery. Other clients can use a different configuration shape. Self-hosted GrowthBook users should follow the local stdio or self-hosted HTTP instructions rather than point a private instance at the Cloud service.

Next, let the connected agent discover the flag-creation workflow. These are tool-call parameters, shown separately in the order to invoke them:

{"name":"growthbook_list_skills","arguments":{}}

{"name":"growthbook_read_skill","arguments":{"name":"feature-flags"}}

{"name":"growthbook_read_skill","arguments":{"name":"feature-flags/references/flag-create"}}

The returned guidance should describe the prerequisites for creating the flag. If the workflow isn't present, resolve the server's bundled skill version before substituting instructions from an older integration.

Resolve real IDs and construct one payload

The flag-creation workflow checks existing keys, lists projects, and resolves environments. In the MCP client, invoke the read tool with each of these argument objects:

For the CLI route, perform that discovery with the corresponding command groups documented in the CLI command catalog:

growthbook features list-keys --output-format json
growthbook projects list --output-format json
growthbook environments list --output-format json

Inspect each command's help if your installed version differs. The returned keys, project IDs, and environment IDs are the inputs to the payload below. In MCP, the equivalent read-tool arguments are:

{"path":"/api/v2/feature-keys"}

{"path":"/api/v1/projects"}

{"path":"/api/v1/environments"}

Verify that checkout-redesign isn't taken, identify the sandbox project's ID, and enumerate every actual environment. These read responses determine the payload; don't treat example environment names as a complete inventory of your organization.

Save the following as flag.json, replacing <PROJECT_ID> and the environment keys with the discovered values. This example assumes the only environments are staging and production:

{
  "id": "checkout-redesign",
  "valueType": "boolean",
  "defaultValue": "false",
  "description": "Control exposure to the redesigned checkout",
  "project": "<PROJECT_ID>",
  "environments": {
    "staging": { "enabled": false },
    "production": { "enabled": false }
  }
}

The API represents defaultValue as a string, even for a Boolean flag. Explicitly disabling each environment avoids relying on the organization's default state for newly created flags. Use a PAT for this example so the caller has a user identity for attribution.

Review the permanent key, project, value, and disabled states before submitting. If the intended job is an experiment, route to experiment design first instead of creating an unrelated flag and hoping to connect it later.

Submit once, then read the object back

With the CLI, the create command accepts a JSON body. In a Bash-compatible shell, inspect the request with dry run before submitting:

growthbook features create --body "$(cat flag.json)" --dry-run
growthbook features create --body "$(cat flag.json)" --output-format json

The first command previews a request; it doesn't prove the server will accept it. Run the second only after checking the preview and approving this creation. In PowerShell, use that shell's file-reading syntax rather than paste the Bash substitution unchanged.

The PowerShell equivalent, using the same reviewed file, is:

$flagPayload = Get-Content -LiteralPath .\flag.json -Raw
growthbook features create --body $flagPayload --dry-run
growthbook features create --body $flagPayload --output-format json

The same distinction applies: preview first, approve the specific creation, then submit. Neither command changes the application's deployed code.

Through MCP, call growthbook_api_write with the same payload serialized into its body string. Substitute the same resolved project and environment names:

{
  "method": "POST",
  "path": "/api/v2/features",
  "body": "{\"id\":\"checkout-redesign\",\"valueType\":\"boolean\",\"defaultValue\":\"false\",\"description\":\"Control exposure to the redesigned checkout\",\"project\":\"<PROJECT_ID>\",\"environments\":{\"staging\":{\"enabled\":false},\"production\":{\"enabled\":false}}}"
}

Choose one creation route. Don't run both against the same key; they target the same product object, and the second creation should encounter an existing key.

Verify through the CLI get command:

growthbook features get --id checkout-redesign --output-format json

Or invoke growthbook_api_read with:

{"path":"/api/v2/features/checkout-redesign"}

Check the returned key, project, value type, default value, disabled state of every environment, and absence of unintended rules. Report those facts rather than “success.” No targeting or exposure change is included in this example.

Choose the route that fits the runtime

Once the procedure and API operation are shared, the remaining choice is operational. Where does the agent run? What can it execute? How does it authenticate? What does your team need to inspect afterward?

Shell access favors terminal composition

A shell-capable agent can run the CLI alongside repository tools, save structured output, and combine product inspection with local checks. That is useful when the job includes code changes, generated files, or a reproducible script.

For example, a flag audit can produce an inventory file while an engineer separately checks code references. The agent should preserve the association between the GrowthBook key and each repository reference. The final artifact can include both the product state and the proposed code diff.

The shell makes composition available; it doesn't make a workflow safe by default. A script can select the wrong profile or swallow an error. Decide which failures stop the workflow, where outputs are stored, and which steps need a reviewable artifact.

Remote tool access favors MCP clients

MCP is useful when the host already provides a tool connection but doesn't expose a terminal. A PM can inspect an experiment through a compatible assistant without installing a product CLI inside that assistant's runtime.

The MCP authorization specification defines an authorization framework for HTTP transports. GrowthBook Cloud uses browser OAuth, while local stdio configurations use credentials supplied to the process. Those are distinct deployment paths.

MCP also gives hosts a structured description of tools. The tool specification includes annotations such as read-only and destructive hints. Hosts can use those signals in their interaction design, but the signals are not an access-control boundary. The product must still enforce the caller's actual permissions.

Use a small decision table

Your immediate requirement	Route to consider	What to verify
Combine product operations with local files and scripts	CLI plus maintained skills	Shell access, profile, installed version, structured output
Connect a compatible assistant without a shell	Hosted MCP	HTTP/OAuth support, identity, tools, bundled workflow
Keep requests inside a self-hosted deployment	CLI or self-hosted MCP	Network reachability, server version, credentials, trusted app origin
Reuse a procedure across different team clients	Shared skills through either route	Workflow revision, API compatibility, expected review artifact

You can support more than one row. An engineer's CLI and a PM's MCP connection can operate on the same objects with different identities and permissions. Standardizing the workflow is often more useful than standardizing everyone's client.

Compare actual outcomes, not tool counts

Four MCP tools can address a large API. A large CLI command catalog can expose the same operations. Counting entries tells you little about whether an agent completed the intended job correctly.

Likewise, there is no universal token or speed result implied by the architecture. The client, model, instructions, response size, pagination, and number of requests all affect a real run. Measure the work you need instead of importing a result from an unrelated integration.

Keep the evaluation task constant

Use one sandbox project and one bounded task. For a read test, ask each route to find a named flag and identify its enabled environments, current revision, and relevant dependencies. For a write test, use separate temporary keys with the same reviewed disabled-flag requirements.

Give both runs the same expected output. Record whether each identified the correct objects, loaded the relevant workflow, handled missing data, and preserved the boundary between creation and exposure. Measure elapsed time and context consumption only alongside those correctness checks.

If one run gets rich workflow guidance and the other gets only a command name, you're testing the guidance as well as the interface. That can still be a useful product evaluation, but describe what changed.

Test failure behavior deliberately

Try a nonexistent object, an existing key, and a credential without the required permission. The agent should identify the failure and stop the dependent action. It should not invent an ID, silently pick another environment, or reinterpret a permission error as permission to retry with a broader credential.

The revision workflow adds another important test: multiple open drafts. When a chain of changes belongs in one revision, the agent must preserve its version rather than repeatedly ask for whichever draft is latest.

A creation receipt and a publication receipt are different artifacts. If your evaluation only checks that a request returned without error, it can miss the point where a proposed change became live or where a draft failed to include the intended edits.

Record the versions behind the result

Capture the CLI version, server version when available, client configuration, and the skill text or revision used. With MCP, inspect the guidance actually returned by the connected server. With installed skills, inspect the local instructions actually loaded.

That record makes a later regression diagnosable. A changed result may come from new guidance, a client update, a changed API, or different product state. “MCP became worse” is too broad to help you fix it.

Build workflows that survive the next client

The most useful consequence of GrowthBook's design is organizational. Engineers, PMs, and analysts can choose the agent surface that fits their work while referring to the same flags, experiments, and metrics.

An engineer can prepare a disabled flag from a terminal. A reviewer can inspect its draft in the app. An assistant connected through MCP can retrieve the same object and explain its state. The handoff should carry an ID, a revision where applicable, the proposed behavior, and the decision still required.

This is where the CLI-versus-MCP question loses much of its importance. Your release procedure doesn't need to be rewritten around each client. The product holds state; the API exposes operations; skills explain the work; the chosen interface delivers access.

Start with one inspection or sandbox creation. Require the agent to return a concrete object and its verified state. Then add the next workflow only when the current one is reliable enough for your team to review.

GrowthBook's feature flag platform gives that work a common place to live. The durable choice is how you connect release control and measurement to your team's process. CLI and MCP are ways to carry that process into the tools people already use.

Start with a common foundation

Set up GrowthBook for the flags, experiments, and metrics your agents will share. Then choose the CLI or MCP connection that fits each team member's work.

Get Started With GrowthBook

Experiments

Analytics

Standard deviation vs variance: Key differences for A/B testers

Oct 21, 2026

min read

A metric can have wide user-to-user variation and still yield a precise experiment estimate.

That is why confusing standard deviation with standard error creates bad A/B testing decisions. A dashboard may show that users differ substantially while the average effect is estimated closely enough to guide a release. Those statements can both be true.

Variance and standard deviation describe spread. Variance works on a squared scale; standard deviation takes its square root to return to the metric's original units. Neither is automatically the uncertainty around an experiment result.

For A/B testers, the useful distinction is practical: use standard deviation to communicate outcome variability, use variance in the calculations that combine uncertainty, and inspect the standard error or interval for the treatment effect. Start by making sure all three describe the correct unit of analysis.

Understand the difference in units

Variance uses squared distances

Outcomes above and below a mean have deviations with opposite signs. Simply averaging those signed deviations gives zero. Variance avoids that cancellation by squaring each deviation before combining them.

The sample variance formula is s² = Σ(xᵢ − x̄)² / (n − 1), where x̄ is the sample mean and n is the number of observations. The symbol Σ means to add the terms across observations.

If the outcome is measured in seconds, variance is measured in seconds squared. That is mathematically useful but less intuitive for explaining a product experience. “Variance is 4 seconds squared” does not mean users differ by 4 seconds.

Standard deviation returns to the original scale

Standard deviation is s = √s². A variance of 4 seconds squared corresponds to a standard deviation of 2 seconds. Both describe the same dispersion, expressed on different scales.

Do not define standard deviation as the average absolute distance from the mean. That is a different measure. Squaring deviations gives large deviations more influence, and taking the square root afterward does not undo that weighting.

Quantity	What it describes	Units for task time
Variance	Squared dispersion in outcomes	Seconds squared
Standard deviation	Outcome dispersion on the original scale	Seconds
Standard error of the mean	Sampling variability of an estimated mean	Seconds

Neither variance nor standard deviation requires a normally distributed population merely to be calculated. Normality matters for particular interpretations and inferential procedures, not for the existence of these summaries when the required moments are finite.

Calculate both from the same observations

Work through a small dataset

Use four synthetic task times: 2, 4, 4, and 6 seconds. This tiny dataset is for arithmetic, not evidence about a real product or a basis for confident inference.

The mean is (2 + 4 + 4 + 6) / 4 = 4 seconds. Deviations from the mean are −2, 0, 0, and 2 seconds. Their squares are 4, 0, 0, and 4 seconds squared, totaling 8.

Treating these observations as a sample, the usual sample variance is 8 / (4 − 1) = 8/3, or about 2.667 seconds squared. The sample standard deviation is √(8/3), or about 1.633 seconds.

This is a useful check on a report. If someone describes the same sample's variance as 1.633, they have probably substituted the standard deviation. If the units are missing, the mistake is much harder to spot.

Choose sample or population formulas deliberately

If those four observations constitute the entire finite population you want to describe, population variance divides by 4: 8/4 = 2 seconds squared. Population standard deviation is √2, or about 1.414 seconds.

The population and sample distinction depends on what you are estimating. Having every recorded event from an experiment does not necessarily make those events the full population of future users or future conditions you care about.

The n − 1 denominator makes the usual sample variance an unbiased estimator of population variance under independent, identically distributed sampling with finite variance. It does not make the square root exactly unbiased for population standard deviation. Nor does it correct biased sampling or dependence between observations.

Verify the calculation with explicit function names

Python's statistics functions distinguish sample calculations from population calculations. With Python 3 installed, this standalone example requires no external packages or data:

from statistics import variance, stdev, pvariance, pstdev

times = [2, 4, 4, 6]
print(round(variance(times), 3), round(stdev(times), 3))
print(round(pvariance(times), 3), round(pstdev(times), 3))

The first line of output is 2.667 1.633; the second is 2 1.414. The different denominators explain the difference. Before comparing statistics from two tools, check their sample/population defaults and handling of missing values.

Find the noise you can reduce

Explore practical variance reduction methods, including metric design and CUPED, and the conditions that make each approach useful.

Read the Variance Reduction Guide

Use standard errors for experiment uncertainty

Distinguish users from estimated means

Standard deviation describes how users' outcomes vary. Standard error describes how an estimator would vary across repeated samples. For an independent sample, the estimated standard error of its mean is s / √n.

Suppose a user-level task-time metric has a sample standard deviation of 20 seconds among 10,000 independent users. The estimated standard error of the mean is 20/√10,000 = 0.2 seconds. Large differences between users coexist with a much smaller uncertainty around their average.

Collecting four times as many independent users would halve this standard error if the outcome variance stayed similar. It would not mechanically halve user-level standard deviation. The individual experience remains variable even as the estimated mean becomes more precise.

Combine arm variances before taking the square root

For independent treatment and control groups, a simple unadjusted difference in means has estimated variance s²ₜ/nₜ + s²꜀/n꜀. Its standard error is the square root of that sum. You do not add the two arm standard deviations and call the result the experiment's uncertainty.

Take a separate synthetic summary with 10,000 independent users in each arm and a 20-second standard deviation in both. The variance of the estimated difference is 400/10,000 + 400/10,000 = 0.08 seconds squared. Its standard error is √0.08, about 0.283 seconds.

If the observed treatment-minus-control difference is −1 second, an illustrative large-sample, unadjusted 95% normal interval is −1 ± 1.96 × 0.283, approximately −1.554 to −0.446 seconds. Negative values mean faster completion under this metric definition.

This example assumes independent groups and an appropriate large-sample approximation. It is not the exact output of GrowthBook's relative-lift, sequential, covariate-adjusted, or ratio-metric methods. GrowthBook's statistical details describe how the platform estimates effects and their uncertainty.

Avoid interpreting intervals as outcome ranges

That interval concerns the mean treatment effect. It does not say that 95% of individual users improved by between 0.446 and 1.554 seconds. An individual user's causal effect is a different quantity, and the experiment does not observe both potential outcomes for the same person.

Likewise, mean ± one standard deviation is not a universal statement about where 68% of users fall. That familiar interpretation depends on an approximately normal distribution. Skewed task times or revenue can look very different.

The frequentist confidence level describes the repeated-sampling coverage of the interval procedure under its assumptions. It is not a posterior probability assigned to a fixed realized interval.

Check what your metric actually measures

Match the variance to the randomization unit

If users are randomized and each contributes many requests, request rows are not automatically independent experimental units. Treating 100 requests from one person as 100 unrelated users can make uncertainty look smaller than it should.

Define the outcome at the intended unit or use a method that accounts for clustering. GrowthBook's aggregate data requirements are relevant when supplying precomputed summaries: the sample size, sum, and dispersion must correspond to the statistic and analysis method.

Two teams can report different standard deviations without either making an arithmetic mistake. One may summarize per-request latency while the other summarizes each user's average latency. They are describing different outcome distributions. Clarify the grain before comparing their numbers.

Paired measurements need another distinction. If the same users provide before-and-after outcomes, the measurements within each pair can be correlated. Analyze each user's difference or use an appropriate paired method; do not apply the independent-arm formula automatically. The paired-means framework treats those differences as the observations of interest. Pairing also does not remove time trends or establish causality by itself; the design still determines which comparison is defensible.

Inspect tails instead of deleting them casually

Large observations influence both variance and standard deviation. A few unusually long tasks can dominate a time metric; a few large purchasers can dominate revenue. Determine whether those values reflect valid behavior, instrumentation defects, retries, or unit errors.

A valid expensive user is not automatically an outlier to discard. Removing high values after seeing an experiment result can change the estimand and introduce selective analysis. Decide justified exclusions or transformations before interpreting the test.

Report complementary summaries when they answer useful questions: a histogram, median, or a relevant tail percentile can show what a single spread measure hides. Those summaries do not replace the mean effect if the mean remains the decision metric.

Keep units consistent across pipelines

Converting seconds to milliseconds multiplies every value by 1,000. Standard deviation also multiplies by 1,000, but variance multiplies by 1,000,000. A dramatic numerical variance increase can therefore be a unit conversion rather than a product regression.

For the four-value example, sample standard deviation becomes approximately 1,633 milliseconds and variance becomes approximately 2,666,667 milliseconds squared. The underlying task times have not changed. Store units with metric definitions and inspect them when comparing dashboards.

A lower variance caused solely by rescaling a metric does not create more statistical information. The effect size rescales too. Separate a numerical change in units from an actual improvement in the precision of your estimate.

Also avoid averaging standard deviations from daily reports and labeling the result a weekly standard deviation. The complete-period spread depends on both variation within days and differences between daily means. Keep the underlying observations or sufficient summaries that support the intended calculation. A simple average of displayed daily values generally loses information needed to reconstruct the full distribution.

Turn spread into a better testing plan

Define the outcome before planning the sample

Start with the metric definition: population, identity, aggregation, window, units, exclusions, and desired effect. Estimate variability from representative historical data at that same grain.

Historical variance is an input, not a guarantee. New traffic, a different user mix, or a changed measurement window can alter it. Explore plausible variance and effect-size assumptions rather than treating one historical number as exact.

Keep the two arm variances separate when appropriate. A treatment can change variability as well as its mean. The simple equal-standard-deviation illustration above is a teaching assumption, not a requirement that production groups have identical spread.

Reduce estimation noise without hiding behavior

CUPED and post-stratification use suitable pre-experiment information and covariates to improve effect estimation. CUPED stands for Controlled-experiment Using Pre-Experiment Data. Its value depends on useful predictive information and a valid implementation.

Adjustment can reduce uncertainty in the treatment effect without making real users' outcomes less variable. That distinction prevents a reporting error: a narrower adjusted interval is not evidence that customers now experience a more consistent product.

Also separate missing data from genuine zeros. A user with no purchase may correctly contribute zero revenue; a missing instrumentation event might represent an unknown outcome. Treating both identically can distort the mean and its variance before any statistical engine sees the data.

Report the right quantity for the decision

Reader's question	Useful quantity
How much do individual outcomes differ?	Standard deviation, with distribution context
What enters this uncertainty calculation?	Variance at the correct analysis unit
How precisely is the effect estimated?	Standard error and an appropriate interval
Is the measured difference worth acting on?	Effect size, uncertainty, guardrails, and decision context

For a clear experiment readout, show the metric's unit and analysis grain, the estimated effect, and its appropriate uncertainty interval. Add outcome spread when it helps readers understand the population or diagnose a noisy metric.

GrowthBook's experimentation platform brings defined metrics and statistical analysis together. A trustworthy interpretation still starts with your measurement choices: describe spread with standard deviation, calculate with variance where required, and judge precision using the estimator's uncertainty.

Build a clearer experiment readout

Connect your metrics to GrowthBook and evaluate effects with the analysis method and measurement definitions your team has chosen.

Get Started With Experimentation

Experiments

Analytics

How to calculate true positive rate (TPR)

Oct 21, 2026

min read

Catching 80 defects means little until you know how many defects existed.

Suppose a detector flags 125 suspicious software changes. Engineers inspect them and confirm 80 real defects. That tells you something about alert quality, but it does not tell you how many defects the detector missed. To calculate true positive rate, you need the actual positives outside the alert queue too.

True positive rate, or TPR, measures the share of actual positive cases that a system correctly identifies. It is also called recall or sensitivity. The formula is TP / (TP + FN), where TP means true positives and FN means false negatives.

The arithmetic is simple. Choosing the right population, labels, threshold, and denominator is the difficult part. This guide works through an engineering example, distinguishes neighboring metrics, and explains what changes when the “test” is an A/B experiment rather than an item classifier.

Start with actual positives, not alerts

Define the positive class

“Positive” means the condition you want to detect. It does not necessarily mean a good outcome. In a defect detector, positive means a real defect. In a spam filter, it means spam. Write that definition before calculating anything.

A true positive is an item that is actually positive and predicted positive. A false negative is actually positive but predicted negative. Those two groups together form the denominator: all actual positives in the evaluation sample.

The recall definition is therefore conditional on actual class membership. TPR asks, “Among the real defects, what proportion did we catch?” It does not ask, “Among the alerts, what proportion were real?”

Fix the unit and observation window

Choose a consistent evaluation unit. A code change, an individual finding, and a repository are different units. If one change generates 12 duplicate alerts, counting alerts in the numerator and changes in the denominator produces a meaningless rate.

For the example, each unit is one code change, evaluated once by one fixed detector version at one fixed threshold. An independent review process eventually labels each change as defective or not defective. The observation window is long enough for those labels to mature.

This resembles the discipline needed when defining experiment metrics: the event, identity, time window, and denominator must describe the same question. A correct formula applied to incompatible counts still gives the wrong answer.

Calculate TPR from the confusion matrix

Put every evaluated item in one cell

Consider an original synthetic evaluation of 1,000 code changes. Reviewers establish that 100 contain a defect and 900 do not. The detector catches 80 defects, misses 20, and incorrectly flags 45 nondefective changes.

Actual label	Predicted positive	Predicted negative	Total
Defective	80 true positives	20 false negatives	100
Not defective	45 false positives	855 true negatives	900
Total	125	875	1,000

Read the labels rather than relying on cell position. Libraries and reports can arrange axes differently. The scikit-learn confusion matrix convention uses actual classes on rows and predicted classes on columns, with class order also affecting where each value appears.

Check the totals before calculating rates. The four cells sum to 1,000. True positives plus false negatives sum to 100 actual defects. Predicted positives total 125, which is a different denominator.

Divide detected positives by all actual positives

The calculation is:

TPR = TP / (TP + FN) = 80 / (80 + 20) = 80 / 100 = 0.80 = 80%.

Interpret it in words: the detector caught 80% of the actual defects in this evaluation sample. It missed 20%, so the false negative rate is 20%. For a binary task with complete labels, TPR and false negative rate sum to 100%.

Do not divide 80 by 1,000. That would measure true positives as a share of all evaluated changes, not recall. Do not divide 80 by 125 either. That is precision, which answers the alert-quality question.

The result describes this sample and threshold. It does not establish that every future workload will have the same recall, or that all defect categories are equally detectable. Keep the counts beside the percentage so readers can see the support behind it.

Handle zero positive support explicitly

If the sample contains no actual defects, TP + FN equals zero. Sample TPR is undefined. Returning 0% would imply the detector missed known defects; returning 100% would imply it successfully found them. Neither interpretation is supported.

Software libraries need a practical zero-division convention. That convention can return a placeholder or warning, but it does not create evidence. Report “TPR unavailable: zero actual positives” and preserve the support count. A pipeline can then distinguish an unevaluable slice from measured poor performance.

Check the experiment error risks

See how power, repeated checks, and multiple comparisons affect the reliability of experimental decisions beyond a single classifier score.

Read the False Positives Guide

Compare TPR with precision and false positives

Keep the denominators visible

For the same 1,000 changes, the other common metrics tell different stories:

Metric	Calculation	Result
True positive rate / recall	80 / 100 actual positives	80%
Precision	80 / 125 predicted positives	64%
False positive rate	45 / 900 actual negatives	5%
Specificity	855 / 900 actual negatives	95%
Accuracy	(80 + 855) / 1,000 changes	93.5%

The classification metric definitions differ because they condition on different populations. Here, 36% of alerts are false, but the false positive rate is 5%. Calling both numbers “the false positive rate” hides the operational distinction.

Accuracy can also look reassuring in an imbalanced population. A detector that labels all 1,000 changes nondefective would be 90% accurate while detecting none of the 100 defects. TPR exposes that failure, though it still needs the false-positive side of the picture.

Choose a threshold for the decision

Suppose a stricter threshold, evaluated on the same labeled changes, catches 60 defects and incorrectly flags 10 nondefective changes. Its TPR is 60%, its precision is about 85.7%, and its false positive rate is about 1.11%.

This threshold reduces the review queue from 125 changes to 70. It also misses 40 defects instead of 20. Whether that tradeoff is acceptable depends on the consequences of a missed defect, the review team's capacity, and what action follows an alert.

A ROC curve plots TPR against false positive rate across thresholds. It helps compare operating points. It does not pick the operational threshold for you, and an area-under-curve summary does not specify the queue size at the threshold you intend to deploy.

If you tune thresholds using the same labels you later report as a final evaluation, the score can be optimistic. Use separate validation and held-out evaluation data, or another defensible evaluation design, and freeze the decision rule before the final assessment.

Specify the action at each threshold too. A low-confidence finding might create a suggestion, while a high-confidence finding might block a merge. Those are different interventions with different error costs. One aggregate recall figure should not imply that every detected item received the same treatment.

Make sure the labels can support the claim

Audit predicted negatives too

If engineers review only alerts, you can estimate precision within the reviewed alert population. You cannot directly count missed defects among unreviewed negatives. Assuming those changes are clean makes FN artificially small and TPR artificially high.

An evaluation may need independent review of a representative sample of predicted negatives, delayed incident labels, or a curated benchmark with known defects. Each approach has limits. Incident labels miss undetected failures; curated examples can differ from ordinary production work.

If the evaluation oversamples suspicious changes or certain repositories, report that design. Use appropriate weighting when estimating performance for the real workload. A simple unweighted confusion matrix describes the sampled dataset, which may not represent the deployed population.

Preserve uncertainty and segment support

An 80% estimate based on 4 detected defects out of 5 is much less precise than one based on 800 out of 1,000. The important sample size for recall is the count of actual positives, not just the total number of changes evaluated.

Report positive support by important slices, such as language, defect category, or repository type. A high pooled TPR can hide weak performance on a rare but expensive failure mode. At the same time, avoid declaring tiny slices definitively better or worse from a few examples.

Counts alone do not make observations independent. Many similar changes from one repository can share a common pattern. Use uncertainty methods that fit the evaluation design rather than assuming every row is an unrelated trial.

Separate label delay from label absence

A recently deployed change might appear nondefective because nobody has had time to find the bug. Treating it as a confirmed negative gives the newest cohort a different labeling standard from older cohorts.

Define when an item becomes eligible for evaluation and track unresolved labels separately. Keep model version, threshold, label version, and observation dates with the result. That makes a later score change interpretable: you can tell whether the detector changed, the workload changed, or the labels matured.

For a warehouse-native measurement workflow, these definitions belong in the underlying data model. Joining model decisions to review outcomes still requires identity and time alignment; a queryable warehouse does not automatically resolve ambiguous ground truth.

Keep comparable populations together

When comparing detector versions, evaluate them on the same held-out cases where possible. Otherwise, an apparent recall improvement could come from an easier mix of defects. Preserve a breakdown of new, removed, and unchanged cases so reviewers can see what changed between evaluations.

When combining reports, add the underlying TP and FN counts for nonoverlapping, consistently defined populations, then calculate the pooled rate. An unweighted average of repository percentages answers a different question because it gives a tiny repository the same influence as a large one. Either summary can be useful, but label the weighting explicitly.

If the same change appears in several reports, deduplicate or model that repeated observation before pooling. Repeated scans do not create additional independent defects. Keep an audit trail from the final counts back to the evaluated item identifiers so a suspicious rate can be investigated.

Connect detection quality to product decisions

Distinguish classifier recall from statistical power

In hypothesis testing, statistical power is the probability that a testing procedure rejects the null when a specified alternative is true. It is a repeated-test detection probability, depending on assumptions such as effect size, variance, sample size, and decision rule.

That is related to the idea of detecting a real signal, but it is not the same dataset as a labeled-item confusion matrix. In an ordinary A/B program, you do not observe the true causal effect of every experiment with certainty. Dividing declared winners by all experiments does not reveal a ground-truth TPR.

Power planning asks how often your procedure would detect a particular real effect under its assumptions. A classifier evaluation asks how many labeled positives were detected in the evaluated sample. Keep those questions separate when discussing GrowthBook's statistical methods.

Evaluate the intervention as well as the detector

An improved offline detector can still create a worse product workflow. More alerts might overwhelm engineers, delay releases, or catch defects too late to matter. Conversely, a modest recall change might be valuable if it catches the failures with the highest operational cost.

When appropriate, use randomized experiments to evaluate the downstream intervention: which review experience helps teams resolve defects, with what delay and cost? Keep detector recall, review workload, and product outcomes as separate measures.

GrowthBook's experimentation platform can analyze defined outcomes from such tests. Your team supplies reliable assignment, exposure, labels, and metric definitions. Neither a high TPR nor a significant product result eliminates the need to inspect the other evidence.

Use a repeatable reporting checklist

Before sharing a TPR number, state the positive class, evaluation unit, detector version, threshold, and label process. Show TP and FN beside the rate. Include precision or false positive rate so readers can assess the workload created by detection.

Then ask whether predicted negatives were evaluated, whether labels had time to mature, and whether the sample reflects the population where you will act. Mark zero-support slices unavailable and explain material gaps in coverage.

The useful result is a statement someone can audit: “At this threshold, the detector found 80 of 100 labeled defects in these 1,000 evaluated changes, while producing 45 false alerts.” That is a clearer basis for a decision than “our model is 80% accurate.”

Measure what the change delivers

Connect your data and defined metrics to GrowthBook to evaluate how a product or workflow change affects the outcomes your team cares about.

Start Measuring With GrowthBook

Ready to ship faster?

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

Get Started

Book a Demo

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.

Replacing a flag SDK can change who sees a feature. Replacing an analysis pipeline can change which outcomes count. A migration has to control both.

1. Define what moves and freeze a migration contract

Separate assignment, analysis, and product analytics

Inventory behavior, not only flag names

Save an executable mapping fixture

2. Make identities and warehouse records agree

Choose the identifier before copying targeting rules

Define a canonical exposure record

Preserve history without pretending the allocator moved

Prove the data path before importing history

Check the migration architecture

3. Integrate one GrowthBook flag in staging

Create the target with a tested baseline

Install and run the staging smoke check

Wire exposure delivery before measuring a real experiment

4. Validate new experiments and cut over deliberately

Compare contracts before comparing estimates

Run a new A/A test and investigate unhealthy data

Cut over one authoritative path

5. Retire old flag paths after the evidence holds

Remove dependencies in the right order

Review the operating cost of the final architecture

Start with one verified migration

Table of Contents

Related Articles

CLI vs MCP: TLDR, it doesn’t matter as much as you think

Your agent should know how to do the job whichever interface it uses.

The debate combines three different questions

How can the agent reach the product?

What procedure should the agent follow?

What context makes the procedure relevant?

GrowthBook puts the durable parts below the interface

The API provides the operations

Skills provide the workflow guidance

MCP provides a thin bridge

Explore the shared flag workflows

The same disabled flag can use either route

Connect and verify before preparing the write

Resolve real IDs and construct one payload

Submit once, then read the object back

Choose the route that fits the runtime

Shell access favors terminal composition

Remote tool access favors MCP clients

Use a small decision table

Compare actual outcomes, not tool counts

Keep the evaluation task constant

Test failure behavior deliberately

Record the versions behind the result

Build workflows that survive the next client

Start with a common foundation

Standard deviation vs variance: Key differences for A/B testers

A metric can have wide user-to-user variation and still yield a precise experiment estimate.

Understand the difference in units

Variance uses squared distances

Standard deviation returns to the original scale

Calculate both from the same observations

Work through a small dataset

Choose sample or population formulas deliberately

Verify the calculation with explicit function names

Find the noise you can reduce

Use standard errors for experiment uncertainty

Distinguish users from estimated means

Combine arm variances before taking the square root

Avoid interpreting intervals as outcome ranges

Check what your metric actually measures

Match the variance to the randomization unit

Inspect tails instead of deleting them casually

Keep units consistent across pipelines

Turn spread into a better testing plan

Define the outcome before planning the sample

Reduce estimation noise without hiding behavior

Report the right quantity for the decision

Build a clearer experiment readout

How to calculate true positive rate (TPR)

Catching 80 defects means little until you know how many defects existed.

Start with actual positives, not alerts

Define the positive class

Fix the unit and observation window

Calculate TPR from the confusion matrix

Put every evaluated item in one cell